Hearing Loss and Reduced Income Growth: A Longitudinal Socioeconomic Analysis

Abstract

Objective:

Hearing loss (HL) has been linked to negative socioeconomic states, including low income. This study examines the relationship between HL and income growth.

Study Design:

Longitudinal observational study

Setting:

Multi-centered U.S. epidemiologic study (Hispanic Community Health Study; HCHS)

Methods:

Using data from two waves of the HCHS, we analyzed the association between HL and income growth in adults ages 18–74 years using generalized estimating equations. The exposure was HL, measured by 4-frequency pure-tone average (PTA). The outcome was yearly household income growth, with income graded on a 10-bracket scale from <$10,000 to >$100,000. Models controlled for demographics, hearing aid use, and vascular risk.

Results:

1,342 participants met inclusion criteria. At visit 1, average age was 47.6 years (SD=12.2), and average PTA was 13.9 decibels (dB, SD=9.5). Average follow-up was 5.9 years (SD=0.6). There was a significant time*HL interaction: with each 10 dB worsening in HL, the odds of belonging to a higher versus a lower income bracket changed by a factor of 0.979 (p<0.001) between waves 1 and 2. In other words, the odds of belonging to a higher income group decreased with worsening HL. At 38.6 dB, the odds for income growth became <1, indicating income loss rather than growth.

Conclusion:

Increased HL is associated with reduced income growth, including the possibility of negative growth (i.e., income decline). This study emphasizes the value of including socioeconomic measures in randomized controlled trials assessing the impact of HL treatment and the importance of extended follow-up for study participants.

Introduction

Hearing loss (HL) is one of the most common chronic medical conditions in the United States (U.S.). Approximately 25% of individuals aged 60 to 69 have HL, a percentage that rises to over 50% among those aged 70 and older.¹ Hearing aids, indicated for the majority of HL, are non-invasive, widely available, and now more affordable since over-the-counter forms entered the U.S. market in fall 2022.^2,3 For moderately severe HL, cochlear implants are usually an option. Yet, among those aged 70 years and older who may benefit from hearing aids, fewer than 30% have ever tried using them.⁴ With a burgeoning field of research describing associations between HL and comorbidities—such as cardiovascular disease,⁵ depression,^6–9 cognitive decline and dementia,^10–16 falls,¹⁷ and mortality^18,19—this undertreatment is cause for serious concern. This research shows that a large population remains at risk for the multitudinous negative health consequences associated with HL.

HL has also been studied relative to social determinants of health,²⁰ which represent the nonmedical conditions and environments that impact a person’s ability to live, work, thrive, and stay healthy. These include factors like economic status, education, housing, social supports, and access to high-quality healthcare. HL has been associated with low socioeconomic states, such as low income²¹ and unemployment or underemployment.²² Additionally, prior studies have shown that individuals with HL, compared to those with normal hearing, have a higher odds of belonging to a low reading or educational level.^22,23 The mechanisms potentially underlying these relationships remain unclear.

The goal of this study is to further understand the relationship between HL and socioeconomic status, focusing on income growth, by leveraging a large, multicentered, longitudinal analysis, the Hispanic Community Health Study (HCHS). The novel contribution of this study is twofold. First, although HL has been studied in relation to socioeconomic factors, such as income, employment,²² and retirement status,²⁴ it has not yet, to our knowledge, been examined alongside income growth, an important marker of economic potential.^25,26 Second, by relying on data from the HCHS, this project offers a much-needed analysis of how HL may be impacting Hispanics, a rapidly growing fraction of the United States. According to 2020 Census data, the U.S. Hispanic population is now the second largest racial or ethnic group in the country and is also one of the fastest growing, increasing in size by 23% from 2010 to 2020.^27,28 By building on previous work analyzing the HCHS dataset,^6,7,12,29 this study seeks to further characterize the socioeconomic environment among adults with HL in a dynamic population that is increasingly representative of the broader country. We hypothesize that the presence of HL is associated with decreased income growth.

