The association of hearing problems with social network strength and depressive symptoms: the cardiovascular health study

Abstract

Background

research on the association between hearing impairment and psychosocial outcomes is not only limited but also yielded mixed results.

Methods

we investigated associations between annual self-reports of hearing problems, depressive symptoms and social network strength among 5,888 adults from the Cardiovascular Health Study over a period of 9 years. Social network strength and depressive symptoms were defined using the Lubben Social Network Scale (LSNS), and the Center for Epidemiological Studies Depression Scale (CES-D).

Results

hearing problems were associated with weaker social networks and more depressive symptoms. These association differed for prevalent versus incident hearing problems. Participants with prevalent hearing problems scored an adjusted 0.47 points lower (95% CI: –2.20, −0.71) on the LSNS and 0.71 points higher (95% CI: 0.23, 1.19) on the CES-D than those without hearing problems. Participants with incident hearing problems had a greater decline of 0.12 points (95% CI: –0.12, −0.03) per year in social network score than individuals with no hearing problems after adjusting for confounders. Females appeared to be more vulnerable to changes in social network strength than males (P-value for interaction = 0.02), but not for changes in depressive score. Accounting for social network score did not appear to attenuate the association between hearing problems and depressive score.

Conclusion

findings suggest that older adults with prevalent hearing problems may be more at risk for depression, but individuals with incident hearing problems may be at greater risk for a winnowing of their social network.

Key Points

• Research on the effects of hearing loss on psychosocial outcomes is limited.

• We found that persons reporting hearing problems were more likely to have weaker social networks.

• Those with a history of hearing problems were more likely to report more depressive symptoms.

• Timing of hearing loss onset may be an important factor to consider when designing studies on hearing loss.

Introduction

In the USA, the prevalence of hearing loss is approximately 6.2%, according to the WHO 2021 World Report on Hearing [1]. This prevalence is expected to double in the next 30 years due to population ageing [2]. Hearing loss is prevalent in adults aged 70 and older, as an estimated 66% have disabling hearing loss [3, 4]. As such, the subpopulation of adults over the age of 70 may be particularly vulnerable to health outcomes that have been linked to hearing loss. Hearing loss has been associated with a lower quality of life, communication difficulties, impaired activities of daily living, cognitive decline and physical decline in older adults [5–8]. Despite these known adverse outcomes, studies on the association of hearing loss and psychosocial outcomes such as depression and social structure are sparse and have conflicting results. A comprehensive review conducted in 2017 found that most longitudinal studies were conducted outside of the USA; and that some studies showed a significant association between hearing loss and increased risk of depression, whereas others found no association [9]. In a more recent 2020 study, hearing loss was found to be significantly associated with depressive symptoms in older adults enrolled in the Hispanic Community Health Study in the USA [10].

Findings on the association of hearing loss with social factors such as social isolation, networks and support are similarly inconsistent. Some studies have found that hearing loss is associated with worse social functioning scores and deteriorations in social support, respectively [11, 12]. However, one study found a significant association between hearing loss and social isolation only in the subgroup of women aged 60–69 [13], and another found that the association between hearing loss and tangible social support was stronger in females than males in the Canadian Longitudinal Study on Aging [14]. One potential explanation for these heterogeneous findings is that most studies have not discriminated between recent and long-standing hearing loss, which may alter the association with health outcomes.

The present study will evaluate the relationship between hearing problems, social network strength and depression using longitudinal data from the Cardiovascular Health Study (CHS), based in the USA. We sought to determine if there was an association: (1) between hearing problems and social network strength, and (2) between hearing problems and depressive symptoms. We also investigated if sex was an effect modifier of the association of hearing problems with social network strength or depressive symptoms. This study will provide important evidence of whether timing (prevalent or new onset) is a significant factor in understanding the association between hearing problems, social networks and depression.

