Association of Age-Related Hearing Impairment With Physical Functioning Among Community-Dwelling Older Adults in the US

This cohort study assesses the association between hearing impairment and decline in physical function in terms of balance and walking endurance in a group of community-dwelling older adults in the US.


Introduction
Physical functioning is necessary for independent living and tends to decline with age. [1][2][3] Hearing impairment, which affects approximately two-thirds of adults older than 70 years, 4 is a risk factor for various adverse outcomes. [5][6][7] Hearing impairment may also adversely affect physical functioning through reduced perception of auditory input that contributes to walking and balance. 8 However, research characterizing the association between hearing impairment and objective physical function and walking endurance measures is limited.
Associations between self-reported hearing impairment and poorer physical function have been reported previously. 9,10 However, self-reported hearing impairment is prone to measurement error and has been shown to underestimate associations with objective measures of function. 11 Although studies with audiometrically assessed hearing, the criterion-standard clinical measure, have revealed associations with slower gait and poorer physical function, [12][13][14] these studies did not assess associations with physical function components separately. Moreover, studies of the association between hearing impairment and walking endurance-the ability to walk longer distances, an early factor associated with disability 15 -are scarce and have inconsistent findings. 10,16 Therefore, we investigated the association of hearing impairment with physical function and walking endurance in a cohort of community-dwelling older adults in the US. We hypothesized that participants with hearing impairment would have poorer concurrent physical function and walking endurance and a faster decline in physical function over approximately 8 years compared with participants with normal hearing.

Study Population
For this cohort study, we used a population from the Atherosclerosis Risk in Communities (ARIC) study, 17

Exposure
Pure tone audiometry, assessed at ARIC visit 6 (2016-2017), was completed in a soundproof booth using insert earphones (EARTone 3a; 3M) and an Interacoustics AD629 or Equinox audiometer (Interacoustics A/S). Measurement of air conduction was completed at standard octaves from 500 to 8000 Hz. For each frequency, the hearing threshold was recorded in decibels hearing level (dB HL). We calculated a 4-frequency (0.5, 1, 2, and 4 kHz) pure tone average for each ear and modeled a

Outcomes
The Short Physical Performance Battery (SPPB) was used to assess physical function, and a fastpaced 2-minute walk (TMW) test was used to measure walking endurance. The SPPB was measured at visits 5, 6, and 7 and included 3 components: balance, gait speed, and chair stands. Each component was scored on a scale of 0 to 4 points, with higher scores indicating better physical function (eTable 1 in the Supplement). The SPPB has an established scoring system, which we used to derive cutoffs for each test from the performance distribution. 20 The composite SPPB score (range, 0-12) was the sum of the 3 scores. We examined the composite SPPB score continuously and used a binary threshold for poor physical performance (SPPB composite score Յ6). 1,20 For each SPPB component, we also used a binary threshold (SPPB component score Յ2) to indicate poor performance, consistent with prior work (eTable 1 in the Supplement). 21 The balance test consisted of holding 3 standing positions for 10 seconds each: side-by-side (easiest), semi-tandem, and full-tandem (hardest) 20 ; participants progressed to a more difficult position if they succeeded in 1 of 2 trials. Each participant's usual pace gait speed was measured twice over a 4-m walk (walking aids permitted). We calculated gait speed (in meters per second) using the trial with the faster result. For the chair stand test, participants stood up from a chair and sat back down with arms crossed 5 times as quickly as possible. Gait speed and time to complete the chair stands were also examined continuously.
The TMW test was administered on a 15.24-m (50-ft) course at visit 6 only. Participants were instructed to walk as fast as they could for 2 minutes. The distance covered was recorded in meters.
Participants unable to complete the SPPB 4-m walk test unaided were excluded from the TMW test, leaving 2535 participants for this analysis.

Additional Independent Variables
Age; sex; educational level (less than high school, high school, or high school or higher); race-study site, owing to the correlation between race and site in the ARIC study 17 (White participants from Minneapolis, Washington County, and Forsyth County and Black participants from Forsyth County and Jackson); body mass index; occupational noise exposure (very loud sounds for >10 hours/wk in the workplace); and smoking status (never, former, or current) were assessed at visit 6. Using data from the annual follow-up interviews, we created a multimorbidity index to control for the presence of chronic conditions. This index was modeled after the Charlson-Deyo Comorbidity Index and included myocardial infarction, stroke, intermittent claudication, heart failure, chronic obstructive pulmonary disease (emphysema or bronchitis), chronic kidney disease, diabetes, and cancer. Stroke and cancer were assigned double weight (counted as 2). No data were available regarding peptic ulcer disease, AIDS or HIV infection, liver disease, or rheumatologic disease. Dementia was not included in the index because we considered it to be a mediator and not a confounder in the association between hearing and physical function. In sensitivity analyses, we tested whether including dementia in the index had any effect on our results. All covariates were assessed at visit 6 and treated as fixed variables.