Methods

Participants

The sample population was drawn from the HCHS, a large multicentered, U.S. national epidemiologic study focused on prevalence and development of disease. Participants are self-identified U.S. Hispanics from the ages of 18 to 74 years old living in Chicago, Miami, San Diego, and the Bronx borough of New York City. The present study included data from wave 1 (2008–2011) and wave 2 (2014–2017) of the HCHS. The HCHS contains a variety of data, including comprehensive physical exams, an assortment of surveys querying participants’ medical histories and socioeconomic details, and audiometric hearing tests. HCHS participants with complete audiometric data, socioeconomic data (as measured on income surveys), and covariate data were included in the cohort for analyses. The process of participant inclusion and exclusion is shown in Figure 1.

Figure 1.

Process of Participant Inclusion and Exclusion

Abbreviation: HCHS, Hispanic Community Health Study.

¹Extra covariates refer to other covariates that were included as part of a larger pre-specified analysis but not the final study.

Hearing (Primary Exposure)

Hearing was assessed according to HCHS protocols in a soundproof booth following American Speech-Language Hearing Association (ASHA) guidelines.^30,31 Unaided (no assistive devices, such as hearing aids) and sided (right vs. left) pure-tone thresholds were recorded from 0.5 to 8 kilohertz (kHz) and measured in decibels (dB). HL was operationalized as the pure-tone average (PTA) calculated from 4 frequencies: 0.5, 1, 2, and 4 kHz. The use of the 4-frequency PTA as a measure of hearing health is supported in prior epidemiologic studies of HL and the subject of an ongoing public health campaign.³² The PTA in the better hearing ear from wave 1 was used for analysis. HL was assessed continuously and defined categorically as follows: normal hearing (−20 to 25 dB), mild HL (26 to 40 dB), moderate HL (41 to 55 dB), moderately severe HL and above (>55 dB).

The HCHS does not contain information on the etiology of a participant’s HL, so we are unable to differentiate between congenital, genetic, or secondary causes of HL. The HCHS is an epidemiologic database that recruited roughly 16,000 healthy Hispanic volunteers. As such, we can predict that the prevalence of the different etiologies of HL in HCHS will be similar to the prevalence of different HL etiologies in the general population.

Income / Socioeconomic Measures (Outcome)

The primary outcome was grouped yearly household income, measured in U.S. dollars per year. The combined income of everyone in the household (including the participant as well as both family and non-family members) was recorded for this measure. All income sources³³ were included in this amount. While a year was considered to be any 12-month period, participants were encouraged to enter an amount reflective of the previous income tax year from January 1 to December 31.

Income was graded on a 10-point scale of income brackets ranging from below $10,000 per year to above $100,000 per year as follows: less than $10,000; $10,001–$15,000; $15,001–$20,000; $20,001–$25,000; $25,001–$29,999; $30,000–$40,000; $40,001–$50,000; $50,001–$75,000; $75,001–$100,000; and more than $100,000. Income data from waves 1 and 2 were used for analysis.

Covariates

For our longitudinal model we controlled for the following potential confounders: age, gender, education (total years of schooling completed), study site (Chicago, Miami, San Diego, or the Bronx), hearing aid usage in the past 12 months (yes/no), and a composite vascular risk score. The model also included elapsed time between visits 1 and 2. This vascular risk score was created by adding component scores for hypertension (scored as 1 if the participant’s systolic or diastolic BP was greater than or equal to 140/90 or if the participant self-reported as taking antihypertensive mediations, or 0 if neither of these conditions was met) and diabetes (scored as 0 for normal glucose regulation, 1 for impaired glucose tolerance, or 2 for diabetes, based on fasting glucose testing and, if available, oral glucose tolerance and A1C testing). Vascular risk scores ranged from 0 to 3, with higher scores indicating higher vascular risk. Excluding the elapsed time variable, all data for these covariates were collected during wave 1.

Statistical Analysis

A basic demographic analysis was performed to assess the makeup of the study sample after applying the inclusion and exclusion criteria. This analysis consisted of calculating frequencies (number and percent) for all categorical variables as well as the mean and standard deviation for all continuous variables.