Methods

Study population

The CHS is a population based-cohort study of 5,888 adults, aged 65 and older, recruited from Medicare eligibility lists at four sites (Sacramento County, CA; Pittsburgh, PA; Washington County, MD; Forsyth County, NC). The primary aim was to evaluate risk factors for the development and progression of cardiovascular disease in older adults. Investigators enrolled a total of 5,888 participants in two waves: 5,201 participants in 1989–90, and an additional 687 predominantly African American participants in 1992–93. Participants completed clinic visits at enrollment, followed by telephone calls at the 6-month mark of each year and annual clinic visits. These visits consisted of an interview, health questionnaire, physical examination and collection of blood specimens. All participants were asked to identify a proxy respondent in the event they became unable to answer study questions. The CHS study was approved by the Institutional Review Boards of all sites. Additional study details have been described elsewhere [16].

Measurement of hearing problems

As part of the annual CHS check-in, participants or their proxy respondents answered three yes/no questions related to hearing problems: (1) Can you hear well enough (with or without) a hearing aid to listen to the radio; (2) Can you hear well enough (with or without) a hearing aid to use the telephone; (3) Can you hear well enough (with or without) a hearing aid to carry on a conversation in a crowded room? These questions were asked between 1989–94 and 1996–99. In this analysis, we defined a hearing problem as prevalent if the participant answered ‘no’ to any of the three questions at baseline. New onset, also referred to as incident hearing problem, was defined as a ‘no’ to at least one of the hearing questions at any point during the participant’s time in the cohort.

Measurement of social networks and depression

Social network strength was measured using the Lubben Social Network Scale (LSNS), which was administered annually in 1989–94, and in 1998–99. The LSNS is adapted from the Berkman-Syme Social Network Index to be used with older populations. It includes 10 questions coded on a scale of 0–5, that sum to a total score ranging from 0 to 50. Higher scores indicate stronger social networks. The Cronbach alpha for the LSNS is 0.7 [17].

Depression was assessed annually using the 10-item Center for Epidemiological Studies Depression Scale (CES-D). Total scores can range from 0 to 30, with a higher score indicating greater depressive symptoms [18]. In a study of older adults, the shortened CES-D was found to be reliable for assessing clinical depression when compared to the validated full-length CES-D, with a Cohen k statistic of 0.97 [19].

Potential confounders

Information on potential confounders was obtained from the baseline visit in 1989–90 or 1992–93 visit. Demographic characteristics included enrollment wave, age, sex (male, female), race (white/other, black) and education (less than high school, high school or General Education Diploma, some college or higher degree). Behavioural characteristics included smoking status (never, former, current), alcohol use defined as none (0 drinks per week), some use (1–7 drinks per week for women, or 1 to 14 drinks for men), or frequent use (>7 drinks per week for women or >14 drinks per week for men), living alone, activities of daily living (0, 1 or more) and BMI, calculated as weight(kg)/height(m)². Baseline cognitive function, measured by the Modified Mini Mental State Examination (3MSE), was also assessed as it could affect reporting of psychosocial outcomes.

We also assessed several clinical factors that have been previously associated with hearing loss. In our analysis, diabetes was defined as a fasting glucose greater or equal to 126 mg/dl or reported use of diabetic medications. Hypertension was defined as hypertensive medication use, or blood pressure greater than 140/90 mm Hg. Cardiovascular disease included history of coronary heart disease, heart failure and stroke.

Statistical analysis

We used Kruskal–Wallis tests for continuous variables and chi-squared (χ²) tests for categorical variables to evaluate differences between participants with incident, prevalent or no hearing problems. Mixed effect models were used to model changes in social networks or depressive symptoms, with a fixed effect for time. We evaluated changes in social networks or depressive score by including an interaction term between hearing problem and time. To ensure time-ordering, we included a 1-year lag after self-reported hearing problems for the outcome of interest. The final, adjusted multilevel model included demographical confounders, activities of daily living and cognitive function. Additionally, we tested effect modification by sex by including an interaction term between type of hearing problems and sex in the multilevel model. If the effect modification was significant (P < 0.05), we further stratified by sex to understand the differential impact of hearing problems on the outcome.

