gms | German Medical Science

Reports that many people with hearing impairment are not fitted with hearing aids are numerous. Even a search for the actual number of people with hearing impairment results in very different findings. For example, the German federal association of hearing aid acousticians (Bundesinnung der Hörakustiker KdöR; biha) gave the number of people “with an indicated hearing impairment” in Germany as 5.4 million [1]. In von Gablenz et al. [2], on the other hand, the number of hearing-impaired people in Germany was estimated to be 11.1 million. The list of such sources could be extended by several variants. This raises the question of which of these numbers – if any – are “correct”. Against this background, the available epidemiological data on hearing abilities from the HÖRSTAT project, conducted in Oldenburg and Emden, and the Aalen study “Hearing in Germany”, were analyzed in different ways.

The extent of the problem of comparing numbers has been shown, inter alia, in the reports of Shield [3], [4]. A comparison of all studies is almost impossible due to the different criteria defining hearing impairment. For example, in the current overview of Shield [4], adoption rates between 11 and 59% were reported for the UK, depending on the data source. For Germany, the EuroTrak analysis [5], which was the only source for our country included in the report of Shield [4], resulted in an adoption rate of 35%. This rate increased to 37% in the current EuroTrak analysis [6] and was of the same order of magnitude as in France, but lower than in the Scandinavian countries and the UK.

The EuroTrak analysis was based on surveys of an online panel. The information on hearing difficulties was neither verified by pure-tone audiometry nor by speech audiometry and could therefore be biased. However, the measurement of pure-tone audiograms is also no guarantee for comparability. Already in von Gablenz and Holube [7] different criteria had to be used to compare the prevalence of hearing impairment between international studies. Most well-known is the WHO classification [8], which assumes hearing impairment for average hearing thresholds of more than 25 dB HL, but also indicates further levels for subdividing the severity of the hearing impairment. In addition, in the international epidemiological joint project “Global Burden of Disease” (GDB), a limit of a hearing threshold of 35 dB HL was used, since hearing aid benefit was only definitely assumed beyond this threshold [9].

Overall, however, it is questionable whether hearing disabilities can be assessed solely from the pure-tone audiogram. Although the guideline for assistive devices (Hilfsmittelrichtlinie in German) [10] uses a criterion from the pure-tone audiogram, namely the examination of the hearing thresholds at individual frequencies at a threshold of 30 dB HL, it additionally takes speech recognition in quiet into account. In addition, the pure-tone audiogram-based classification tables established by Röser [11] should be supplemented by speech-audiometric data. Assuming that hearing impairment is more likely to impact speech recognition in noise than in quiet and that this is more relevant to everyday problems, Smits et al. [12] provided a screening test for the detection of hearing impairment using digits in noise. In audiological practice, the Göttingen sentence test (GÖSA, [13]) can be used as a more accurate measurement method, for which criteria for the classification of hearing impairment have also been proposed [14].

The possibilities for the classification of hearing impairment are therefore manifold. In order to analyze the effects of the various criteria on the hearing aid adoption rate, the data on hearing abilities from Oldenburg, Emden, and Aalen were analyzed and presented with regard to this question.

For the analysis, the data from the projects HÖRSTAT and “Hearing in Germany” in Aalen was pooled. For a detailed description of the data collection procedures refer to von Gablenz et al. [2] and von Gablenz and Holube [7]. Both databases were collected from random samples of the general population from the age of 18 years upwards. In total, N=3,105 cases with pure-tone audiograms measured on both ears were available. All human studies described were conducted with the approval of the responsible ethics committee, in accordance with national law and the Helsinki Declaration. All the subjects involved signed a declaration of consent.

