Comparison of Self-reported Measures of Hearing With an Objective Audiometric Measure in Adults in the English Longitudinal Study of Ageing

Key Points Questions Among middled-aged adults, are self-reported measures of hearing concordant with audiometry findings, and is potential discordance associated with age or lifestyle factors? Findings In this cross-sectional study of 9666 participants in the English Longitudinal Study of Ageing, self-report measures of hearing had limited accuracy and were not sufficiently sensitive to detect hearing loss. Hearing loss went undetected by the self-reported measures. Meaning These findings may inform public health policies relevant to selection of appropriate and validated tools for detecting hearing problems among middle-aged and older adults.


Introduction
Hearing loss (HL) is an important public health concern, 1 with an estimated 432 million adults worldwide and one-third of people older than 65 years affected by disabling HL. 2 As a noncommunicable disease, HL is far beyond a sensory disorder and can have profound effects on people's quality of life, [3][4][5][6] which reinforces the importance of early detection and intervention for the maintenance of physical and emotional well-being among older adults, where the burden of disease is the highest. 4,7 The World Health Organization has highlighted the pressing need for measures to promote public health action by facilitating early identification of hearing difficulties that supports prioritization of service provision at the community level and integration within primary care systems. 8 Large-scale hearing screening programs and tools to detect HL in major health care sectors, such as primary care, do not exist globally, including in high-income countries. 1 This lack of screening programs 9 excludes the early detection and treatment of patients with gradually progressive HL, 5 and the annual cost of unaddressed HL exceeds $750 billion globally. 8 Moreover, in the absence of HL screening programs that could identify those who are unaware that they have HL (ie, unacknowledged HL), 10 hearing help-seeking depends on self-recognition of hearing difficulties 9 as a crucial step for the initiation of contact with a health care professional in primary health care settings 11 and consequently the referral to ear specialists and hearing aid provision (D.T. et al, unpublished data, 2020).
Self-reported measures are frequently used to gather hearing health data in population-based epidemiologic studies. Evidence indicates a discordance between self-reported and objective measures of hearing because adults self-report HL according to their beliefs, which are influenced by a range of contextual factors. 12,13 However, the validity and the factors associated with the concordance between self-reported HL and manual audiometry remain mostly unknown. 7 The hearing measures in the English Longitudinal Study of Ageing (ELSA) are comparable to 7 other global aging surveys with harmonized physical and anthropometric measurements. 14 Thus, the validation of hearing measures is essential for hearing data quality evaluation and can help explain some of the inconsistencies in findings regarding the association of HL with functional outcomes in older adults. 15,16 The aims of this study were to examine the concordance of self-reported measures of hearing difficulty in ELSA, with objective hearing data measured by a handheld audiometric screening device, and the factors associated with the potential discordances among these measures across different population subgroups of a representative sample of people 50 years and older in England.

Study Population
We used data from ELSA, which is a large, population-based, prospective cohort study that provides a unique resource for exploring issues associated with aging in England in the 21st century. 17 The full analytic cohort was composed of 9666 individuals participating in the wave 7 of ELSA, which collected information from June 1, 2014, to May 31, 2015. For the purpose of this cross-sectional analysis, we further analyzed a sample of 8529 adults 50 to 89 years of age who had an assessment of their hearing by self-reported measures, consented for assessment by a qualified nurse via a hearing screening device, 18 and did not have an ear infection or a cochlear implant. All participants gave written informed consent at the recruitment wave to participate in ELSA and at each subsequent wave. All data were anonymized. Ethical approval was granted by the National Research and Ethics Committee. 19 This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Self-reported Hearing Difficulty
According to ELSA documentation, 19 hearing difficulty is defined as having declared fair or poor hearing on a 5-point Likert scale (with 1 indicating excellent; 2, very good; 3, good; 4, fair; and 5, poor) or finding it difficult to follow a conversation if there is background noise (such as television, radio, or children playing). The participants who positively answered the last question then answered in a separate question whether they had slight, moderate, or great difficulty in following a conversation if there is background noise. We used that response for a further classification of their hearing difficulty into categories, eliminating those who had indicated slight difficulty following a conversation if there is background noise to allow for a fair comparison with the categories of moderate and moderately severe or severe objectively measured HL (eFigure in the Supplement).