As a comparison study, we evaluated the concordance of the self-reported hearing problem variable with use of a proxy respondent, hearing aid use and audiometric results, which were captured as part of bi-annual clinic visits. CHS collected information on the reason for the use of proxy respondent (cognitive, hearing, hospitalised, other illness, other), so we evaluated use of proxy respondent when hearing problems were indicated as the reason. We also explored the concordance of hearing problems and use of a hearing aid, which was only assessed at the 1998–99 visit. Additionally, we compared hearing problems to hearing function as measured via audiometry at the 1998–99 visit. We tested whether this association varied by sex and cognitive function to evaluate whether these factors influence the reporting of hearing problems.

All analyses were performed in SAS (SAS University Edition, Cary, NC, USA). Significance was defined for all analyses as P < 0.05.

Results

Participant characteristics

Among the 5,888 adults included in the analysis, approximately 8% (n = 443) reported having hearing problems at baseline, and 20% (n = 1,176) reported acquiring hearing problems during their time in the CHS study. Over 9 years of follow-up, the median time to incident hearing problems was 6.55 years (IQR: 4.00–7.00). Participants who never reported hearing problems were younger, more likely to be female and African American, have a higher body mass index (BMI), and to have some college or higher degree. In addition, individuals with no reported hearing problems appeared healthier at baseline. They were more likely to report no activities of daily living (ADL) limitations, less likely to have cognitive impairment, and less likely to have a history of cardiovascular disease. Covariate data had a minimal amount of missingness; the maximum missingness was in the living alone variable (12%). At baseline, participants with no reported hearing problems had a higher mean social network score than participants with prevalent hearing problems and lower mean depressive score compared to both those with prevalent and those incident hearing problems (Table 1).

Table 1.

Characteristics of CHS participants at baseline

Characteristics	Incident		Prevalent		None
	1,176	19.9%	443	7.5%	4,269	72.5%
	No.	%	No.	%	No.	%	P *
Age mean (SD)	74 (6)		75 (7)		72 (5)		<0.001
Cohort							<0.001
Wave 1 (1989–90)	1,093	92.9	408	92.10	3,700	86.7
Wave 2 (1992–93)	83	7.1	35	7.9	569	13.3
Sex							0.002
Female	676	57.5	220	49.7	2,497	58.5
Male	500	42.5	223	50.3	1,772	41.5
Race							<0.001
African American	165	11.5	48	10.8	741	17.4
White/Other	1,041	88.5	395	89.2	3,528	82.6
Education							<0.001
Less than high school diploma	385	32.9	166	38.1	1,175	27.6
High school diploma	289	24.7	114	26.2	1,217	28.6
Some college or higher degree	497	42.4	156	35.8	1,866	43.8
Alcohol use							0.10
Never	595	51.0	239	54.2	2,103	49.4
Light	444	38.1	170	38.6	1,713	40.3
Frequent	128	10.9	32	7.3	438	10.3
Current smoker							0.02
Never	568	48.3	227	51.2	1,943	45.6
Former	487	41.5	175	32.5	1,782	41.8
Current	120	10.2	41	9.3	539	12.6
Body mass index (kg/m2) mean (SD)	26.2 (5.2)		26.3 (4.5)		26.7 (4.7)		0.01
Living alone							0.94
Yes	145	14.4	58	14.5	561	14.9
Cognitive function							<0.001
<80	133	12.8	72	21.0	383	10.7
80	910	87.3	271	79.0	3,201	89.3
Activities daily living							<0.001
0	1,050	89.6	370	83.5	3,982	93.4
1 or more	122	10.4	73	16.5	281	6.6
History of coronary heart disease							0.03
Yes	15	21.1	100	22.6	892	18.7
History of heart failure							0.09
Yes	5	8.5	30	6.8	183	3.8
Hypertension							0.53
Yes	829	70.5	300	67.7	298	69.3
History of stroke							0.65
Yes	50	4.25	15	3.39	184	4.31
History of diabetes							0.50
Yes	171	14.5	72	16.3	678	15.9
Social network score mean (SD)	31.8 (7.4)		31.0 (7.8)		32.6 (7.3)		<0.001
Depressive score mean (SD)	4.8 (4.6)		5.8 (5.0)		4.3 (4.4)		<0.001

Cognitive function measured by Modified Mini Mental State Exam; hypertension defined as hypertensive medication use, SBP 140 mmHg, or DBP 90 mmHg; diabetes as diabetic medication use or glucose 126 mg/dL; social network score measured by Lubben Social Network Scale, depressive scores measured by CES-D.