In HÖRSTAT, the GÖSA in noise [13] was additionally conducted on both ears via headphones with a noise level of 65 dB SPL (for 47 participants 75 dB SPL) and the participants were asked about subjective hearing difficulties. This global question was “Which of the following are you suffering from?” and allowed the selection of “hearing difficulties” along with other symptoms (poor eyesight, high blood pressure, back problems, none of these). It was taken from the EuroTrak survey [15]. In relation to the 1,866 participants from HÖRSTAT, which were included in von Gablenz et al. [2], there were no complete or sufficiently reliable results for 56 participants for the GÖSA and self-reported hearing difficulties. Of these, the pure-tone hearing ability was so poor in 15 participants that the GÖSA could not be measured at one or both ears. Measurements of these participants were included in the analysis by setting the hearing loss for speech (see below) to 100% for each ear if the speech test was deemed unfeasible. The reasons for the missing GÖSA results of 40 participants were, amongst other things, of a technical or organizational nature. Therefore, a random lack that was independent of the hearing ability was assumed. The influence of the exclusion of these participants on the prevalence data was assumed to be negligible. For one participant, moreover, the response to the global question about hearing difficulties is missing. Criteria based on the GÖSA and the global question were therefore limited to N=1,825 cases.

In Table 1 [Tab. 1], the numbers of participants by gender and by age groups 18 to 24 years, 25 to 34 years, etc. and over 84 years for the whole group are reported. Also listed is the number of unilateral and bilateral hearing-aid adoptions. Altogether, 196 of the 3,105 participants were fitted with technical hearing devices. These included one participant with a cochlear implant and one participant with a CROS device. For the sake of simplicity, however, these provisions were not further differentiated, but generally subsumed under hearing-aid adoption. Figure 1 [Fig. 1] shows the provisions as a function of the average hearing loss (PTA, mean of 0.5, 1, 2, and 4 kHz) of the better and the worse ear. As expected, unilateral hearing-aid fittings were mainly undertaken in asymmetric hearing impairments, i.e. on the worse ear, with still good hearing abilities of the better ear, and on the better ear when the worse ear was non-fittable.

To compensate for variations in age-related sample sizes from the general population (see Table 1 [Tab. 1]), the results were weighted according to the data of Destatis [16]. For all figures and tables (with the exception of the case numbers in Table 1 [Tab. 1]), the same procedure was used for weighting. The procedure was applied to 3,105 or to 1,825 cases, depending on the question addressed. The age distribution of the sample was adjusted to microcensus data [16] using 5-years bands, and an equal age distribution of men and women was adopted. Similar to von Gablenz et al. [2], the vocational education was adjusted to the distribution of qualifications (unspecified, without professional qualification, vocational training, technical college, university) also based on microcensus data [16] in the age groups 20 to 34, 35 to 64, and above 65 years. Due to the small number of cases, the groups “without professional qualification” and “vocational training” had to be combined for the HÖRSTAT subgroup with N=1,825 for men from the age of 65 years.

Various criteria were used to calculate the prevalence and the adoption rate (see Table 2 [Tab. 2]). Some of the criteria are based on the average hearing loss PTA already described above. In most publications, the criterion WHO1 or the GBD criterion was derived. In order to remain in the WHO classification system and still choose a further limit (from which it can definitely be assumed that there will be a benefit from hearing aids), the second level (WHO2, [8]) was the obvious choice. These PTA criteria were compared, inter alia, with the pure-tone audiometry criterion of the guideline for assistive devices. Unfortunately, data on the Freiburg monosyllable speech test in quiet were not available. In addition, a threshold (FT20) was taken from the percent hearing loss %FT from the 4-frequency table according to Röser [11]. In order to compare the value %FT with the PTA, hearing loss for pure-tones in percent %PTA was calculated based on Röser’s table:

To compare %PTA with a comparable percentage level in speech recognition, the percentage of hearing loss for speech in noise was calculated according to Thiele et al. [14] from the signal-to-noise ratio (SNR) for a speech recognition score of 50% in GÖSA, i.e. the speech recognition threshold (SRT). As thresholds for the SRT, either –4 dB SNR (SRT-4, [14]) or –3 dB SNR (SRT-3, [17]) was used. In addition, the subjective indication of hearing difficulties (SHP, global question) was used as a prevalence criterion. Although the criteria of WHO and GBD internationally only consider the better ear, all criteria except the global question were considered separately for the better and the worse ear.