Objectively Measured HL
The objective measurement of hearing acuity was performed by the HearCheck Screener (Siemens), a handheld audiometric screening device. 18 The HearCheck Screener automatically generates 6 tones in total: a fixed series of 3 midfrequency sounds with decreasing volume at 1 kHz (at 55 dB HL, 34 dB HL, and 20 dB HL) and afterwards another 3 pure high-frequency sounds at decreasing intensities at 3 kHz (at 75 dB HL, 55 dB HL, and 35 dB HL), testing for audibility for each sequence and per each ear. Participants indicated when they hear the sound by raising their finger.
The HearCheck Screener is an accurate tool in detecting HL when compared with pure-tone air conduction averages, which are designated as gold standard values. In cases of moderate or worse HL, the HearCheck Screener fulfills all the criteria of high sensitivity, high specificity, and high positive predictive values. 20 Hearing level was defined as greater than 35 dB HL at 3.0 kHz in the better-hearing ear because this is the level at which intervention for HL is definitely beneficial. 21 Those with HL were further subdivided according to a categorization that has been previously used in the literature for the characterization of those assessed by the same audiometric screening device 21 as follows: (1) moderate HL (tones heard at 75 dB HL and 55 dB HL but not at 35 dB HL) or (2) moderately severe or severe HL (tone heard or not at 75 dB HL and tones not heard at 55 dB HL and 35 dB HL). net household income and the total nonpension wealth (first quintile indicating the lowest and fifth quintile indicating the highest).
We considered as covariates age, sex, and lifestyle factors (such as body mass index, physical activity, and tobacco and alcohol consumption) because these are key risk factors for HL among older adults. 22 We dichotomized marital status into currently married (married, first and only marriage; in a registered civil partnership; or remarried, in a second or later marriage) or not (single, ie, never married and never registered in a marriage; separated but still legally married; divorced; or widowed).
Retirement status was also dichotomized to being currently retired or not.

Statistical Analysis
Bivariate analyses were performed from July 1 to December 30, 2018, and multivariate analysis from January 1 to June 30, 2019. Descriptive statistical measures were reported on hearing difficulties, hearing in noise, quality of care in hearing, and hearing aid recommendation in ELSA wave 7.
Participants' self-reported and objectively measured HL (moderate and moderately severe or severe) was reported as absolute number (relative frequency). We fitted multiple logistic regression models to identify factors associated with the false-negative report of hearing difficulties in people with objectively identified HL. Age was categorized into 3 groups (50-64, 65-74, and 75-89 years) to allow for a comparison with the study by Benova et al, 9 which examined the self-reported hearing difficulty in ELSA wave 2. There were no missing values in the hearing data of the final analytical sample, which was specifically chosen for the study to fulfill the criteria of completed assessment of hearing by self-reported measures, with given consent for assessment by pure-tone audiometry and without any ear infection or cochlear implant. Separate analyses were conducted for moderate and moderately severe or severe HL. Because some data were missing at random on many variables, we excluded records with missing data from our analyses, concluding that this would be unlikely to affect the validity of our findings. 23,24 For all models, odds ratios (ORs) and 95% CIs are presented. The performance of self-reported hearing difficulty with second stage pure-tone audiometry screening (sensitivity, specificity, and positive and negative predictive values as overall test accuracy) was calculated, and the area under the receiver operating characteristic curve represents the accuracy of all models. We used the Hosmer-Lemeshow test as a postestimation tool, which demonstrates the goodness of fit of logistic regression models. A 2-tailed P Յ .05 was considered to be statistically significant. All data were analyzed using Stata, version 14 (StataCorp).  (Figure). Within the cohort, 3801 (39.3%) reported that they had hearing difficulties. Of those 3801 individuals with self-reported hearing difficulty, 1949 (51.3%) did not tell a physician or nurse about their hearing problems, thereby missing the opportunity to be referred for further assessment.

Self-Reported Hearing Acuity
Examining the characteristics of the 2 separate categories (not mutually exclusive events) that the self-reported hearing difficulty was composed of, we found that not informing a health care professional was more common among those who reported difficulty in following conversations in the presence of background noise (1753 of 3424 [51.2%]) compared with those who had fair to poor self-reported hearing (691 of 2086 [33.1%]). Importantly, 1894 of the 3425 participants (55.3%) who had reported that they found it difficult to follow a conversation if there is background noise (such as television, radio, or children playing) and did not have hearing aids had reported that they have good, very good, or excellent hearing, which indicates that more than half of them had unacknowledged HL, with 718 of 1894 (37.9%) of them having moderate or great difficulty ( Table 1).  Table 3 gives the distribution of sociodemographic characteristics of participants' self-reported and objectively measured HL. Table 4 gives the summary of multiple logistic regression for variables Is your hearing (using hearing aids as usual)...