^*Chi-square test for categorical variables, Kruskal–Wallis test for continuous variables.

Association of hearing problems and social network strength

The association of hearing problems with social network score varied across type of hearing problems in both the unadjusted and adjusted models (Table 2). Individuals that were categorised as having prevalent hearing problems had lower social network scores than individuals with no hearing problems. In the adjusted model, individuals with prevalent hearing problems scored 0.47 points lower (95%CI: –2.20, −0.71) on the LSNS than individuals without hearing problems. The difference in social network score was not significant between individuals with incident hearing problems and those without hearing problems. However, individuals with incident hearing problems had a greater decline in social network score than individuals with no hearing problems after adjusting for confounders; with 0.12 points (95%CI: –0.12, −0.03) decrease in LSNS score per year.

Table 2.

Association of hearing problems and social network score

	Unadjusted		Model 1a
	Coefficient	95% CI	Coefficient	95% CI
Prevalent hearing loss	−1.46*	−2.20, −0.71	−0.47*	−1.18, −0.26
Incident hearing loss	−0.04	−0.71, 0.63	0.52	−0.20, 1.24
Time	−0.01	−0.03, 0.02	0.03	0.00, 0.06
Time* Prevalent hearing loss	0.02	−0.08, 0.11	−0.05	−0.15, 0.06
Time* Incident hearing loss	−0.07	−0.15, 0.01	−0.12*	−0.21, −0.03

Social network strength was measured using the Lubben Social Network Scale with a score range of 0–50, 50 indicating strongest social networks. Coefficients are from multilevel models.

^* P < 0.05.

^aAdjusted for age, cohort, sex, race, body mass index, education, activities of daily living, cognitive function.

^bIncludes variables in Model 1 and interaction term for hearing loss and sex.

We found evidence of effect modification of incident hearing problems by sex (P-value for interaction = 0.02). Although females with incident hearing problems appeared to have higher social network scores at baseline than those without hearing problems (1.15 points, 95% CI: 0.18, 2.12), they experienced a greater decline in social network scores over time than females without hearing problems (−0.21 points per year, 95% CI: –0.33, −0.09) after the first hearing problem was reported (Table 3).

Table 3.

Association of hearing problems and social network score, stratified by sex

	Men (n = 2,495)		Women (n = 3,393)
	Coefficient	95% CI	Coefficient	95% CI
Prevalent hearing loss	−0.11	−1.15, 0.93	−0.68	−1.68, 0.33
Incident hearing loss	−0.18	−1.25, 0.89	1.15*	0.18, 2.12
Time	0.15*	0.10, 0.93	−0.05*	−0.09, −0.02
Time* Prevalent hearing loss	−0.05	−0.20, 0.10	−0.09	−0.23, 0.05
Time* Incident hearing loss	−0.04	−0.18, 0.10	−0.21*	−0.33, −0.09

^* P < 0.05.

Association of hearing problems and depressive symptoms

Individuals with hearing problems had higher CES-D scores compared to those without hearing problems, and this difference was significant for both prevalent and incident hearing problems (Table 4). Compared to individuals without hearing problems, individuals with prevalent hearing problems scored 0.71 points higher (95% CI: 0.23, 1.19) on the CES-D, and individuals with incident hearing problems scored 0.56 points higher (95% CI: 0.09, 1.03). Only individuals with prevalent hearing problems had a small increase in depressive symptoms over time compared to those without hearing problems, with CES-D scores increasing 0.06 points per year (95% CI: 0.00, 0.11). Unlike social network strength, there did not appear to be any effect modification by sex for depressive symptoms (P-values for interaction 0.41 and 0.65 for prevalent and incident, respectively).