In order to present the patterns of percentage hearing loss and prevalence over age, results were aggregated into 10-year cohorts and quadratic functions adjusted. Neither the number of cases nor confidence intervals were taken into account. The quadratic functions were visually checked and replaced with linear functions in the younger participant groups with insufficient data fits. The transition point was chosen separately for each function.

The detailed numerical values for percent hearing loss and prevalence in age groups are given in the table in Appendix 1 (Attachment 1 [Attach. 1]). The following graphical representations are more descriptive.

Figure 2 [Fig. 2] shows the percentage of hearing loss from the 4-frequency Röser table (%FT), the mean percentage hearing loss for pure-tones (%PTA), and for speech in noise (%SRT) for the better and the worse ear as a function of age. The two quantities %FT and %PTA calculated from the pure-tone audiogram are very similar, with %PTA increasing slightly more with age than %FT. The slope on the worse ear is similar to the hearing-aid adoption rate. The quantity %SRT calculated from the speech recognition in noise indicates a higher percent hearing loss until the age of about 70 years than for the other criteria in comparison.

In order to determine the most suitable indicator for hearing-aid adoption, ROC curves were calculated for the HÖRSTAT data (see Figure 3 [Fig. 3]). The curves illustrate the sensitivity and the specificity for a unilateral or a bilateral hearing-aid adoption. As indicators, the PTA of the better, the worse and of both ears, as well as the SRT of the better and the worse ear and age were used. The best indicator was the PTA of the worse ear (area under the curve AUC=0.970). Speech recognition in noise showed a lower sensitivity and specificity (AUC=0.929). Pure-tone thresholds and speech recognition become progressively worse with age, but age itself was by comparison the worst predictor for HA use (AUC=0.842).

Hearing-aid users in Oldenburg were fitted with hearing aids at a lower PTA than in Aalen. When choosing Oldenburg as the reference, the odds ratios for Emden were OR=0.7 (0.4-1.2) and for Aalen OR=0.4 (0.3-0.6). Thus, the chance of a hearing-aid fitting in Oldenburg is about two and a half times higher than in Aalen, if the PTA of the worse ear and age are taken into account. With an odds ratio of OR=0.9 (0.6-1.3), gender has no significant influence on hearing-aid adoption.

The percentages of hearing-impaired subjects as a function of age when using the various prevalence criteria are shown in comparison to the percentage of hearing-aid adoptions in Figure 4 [Fig. 4]. The criteria FT20 from Röser’s table and WHO1 were roughly the same. They were well above the other two PTA-based criteria, GBD and WHO2, which came closest to the proportion of hearing-aid adoption and increased approximately similarly with age. By contrast, the criterion of the guideline HMR resulted in a significantly higher prevalence, which roughly corresponds to the prevalence according to the GÖSA SRT-3 criterion. The stricter criterion SRT-4 showed the highest prevalence and led to almost all participants aged 75 years and over. The subjective hearing impairment SHP showed a considerably different slope with a higher prevalence in young and a lower prevalence in older age groups.

Using the various prevalence criteria, it is possible to verify the proportion of participants classified as hearing impaired that have one or two hearing aids. Figure 5 [Fig. 5] shows the percentage of participants with hearing impairment for the better and the worse ear for the different criteria, together with the proportion of hearing-aid adoptions (unilateral or bilateral). The speech-in-noise based SRT-3 and SRT-4 criteria classified most participants as hearing impaired. This proportion was lowest for the pure-tone audiogram criterion WHO2. Thus, the highest adoption rate was observed for this latter criterion. Nevertheless, when limiting to the better ear, a noteworthy number of participants classified as normal hearing were fitted with hearing aids. This applies, albeit to a lesser extent, to the classification with the GBD criterion. When considering the worse ear, the proportion of hearing aid fittings of normal-hearing participants was negligible except for the WHO2 criterion.

As shown in Figure 5 [Fig. 5], about 25% of the participants reported subjective hearing difficulties. Of these, about 25% were fitted with hearing aids. However, these shares were strongly dependent on age. In the age group over 65, both proportions were around 45% (see Figure 6 [Fig. 6]).