Objectively Measured HL
Have you told the doctor or nurse about your hearing problems?
Do you find it difficult to follow a conversation if there is background noise, such as television, radio, or children playing (using a hearing aid as usual)?
Has any doctor or nurse or ear specialist recommended a hearing aid?
When you told the doctor or nurse about your hearing problems, did he or she refer you to an ear specialist to check your hearing?

Discussion
In this study, we examined the validity of self-reported measures compared with HL as measured by the HearCheck Screener. We found that in a population-based sample of 8529 adults 50 to 89 years of age, nearly one-third of those had objectively identified HL that went undetected by the selfreport measures. These findings suggest that the use of a screening measure for audiometric testing along with a self-report measure in epidemiologic studies and clinical practice is essential for accurately identifying older people with HL. Moreover, we found that female sex, older age, socioeconomic inequalities, and unhealthy lifestyle (tobacco use, alcohol intake above the low-risk level guidelines, and lower levels of physical activity), which are recognized as key risk factors for HL among older adults, 22 were largely associated with the inaccuracy of self-identification of hearing difficulties in those with objectively identified HL.

Comparison With Previous Literature
Our findings are consistent with previous studies 7,10,15,26-31 that have examined on a smaller scale the performance of self-reported hearing difficulties in combination with pure-tone audiometry among elderly individuals. However, to our knowledge, our study is the first vigorous examination of the validity of self-reported measures of hearing, including difficulties in background noise, with objective audiometric assessments in such a large and nationally representative cohort.
In general, all studies except for the studies by Diao et al 26  This finding may reflect that people tend to undervalue the importance of hearing and consider its loss as an inevitable accompaniment of getting older 22 and therefore adapt to HL over time, 16 underestimating the magnitude of their HL. 29 Regarding the role of socioeconomic position, our findings are consistent with previous studies that found that agreement rates between subjective and objective hearing measures were relatively lower among those of a lower educational attainment 28 and occupational groups subject to noiseinduced HL. 32 The role of income in the false-negative report of hearing difficulties may reflect financial barriers to the use of and access to hearing health care 33 and the downgrade of HL as a health priority. 26

Implications for Research, Policy, and Practice
These findings have important public health implications and call for a revised assessment approach for HL in older adults; clinical research often relies on self-report measure of HL, but our findings indicate that this could not be regarded as a well-suited and accurate measure to identifying individuals with HL without the additional use of a screening measure for audiometric testing. 27 The underestimation of hearing difficulties poses a significant barrier to HL intervention, and the selfreport measures should not be considered reliable measures of hearing acuity to influence the judgment for referral to secondary care.
The help-seeking behavior for hearing difficulties starts with individuals' self-diagnosis and initiation of contact with a health care professional in primary health care settings. 13 In addition, unacknowledged HL constitutes a significant nonfinancial barrier. The existence of objective hearing measures is crucial, particularly for those belonging to high-risk groups that are most likely to remain unrecognized, such as people who face socioeconomic inequalities and adopt an unhealthy lifestyle, because these factors may affect the initiation of help-seeking and consequently the referral to ear specialists. Our findings address important conflicts in the literature, shedding light on the inconsistencies across studies regarding the association of HL with functional outcomes 15 and may reflect attitudinal differences across different cultures and geographic variation in the acknowledgment of hearing difficulties.

Strengths and Limitations
The main strength of our study is that it provides the largest and most accurate evaluation of the discordance between objective and self-reported measures of HL today. Our study is also the first, to our knowledge, to address the association of lifestyle factors with the agreement rate, which had not been previously examined in the literature. 15 However, the study also has significant limitations. First, the cross-sectional analyses did not allow for causal or temporal relationships among the factors associated with the inaccuracy of self-reported measures. In addition, questionnaires that contain few questions to assess hearing deficits may have validity. 34 A relatively small proportion of participants who responded having good, very good, or excellent hearing were also using a hearing aid, which may have confounded their response. Finally, the comparison of self-reported measure to the results from the HearCheck Screener may contain information bias because the screening tool identified only those with HL greater than 35 dB HL at 3.0 kHz in the better-hearing ear, whereas the self-reported questions did not specify that criterion.

Conclusions
Our study found that self-report measurement of HL had limited concordance with objective measures of HL. In light of these findings, the importance of an effective and sustainable HL screening strategy for the early detection and intervention for HL in older adults is reinforced. The lack of screening programs excludes the early detection and treatment of patients with gradually progressive HL, especially those with unacknowledged HL. These results should be considered by HL researchers who analyze self-reported hearing data as a surrogate measurement of audiometric hearing to identify bias in their observed analytic research results. Future research should examine the role of other environmental and personal factors in the agreement rate between self-reported and objective measures of hearing, for which little is known, 15 and investigate sociospatial hearing health inequalities.