Table 4.

Association of hearing problems and depressive symptoms

	Unadjusted		Model 1a
	Coefficient	95% CI	Coefficient	95% CI
Prevalent hearing loss	1.03*	0.57, 1.49	0.71*	0.23, 1.19
Incident hearing loss	0.85*	0.42, 1.28	0.56*	0.09, 1.03
Time	0.16*	0.15, 0.17	0.17*	0.15, 0.18
Time*Prevalent hearing loss	0.06*	0.01, 0.11	0.06*	0.00, 0.11
Time*Incident hearing loss	−0.02	−0.07, 0.03	0.02	−0.04, 0.06

Depressive symptoms were measured using CES-D with a score range of 0–30, 30 indicating more depressive symptoms. All coefficients are from multilevel models.

^* P < 0.05.

^aAdjusted for age, cohort, sex, race, BMI, education, ADL, cognitive function.

^bIncludes variables in Model 1 and interaction term for hearing loss and sex.

We also investigated if adjusting for social network strength attenuated the association observed between hearing problems and depressive symptoms. The model estimates for the difference in CES-D score over time did not appear to be changed when adjusting for social network strength; 0.71 points unadjusted (95% CI: 0.23, 1.19) versus 0.72 points adjusted (95% CI: 0.23, 12.1) for prevalent hearing problems and 0.56 points unadjusted (95% CI: 0.09, 1.03) versus 0.53 points adjusted (95% CI: 0.05, 1.01) for incident hearing problems.

Comparative study

Information on use of a proxy respondent was available for 8 years of the CHS study (1991–99). Out of 595 person-visits where a proxy respondent was used due to hearing problems, hearing problems were reported by 84.7% of individuals (n = 504 person-visits). Hearing aid data were available for 75% (n = 2,974) of all subjects in the 1998–99 study visit, of which 17.4% (n = 518) reported using hearing aids. Among individuals who reported using a hearing aid, 69% (359/518) reported hearing problems. Among those who did not report using hearing aids, 83% (2,046/2,456) did not report a hearing problem. Among participants who were unable to hear a sound at high frequencies (>1,000 Hz at 40 dB HL) on the audiometric exam administered during the 1998–99 study visit, only 43% (487/1,127) reported hearing problems. Among those who were able to hear all three frequencies, 88% (1,407/1,586) did not report a hearing problem. There did not appear to be differences by sex or cognitive function. Less than half of those with audiometric hearing loss used a hearing aid (37%, 414/1,127), but those who also reported a hearing problem were more likely to use a hearing aid (70% who reported a hearing problem versus 30% who did not, Supplemental Table 1).

Discussion

In this longitudinal study of older adults in the CHS study, we found that persons reporting hearing problems were more likely to have weaker social networks and report more depressive symptoms than those that did not report hearing problems. Individuals with prevalent hearing problems had smaller social networks, but those with incident hearing problems had greater decline in social network over time. Those with prevalent hearing problems at baseline had more depressive symptoms and had a greater increase in depressive symptoms over time. Our findings suggest that it may be important to consider the timing of hearing problems onset when considering the association between hearing problems, depressive symptoms, and social networks. Females seem to be more vulnerable to changes in social network strength than males, but not to changes in depressive symptoms.

Although some evidence suggests that hearing impairment is associated with higher levels of depressive symptoms, others have found no significant association [12, 20]. One thing to note is that though both studies were longitudinal, and population based, they were conducted in Australia and Netherlands, respectively. The conflicting results may be due to cultural differences in reporting social and psychological variables and suggest results may be difficult to extrapolate across countries and cultures. Our study in an US cohort suggests that the associations between hearing problems and an increase in depressive symptoms are limited to those who have prevalent hearing problems at our study baseline. This is contrary to our hypothesis, as we had expected that participants with prevalent hearing problems would have had more time to develop coping strategies. However, it is possible that this observation also is due to confounding by the age of onset; participants were, on average, in their early seventies at baseline. It is possible that those individuals who develop age-related hearing problems in early old age may perceive themselves as less healthy or more limited than their peers, and this could lead to increased depressive symptoms. Hearing loss has been shown to lead to a reduction of quality of life in older adults, by way of impairing communication abilities [5]. Communication is essential for building and maintaining social networks, so it follows that hearing loss may lead to smaller, weaker social networks. Limited communication could not only affect social network size and social isolation, but also lead to increase depressive symptoms [21, 22].