In the whole group of those aged 18 and above, the adoption rate was 5.6%, which corresponded to a number of approx. 3.8 million hearing aid users aged 18 and above in Germany. The prevalence rates for the various criteria and the overall adoption rate were used to estimate the adoption rate of persons with hearing impairment (see Table 3 [Tab. 3]). For the different criteria, adoption rates ranged from 8.8% to 63.5%. The adoption rates calculated according to speech recognition in noise were below the adoption rates calculated from the pure-tone audiogram. The adoption rates using the WHO1 criterion and Röser’s table (FT20) were of the same order of magnitude as those based on the subjective global question.

The analysis of the epidemiological data on hearing abilities from Oldenburg, Emden, and Aalen resulted in very different adoption rates depending on the prevalence criterion used and thus reflects the diversity of the study results compiled by Shield [4]. None of the criteria is wrong in itself. On the contrary, this analysis shows that the criterion used must always be given in order to be able to interpret prevalence estimates, or the number of hearing-impaired persons. Nevertheless, the suitability of the various criteria for hearing impairment must be critically discussed.

The use of speech recognition in noise was based on the assumption that this measurement condition is more closely related to the everyday complaints of people with hearing impairment, than are speech recognition in quiet or the pure-tone audiogram [12]. The percentage hearing loss for speech was higher in the analysis than for pure tones for the younger participants. This effect was expected because %SRT should upvalue the impairment for speech in noise for mild pure-tone hearing losses [14]. However, the intersection of the percentage hearing loss curves was relatively high, near 70 years. Based on Thiele et al. [14] and von Gablenz and Holube [17], two SRT criteria differing in the cut-off value for impairment by 1 dB were applied to the data. These two criteria led to a very high prevalence, especially in the older age groups, and low adoption rates. Therefore, the choice of an age-dependent prevalence criterion for speech in noise should be considered [18].

The criterion from the guideline for assistive devices (HMR) also led to an overestimation of the prevalence and thus to an underestimation of the adoption rate, since only the pure-tone audiogram was used and not speech recognition in quiet. Once a frequency in the pure-tone audiogram reached or exceeded a hearing loss of 30 dB HL, the prevalence criterion was met. However, the recognition of monosyllables at a speech level of 65 dB SPL would probably exceed 80% in this border area.

By contrast, the WHO2 criterion, especially when used on the better ear, resulted in significantly lower prevalence and higher adoption rates. Even subjects who were classified as normal hearing according to this criterion were fitted with hearing aids. Therefore, this criterion is not appropriate. This also applies, albeit to a lesser extent, to the GBD criterion.

The FT20 criterion derived from Röser’s table and the WHO1 criterion both resulted in prevalence and adoption rates of 25–30%. This coincidence is due to the calculation of both criteria from the thresholds in the pure-tone audiogram. A hearing loss of 20% according to Röser’s table thus corresponds approximately to an average hearing loss (PTA) of 25 dB HL. The adoption rates of these two criteria were of the same order of magnitude as the rate for the subjective hearing difficulties criterion. The agreement, however, was limited to the numbers summarized across all age groups. The age-related prevalence of SHP showed a very different trend compared to the pure-tone criteria. In the younger groups of participants, hearing difficulties were reported much more frequently, but in the older age groups the SHP prevalence approaches that of the GBD criterion for the worse ear. The reported hearing difficulties in HÖRSTAT were, with approx. 25%, comparable to those in the British biobank data [19], thus differing from the much lower figures of the EuroTrak survey conducted on the internet. The data agree only in the older age groups.

If the analysis of prevalence and adoption rates is limited to the age group over 65 years, and further assuming that only those with subjective hearing difficulties are eligible for hearing-aid fitting, the adoption rate is approx. 45%. The prevalence of subjective hearing difficulties in this age range was most consistent with the pure-tone related GBD criterion of the worse ear. However, the GBD criterion should not be used as the limit for hearing aid fitting, since even participants classified as normal hearing by this criterion owned hearing aids. A stricter, i.e. worse, threshold in the pure-tone audiogram than currently specified in the guideline is also not appropriate, because the hearing-aid users in Oldenburg had lower hearing losses than in Aalen. This may be due to the presence of hearing research institutions in Oldenburg and the associated successful publicity. Finally, a bias of the Oldenburg sample in HÖRSTAT due to these factors cannot be completely ruled out.