Statistical Analysis
We evaluated the differences in demographic characteristics and medical conditions across hearing categories using χ 2 and analysis of variance tests as appropriate. We assessed the cross-sectional associations (at study visit 6) between continuous hearing, by 10-dB HL increments, and physical function and walking endurance using 2 models. Model 1 was adjusted for age, sex, and race-study center, and model 2 was adjusted for those variables as well as educational level, body mass index, occupational noise, smoking status, and the multimorbidity index.
Because of the ceiling effect of the SPPB (approximately 50% of participants had a score Ն10), we used Tobit regression models to calculate the mean differences in SPPB composite scores. We used logistic regression analysis to estimate the odds ratios (ORs) of low SPPB composite scores (Յ6) and low scores for each SPPB component (Յ2). We used linear regression analysis to assess differences in gait speed and time to complete chair stands as continuous measures.
We also used linear regression analysis to assess the cross-sectional (visit 6) association between hearing and walking endurance (distance walked during the TMW test). We performed all analyses using Stata, version 15 (StataCorp LLC). Two-sided P < .05 was considered statistically significant. (approximate mean time between visits 5 and 6) to allow the rate of decline in SPPB scores to vary between visits. We allowed for between-person heterogeneity in SPPB scores with random effects for intercept and slopes (before and after year 5). We included interaction terms between hearing and both time coefficients (before and after year 5) to estimate the mean differences in the rate of change in physical function across hearing categories in each period. We evaluated the assumptions of our models with residual plots and present fully adjusted longitudinal models. In a secondary analysis, among 811 participants with moderate or severe hearing impairment, we assessed whether there were differences in physical function and walking endurance by hearing aid use (selfreported as yes or no) using fully adjusted models.
For sensitivity analysis, using ordinal logistic regressions, we estimated the OR of having the next-lower SPPB score (the scores were reversed to compute OR) for all hearing groups. Also, among 2520 participants, we examined the association between hearing (per 10-dB HL increments and across hearing categories) and grip strength (a measure of the forearm's maximal strength) using linear regressions adjusted for covariates in model 2. In addition, we built an alternative multimorbidity index including dementia and used it for adjustment in model 2 in lieu of the index without dementia.

Results
A total of 2956 participants attended study visit 6 ( Table 1

Cross-Sectional Analysis Findings Physical Function
The differences in continuous measures of SPPB tests (composite score, gait speed, and time to complete chair stands) at visit 6, per 10-dB HL increase, and across hearing categories (in which  normal hearing was the reference group) are presented in

Walking Endurance
In model 2 (

Longitudinal Analysis Findings
Participants with worse hearing demonstrated faster declines in SPPB composite scores over time  Table 3 and Figure 2).    Abbreviations: HI, hearing impairment; SPPB, Short Physical Performance Battery. a Adjusted for age, sex, race-center site, body mass index, educational level, occupational noise exposure, smoking status, and multimorbidity index. Supplement), the differences in physical function and walking endurance between hearing aid users and nonusers were not significant.

Sensitivity Analyses
In ordinal logistic regression models, the odds of being in the next lower SPPB score were higher with worse hearing for all outcomes, but the OR was not statistically significant for the chair stand test (eTable 3 in the Supplement). No association between grip strength and hearing measures was found.
The models that used the multimorbidity index with dementia yielded similar estimates and the same inferences as the index without dementia. The only exception was for the results for the association between severe hearing impairment and the gait speed binary outcome. Using the alternative multimorbidity index attenuated the estimate from an OR of 2.11 (95% CI, 1.02-4.33) to 1.97 (95% CI, 0.95-4.10).