Previous studies have found that associations between hearing problems and increased social isolation are modified by gender, with a stronger effect in women than in men [15]. In a cross-sectional study using data from the US National Health and Nutrition Examination, greater hearing loss was associated with increased odds of social isolation only among women aged between the ages of 60 and 69 [12]. Our study also found a stronger effect of hearing problems in women, and we observed that a decline in social networks over time was only significant for women but not men with incident hearing problems (although CHS assessed participant sex, we use the terms women/men to compare our findings with existing literature). Differences in psychosocial outcomes for men and women could be explained by differences in coping mechanisms and how social networks are perceived. For example, men perceive social support from friends significantly higher than women, but women report more perceived social support from spouses. [23] It also appears that women are more likely to seek social support from smaller and closer knit social networks than men [24].

The primary strength of this study is that it is one of the first to differentiate between long standing hearing problems and those more recently acquired. However, our study has several limitations, primarily because of the potential for misclassification of self or proxy-reported hearing problems. Audiometric data were only available for 1 year of the study (1998–99 study visit), so we were unable to investigate if those with prevalent hearing problems experienced a deterioration in their measured hearing thresholds over time, which could have contributed to the stronger relationship between hearing problems and depressive symptoms we observed in those with prevalent hearing problems. However, it is likely that we underestimated the prevalence of hearing problems, as older adults can be less likely to perceive their own decline in hearing loss, especially while it is mild enough to not impact day to day life and more easily mediated by other factors such as compensatory strategies, gender and comorbidities [25].

Additionally, since hearing loss typically deteriorates rather than improves over time, we chose to define an individual as having hearing problems from the first time they reported hearing problems onwards. We performed several sensitivity analyses to investigate if this simplifying assumption may have introduced measurement error of the hearing problem variable by comparing the hearing problems variable to hearing aid use, proxy use and audiometric test values. Reported hearing aid usage in our cohort was 17.4%, which is close to the estimated prevalence of hearing aid use in the WHO 2021 World Report on Hearing [1]. Finally, an individual’s report of hearing problems may have potentially been confounded by cognitive decline over time, as we only adjusted for cognitive function at baseline.

Findings from this study suggest that more longitudinal studies may be needed to understand the association between timing of hearing loss and psychosocial outcomes; using audiometric data if possible. Because individuals with prevalent hearing loss appear to be most at risk for worsening depressive symptoms over time, it may be important for future research to investigate the effect of earlier detection and treatment of hearing loss on depression in this subpopulation. Individuals with incident hearing loss, especially women, may benefit from targeted coping strategies and auditory rehabilitation to maintain social networks. Building on this retrospective cohort study, future studies assessing the effect of hearing impairment on psychosocial or other health outcomes could benefit from distinguishing between new onset and prevalent hearing loss.

Declaration of Conflicts of Interest

None.

Declaration of Sources of Funding

This research was supported by contracts HHSN268201200036C, HHSN268200800007C, HHSN268201800001C, N01HC55222, N01HC85079, N01HC85080, N01HC85081, N01HC85082, N01HC85083, N01HC85086, and grants U01HL080295 and U01HL130114 from the National Heart, Lung, and Blood Institute (NHLBI), with additional contribution from the National Institute of Neurological Disorders and Stroke (NINDS). Additional support was provided by R01AG023629 from the National Institute on Aging (NIA). A full list of principal CHS investigators and institutions can be found at CHS-NHLBI.org.