Generalized estimating equations (GEE) with an ordinal logistic link, controlling for the covariates described above, were used to analyze the longitudinal association between HL (exposure) and income growth (outcome). The GEE model was chosen because of its ability to estimate population trends while accounting for within-subject variance, whereas the ordinal logistic link was chosen due to income’s status as an ordinal variable.³⁴ Data was reported with 95% confidence intervals and significance was defined as p≤0.05.

Preparation and basic demographic analysis of this data was conducted in R version 4.2.3 (R Foundation for Statistical Computing, Vienna, Austria) in RStudio version 2023.06.1+524 (Posit, Boston, MA). Longitudinal analysis, as described above, was performed in R version 4.2.2 and SPSS Statistics version 25 (IBM Corp, Armonk, NY). Preparation and analysis were performed in July 2023. The Columbia University Institutional Review Board (IRB) deemed this study of anonymous data to not be human subjects research and thus exempt from IRB approval. This study met the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) criteria checklist.³⁵

Results

Baseline Demographic Characteristics

1,342 participants met inclusion criteria, of whom 778 (58.0%) were female. From wave 1, the average age (SD) was 47.6 (12.2) years, with a range of 18 to 74 years. The average time (SD) between the first and second wave visits was 5.9 (0.6) years. For an overview of the basic demographic characteristics of the dataset, please refer to Table 1.

Table 1.

Participant Characteristics in Wave 1 in the Hispanic Community Health Study (n = 1,342)

Participant Characteristic	No. (%) or Mean (SD)
Age, mean (SD)	47.6 years (12.2)
18–40 years	324 (24.1%)
41–64 years	926 (69.0%)
65 years and older	92 (6.9%)
Gender, no. (%)
Women	778 (58.0%)
Men	564 (42.0%)
HL Severity, mean (SD)	13.9 dB (9.5)
Normal, −20 to 25 dB	1208 (90.0%)
Mild, >25 to 40 dB	109 (8.1%)
Moderate, >40 to 55 dB	19 (1.4%)
Moderately Severe and Worse, >55 dB	6 (0.4%)
Years of Education, mean (SD)	11.7 years (4.0)
Hearing Aid Usage in Past 12 Months, no. (%)
Yes	1334 (99.4%)
No	8 (0.6%)
Site, no. (%)
Bronx	348 (25.9%)
Chicago	323 (24.1%)
Miami	350 (26.1%)
San Diego	321 (23.9%)
Vascular risk score, no. (%)
0	444 (33.1%)
1	451 (33.6%)
2	318 (23.7%)
3	129 (9.6%)

Abbreviation: HL, hearing loss.

Baseline Hearing Characteristics

As described in Table 1, the average PTA (SD) in the better hearing ear using wave 1 data was 13.9 (9.5) dB, with a range of −3.75 to 72.5 dB. 1,208 participants (90.0%) had normal hearing (-20 to 25 dB), 109 participants (8.1%) had mild HL (26 to 40 dB), 19 participants (1.4%) had moderate HL (41 to 55 dB), and 6 (0.4%) had moderately severe HL (>55 dB).

Baseline Income Growth Characteristics

In wave 1, 208 participants (15.5%) had a yearly income less than $10,000, 191 participants (14.2%) had an income between $30,000–$40,000, and 27 (2.0%) had an income more than $100,000. See Table 2 for a comprehensive description of participant income across waves 1 and 2.

Table 2.

Distribution of Income Across Participants in Waves 1 and 2.

Income (U.S. dollars per year)	Wave 1 n (%)	Wave 2 n (%)
Less than $10,000	208 (15.5%)	150 (11.2%)
$10,001–$15,000	255 (19.0%)	234 (17.4%)
$15,001–$20,000	175 (13.0%)	138 (10.3%)
$20,001–$25,000	167 (12.4%)	114 (8.5%)
$25,001–$29,999	112 (8.3%)	129 (9.6%)
$30,000–$40,000	191 (14.2%)	231 (17.2%)
$40,001–$50,000	91 (6.8%)	129 (9.6%)
$50,001–$75,000	79 (5.9%)	113 (8.4%)
$75,001–$100,000	37 (2.8%)	60 (4.5%)
More than $100,000	27 (2.0%)	44 (3.3%)

Generalized Estimating Equations Analysis

A longitudinal model accounting for potential confounders (age, gender, education, study site, hearing aid usage, vascular risk) and including elapsed time between visits was run for the exposure (HL severity as defined by PTA) and outcome (grouped yearly household income) using GEE with an ordinal logistic link across both waves of data. The effect of time on income at a particular HL severity can be estimated from $\frac{\Delta log}{\Delta t}=0.081-0.0021*HL$, where $t$ is the time between waves 1 and 2 and $HL$ is the PTA in the better hearing ear.