Discussion
In this cohort study of community-dwelling older adults in the US, hearing impairment was Collectively, these findings suggest that individuals with hearing impairment may be at greater risk for physical function limitations.
The association between hearing impairment and physical function has been studied previously.
In a previous study, Deal  Our findings are consistent with and expand on findings from this prior work. We studied each SPPB component separately, identifying stronger associations between hearing and balance and gait speed than between hearing and chair stands ( Figure 1). Because of the increased complexity of balance and gait speed, integration of multiple inputs, and coordination of movement, it is likely that these functions rely more heavily on acoustic input from the environment compared with chair stands, which depend more on strength. 8,25 In sensitivity analyses, we found no association between hearing and grip strength, supporting this assumption.  Mikkola et al 10 investigated the associations between self-reported hearing and the SPPB components and did not find any associations after adjustment for confounders. This may be explained by the use of self-reported hearing, which has been shown to underestimate associations compared with objective measures of hearing function. 26 We found an association between hearing and reduced walking endurance. Tests of walking endurance, which correlate with cardiorespiratory fitness, 27 have been shown to be earlier factors associated with mobility limitations and disability compared with self-reported measures, particularly among well-functioning adults. 2,15 In this context, our findings suggest that older adults with hearing impairment may be at an increased risk for mobility limitations. The association between hearing and walking endurance has been documented in only a few studies. Valjean et al 16 assessed the association between audiometrically defined hearing impairment and reduced walking endurance in a 6-minute walk test. If the difference that we found in the TMW pace for participants with severe hearing impairment were maintained for 6 minutes, it would yield a difference (approximately 16 m) similar to that (16 m) reported by Valjean et al. 16 However, their study was limited to Finnish women, and the findings were not statistically significant.
Multiple mechanisms could explain the associations found in the present study. First, a common cause (ie, cardiovascular disease) of hearing and physical impairment 28,29 could contribute to our findings. However, we adjusted for cardiovascular and sociodemographic risk factors, and our estimates were only slightly attenuated. Second, the inner ear hosts both the vestibular and auditory systems. Thus, damage to the inner ear could cause dysfunction in both systems, leading to an association between poorer hearing and balance. Although previous findings suggest that each sensory system has independent risk factors, 30 we cannot rule out this mechanism because we were unable to adjust for vestibular function. Third, the association between hearing and physical function may be mediated through reduced cognitive resources, 31,32 depression, social isolation, and reduced life space, 33-36 resulting in less physical activity. 11,37,38 These mechanisms would be particularly relevant for our findings related to walking endurance. In sensitivity analyses, we observed that when dementia was included in our adjustment variables (part of the multimorbidity index), it attenuated the association between severe hearing impairment and low gait speed scores. This finding is supportive of our theory that cognition is a mediator of the association between hearing and physical function. Fourth, hearing impairment may be associated with reduced physical function via reduced perception of auditory cues from the environment, which contribute to balance. 8,25 We found no differences in physical function or walking endurance in association with hearing aid use (eTable 2 and eTable 4 in the Supplement). Participants who used hearing aids had higher socioeconomic status (associated with better physical function) but worse hearing, and they were older (associated with poorer function). Our measure of hearing aid use was limited to self-reports of yes or no, which did not characterize use patterns. These factors may explain the consistently null findings in the literature. 12 Randomized clinical trials to elucidate the effect of hearing aid use on physical function are warranted. 39

Strengths and Limitations
This study has strengths. The study included a biracial population, 4 US sites, assessment of hearing using pure tone audiometry (clinical standard), performance-based tests of physical function and walking endurance, and repeated measures of physical function over 8 years. Our findings were robust to the adjustment for confounders and consistent across different statistical approaches (binary vs ordinal logistic regression from the sensitivity analysis).
This study also has limitations. First, the hearing assessment was conducted at visit 6; thus, in longitudinal analyses, we included measures of physical function that preceded the assessment of hearing. However, age-related hearing impairment progresses slowly at a rate of 1 to 2 dB HL per year 40,41 ; thus, hearing at visit 5 would be correlated with hearing at visit 6. Also, the physical function at visit 5 would be unlikely to affect hearing acuity at visit 6. Moreover, we observed a significant decline in physical function between visit 6 and visit 7 in our models for participants with mild and moderate hearing impairment ( Figure 2). Second, residual confounding by unmeasured variables, including subclinical cardiovascular disease and vestibular function, was possible. Third, 600 (20%) participants were lost to follow-up between visit 6 and visit 7. However, participants with worse hearing, who likely experienced faster declines in physical function, were more likely to be censored (eTable 5 in the Supplement). Under these conditions, if selection bias were present, it would likely be conservative.

JAMA Network Open | Geriatrics
Only 121 participants had severe hearing impairment. The relatively small sample size of this group could have reduced our statistical power to detect differences in physical functioning between participants with normal hearing and severe hearing impairment. However, we detected a significant association across multiple outcomes except in sensitivity analyses, in which the odds of having a low gait speed score were higher for participants with severe impairment. Although the study population included some Black participants, the sample was predominantly White. The lack of other racial/ethnic groups limits the generalizability of our findings. Further research in cohorts that include larger numbers of individuals of other racial/ethnic groups is warranted.

Conclusions
In this cohort study, hearing impairment was associated with poorer physical function, reduced walking endurance, and faster decline in physical function over time. Because hearing impairment is amenable to prevention and management, it potentially serves as a target for interventions to slow physical decline with aging.