Although not all participants with a mild hearing impairment according to the WHO classification reported hearing difficulties, the prevalence criteria FT20 and WHO1 seem to be the most suitable measures for characterizing hearing loss when limiting to pure-tone audiometry. However, Röser’s 4-frequency table is not well known internationally. Therefore, epidemiological studies should at least indicate the prevalence according to the WHO1 criterion, in order to allow international comparisons. From this recommendation, the impracticality of the HMR criterion should not be inferred. The HMR criterion, as mentioned above, is used in clinical practice in combination with speech recognition of monosyllables. Results from the monosyllables speech test are not included in this database and, therefore, HMR as a combination criterion cannot be compared with the other criteria in this analysis.

It should also be noted that the present data were collected in the years 2008–2012 [2]. Since this period, the adoption rates in the EuroTrak survey increased from 32% [15] to 37% [6]. In addition, there have been sociodemographic changes in society since the data collection. Therefore, the prevalence and adoption rates are likely to have changed numerically. However, the relative differences in prevalence criteria remain unaffected.

Finally, it should be noted that the reported percentage of hearing aid users of 5.6% (4.7–6.6, 95% confidence interval) for the total sample differs from the 5.2% value given in von Gablenz et al. [2]. This difference is due to the different weighting of the data. For the extrapolations in von Gablenz et al. [2], the data for the population projection for 2015 were used and the different age distribution of male and female was retained. Furthermore, combining additional variables from the Aalen dataset, including usage time, and their critical review, revealed that 12 other individuals had not selected the option hearing-aid user but pointed out information on specific questions in this context. The proportion of hearing-aid users reported in von Gablenz et al. [2] was therefore somewhat underestimated, but lies within the 95% confidence interval of the present estimate.

Overall, the present contribution extends the adoption rates reported in von Gablenz et al. [2] and von Gablenz and Holube [7]. The analysis shows the adoption rates in relation to hearing impairment, and illustrates the proportion of persons classified as hearing impaired without hearing aids. This closes the evaluation gap criticized by Löhler et al. [20] for the cross-sectional data from Oldenburg, Emden, and Aalen collected in the years 2008–2012. It should, however, be taken into account that the incompleteness of the data analyzed, as criticized in Figure 3 [Fig. 3] in the article of Löhler et al. [20] does not apply to the present publication or to any other publications of the authors. In both publications, von Gablenz et al. [2] and von Gablenz and Holube [7], all data collected were fully documented. Rather, Löhler et al. [20] probably mistakenly incorrectly assigned the relevant publications and misquoted the number of study participants.

Data for this article are available from the Dryad Repository: https://doi.org/10.5061/dryad.cfxpnvx27 [21]

Parts of this contribution were presented at 1. Interdisziplinäres Kolloquium der KIND Hörstiftung in Berlin, Germany, February 4, 2019.

The authors declare that they have no competing interests.

The authors would like to thank all examiners and participants for data collection and participation and, in Oldenburg, Prof. Dr. Karsten Plotz and the local hearing-aid acousticians for their support.

HÖRSTAT was supported by the European Regional Development Fund (ERDF), by the governmental funding initiative Niedersächsisches Vorab of the Lower Saxony Ministry for Science and Culture, the research focus “Hören im Alltag Oldenburg (HALLO)”, as well as the research fund of the Jade University of Applied Sciences. The Research Association of German Hearing Aid Acousticians (Forschungsgemeinschaft Deutscher Hörgeräte-Akustiker, FDHA) funded the accompanying study of the German short form of the SSQ. We thank the company Auritec for providing the portable audiometers Ear 2.0.

The study “How does Germany hear?” was financed by the German military (Bundeswehr). The audio truck of WTD91 (Meppen) was made available for audiometric measurements.

Language support for this manuscript was provided by STELS-OL http://stels-ol.de.