There was a significant time*HL interaction (p<0.001): with each 10 dB worsening in HL, the odds of an individual belonging to a higher versus a lower income group changed by a factor of 0.979 between waves 1 and 2. More specifically, for individuals with no HL, the odds of belonging to a higher versus a lower income bracket increased by a factor of 1.08 across the time between waves 1 and 2. For individuals with 10 dB HL, the odds of an individual belonging to a higher versus a lower income group increased by a factor of 1.06 across the time between waves 1 and 2. The difference between these two factors (between 1.08 and 1.06), and between all adjacent odds ratios, was statistically significant (p<0.001 on interaction test). Table 3 reports the multivariable odds ratios for income growth stratified by deciles of HL severity. Of note, 38.6 dB HL is the point at which income transitions from growth to decline (defined by the odds ratio becoming <1). The odds of an individual belonging to a higher versus a lower income bracket across the time between waves 1 and 2 increased by a factor of 1.018 for those with 30 dB HL, but at 40 dB HL, the odds of an individual belonging to a higher versus a lower income bracket changed by a factor of 0.997 during that time, representing a decrease in income.

Table 3.

Multivariable Odds Ratios for Income Growth Stratified by HL Severity

HL Severity (in dB)	Odds Ratio for Income Growth1,2
0	1.084
10	1.062
20	1.040
30	1.018
40	0.997
50	0.976
60	0.956
70	0.936
80	0.917
90	0.898

Abbreviation: HL, hearing loss.

¹Odds ratios indicate the odds of belonging to a higher versus a lower income group across time between waves 1 and 2, controlling for age, gender, education, study site, hearing aid usage, vascular risk, and elapsed time between visits. Odds ratio > 1 indicates income growth. Odds ratio < 1 indicates income decline.

²p<0.001 for the difference between any two adjacent odds ratios

Discussion

In this longitudinal analysis across waves 1 and 2 of the HCHS cohort, HL was associated with decreased odds of belonging to a higher income bracket over a mean of 6 years. For individuals with HL worse than 38.5 dB, income tended to decrease, rather than increase, over time. These associations persisted despite controlling for the confounders of age, gender, education, study site, hearing aid usage, and vascular risk.

This paper’s findings are statistically, clinically, and socioeconomically significant. In general, as individuals aged, income tended to increase within our cohort during the study period. However, as we examine participants with increasingly severe levels of HL, the odds of a participant belonging to a higher versus a lower income group steadily decrease. Between the two waves, for each 10 dB worsening in HL, the odds of an individual belonging to a higher versus a lower income group decreased by approximately 2%. The effect of this value becomes more striking as we compare across varying degrees of HL (Table 3). Interestingly, beginning in levels of mild HL at 38.5 dB, income change over time shifts direction—individuals tended to experience reductions in income. For example, an individual with normal hearing would be expected to have an increase of 8% in income over the course of the study period, while an individual with a moderate 50 dB HL would be expected to have a 2% decrease in income over that time.

This study is novel in two key ways. First, to our knowledge, this is the first analysis looking at HL as the exposure and income growth as the outcome. HL has been studied with income and other socioeconomic factors (i.e., education, employment) in the literature. But this study offers a look at HL’s impact on a participant’s chances to achieve income growth, a critical measure for economic potential. Second, it represents the first longitudinal analysis examining both HL and socioeconomic factors together in the HCHS cohort. This longitudinal analysis offers more causal evidence than a cross-sectional study that HL (the primary exposure), is a potential cause for the decrease in income growth over time (the outcome). Although directionality and causality of the relationship between HL and income growth cannot be determined without a randomized controlled trial, longitudinal studies offer stronger evidence than cross-sectional analyses in examining this relationship. This study suggests HL may be an important factor in determining one’s socioeconomic resources over time.

Though we cannot infer causation in this observational study, there are multiple possible mechanisms that could explain the findings. For one, HL is associated with decreased socialization, specifically with less social support, smaller and less diverse social networks, and less family cohesion, including in the HCHS cohort.²⁹ Diminished social support may make it difficult for people to obtain or retain work for income. It may also decrease the likelihood that someone is able to seek assistance from other sources of funding or other social programs, thereby also decreasing yearly income as defined in this study.

This analysis contributes to the robust body of literature exploring the link between HL and various deleterious conditions. There is strong evidence supporting the association of HL with conditions such as depression, cognitive decline, and dementia, including at the meta-analysis level.^36,37 Other studies using the HCHS cohort have similarly shown significant associations between HL and depressive symptoms, despair or negative emotional states, cognition, socialization, schizophrenia or serious mental illness, stress, and substance use.^{6,12,29,38–42} Though this study is the first, to our knowledge, to analyze the association between HL and income growth, HL has also been shown to have independent associations with other socioeconomic factors, including unemployment, reduced school performance, and earlier retirement.²¹ In short, HL should not be considered solely a concern for worse communicative abilities or quality of life. However, the extent to and process by which HL and socioeconomic dysfunction are interlinked remains to be fully understood.

This study has several limitations. First, while this study is longitudinal and provides a temporal directionality (i.e., HL occurred before the change in income), it is still observational. Therefore, while this investigation does identify an association between HL and income, causation cannot be truly inferred, meaning that this study cannot conclude that HL causes reduced income growth over time. Although we adjusted for some of the most common and important confounders for HL, such as age, gender, and cardiovascular risk, we cannot control for all possible confounders, particular those factors yet to be recognized as confounders. Future research should include quantitative metrics for other potential confounders such as health insurance coverage or occupation. Only randomized controlled trials can eliminate confounding and truly assess causality.

Second, we are limited in our definitions of the outcome measures, i.e., income or income growth. Income is not the same as wealth. For example, one person could have low income due to limited socioeconomic resources, while another could have low income for the opposite reason—they possess an abundance of socioeconomic resources, allowing them to forgo employment and other income-seeking activities. Notably, such individuals would likely receive income through more passive means, such as rent or capital payments, that theoretically should be reflected in the HCHS income measure. The number of income contributors in a household can also affect the true wealth of a household. This analysis considers a single income value for all contributors in a household. However, it is worth noting that an income of $100,000 can represent different socioeconomic scenarios depending on the number of income-earners required to achieve that sum. Furthermore, the HCHS only contains categorical income data, i.e. income brackets, rather than continuous yearly income. Future waves of the HCHS and other national cohorts can consider recording continuous income data to allow for further analyses, such as calculating mean income amounts, and more precisely track the relationship between medical conditions and socioeconomic measures over time.

Finally, although this study’s focus on the U.S. Hispanic population is a strength of the project, such focus may limit the generalizability of these findings to the national population or international populations of interest. The recruitment and study sites of the HCHS are also primarily urban environments, which may limit generalizability to suburban or rural locations. However, HL appears to negatively impact one’s income or income growth in other study populations as well.^21,43,44 Our work supports the prevailing idea in the literature that HL can have negative socioeconomic implications.

Future epidemiological studies should aim to include both income and income growth if studying social determinants of health. Particularly for studying HL, it will be important to continue examining its relationship not just to medical conditions, but socioeconomic ones as well. Future studies should also examine additional waves of the HCHS, which are ongoing. This will allow for analyses across longer time intervals and may show larger effect sizes for the relationships between HL and socioeconomic factors like income growth. Longer time intervals could also allow examination of disease states with long latency such as dementia.⁴⁵

In conclusion, HL is not just associated with medical conditions such as dementia or depression. HL is also associated with factors across the social determinants of health, including lower income growth over time. Among those with moderate or worse HL, we observed income decline rather than growth. Our findings may have implications on the ability and motivation of people to access appropriate hearing screenings, treatment, and healthcare. Further research is required to understand the mechanisms and causality of this relationship between HL and income growth, to better improve the health and well-being of those living with HL.