Hearing Difficulties and Tinnitus in Construction, Agricultural, Music, and Finance Industries: Contributions of Demographic, Health, and Lifestyle Factors.

High levels of occupational noise exposure increase the risk of hearing difficulties and tinnitus. However, differences in demographic, health, and lifestyle factors could also contribute to high levels of hearing difficulties and tinnitus in some industries. Data from a subsample (n = 22,936) of the U.K. Biobank were analyzed to determine to what extent differences in levels of hearing difficulties and tinnitus in high-risk industries (construction, agricultural, and music) compared with low-risk industries (finance) could be attributable to demographic, health, and lifestyle factors, rather than occupational noise exposure. Hearing difficulties were identified using a digits-in-noise speech recognition test. Tinnitus was identified based on self-report. Logistic regression analyses showed that occupational noise exposure partially accounted for higher levels of hearing difficulties in the agricultural industry compared with finance, and occupational noise exposure, older age, low socioeconomic status, and non-White ethnic background partially accounted for higher levels of hearing difficulties in the construction industry. However, the factors assessed in the model did not fully account for the increased likelihood of hearing difficulties in high-risk industries, suggesting that there are additional unknown factors which impact on hearing or that there was insufficient measurement of factors included in the model. The levels of tinnitus were greatest for music and construction industries compared with finance, and these differences were accounted for by occupational and music noise exposure, as well as older age. These findings emphasize the need to promote hearing conservation in occupational and music settings, with a particular focus on high-risk demographic subgroups.

Introduction

Hearing difficulties and tinnitus are common risks associated with professions where there are high levels and long durations of noise exposure such as in construction, agricultural, and music industries (Nelson, Nelson, Concha-Barrientos, & Fingerhut, 2005; Sliwinska-Kowalska & Davis, 2012; Stucken & Hong, 2014; Tak, Davis, & Calvert, 2009). Between 7% and 21% of hearing loss in adults is attributable to occupational noise (Dobie, 2008; Lie et al., 2016; Nelson et al., 2005), with noise-induced hearing loss (NIHL) being one of the most common occupational diseases (Fausti, Wilmington, Helt, Helt, & Konrad-Martin, 2005; Le, Straatman, Lea, & Westerberg, 2017; Themann, Suter, & Stephenson, 2013). Compared with 7% of nonnoise exposed workers, 23% of U.S. workers exposed to occupational noise reported having a hearing impairment, while 15% of noise exposed workers reported tinnitus compared with 5% of nonexposed workers (Masterson, Themann, Luckhaupt, Li, & Calvert, 2016). With respect to specific industries, 25% of noise exposed U.S. construction workers (Masterson, Deddens, Themann, Bertke, & Calvert, 2015), 65% to 67% of farmers (Kerr, McCullagh, Savik, & Dvorak, 2003; Marvel, Pratt, Marvel, Regan, & May, 1991), and 37% to 58% of classical musicians (for a review, see Zhao, Manchaiah, French, & Price, 2010) were found to have hearing impairments based on audiometric data. Jansen, Helleman, Dreschler, and de Laat (2009) showed that approximately half of the musicians in their study complained of tinnitus, and Hong (2005) demonstrated that 38% of construction workers reported experiencing tinnitus.

The current permissible noise limit in the United Kingdom is 85 dBA for 8-hr duration, which is reduced by half for each 3 dBA increase in noise intensity (Control of Noise at Work Regulations, 2005; for similar regulations, also see National Institute of Occupational Safety and Health [NIOSH], 1998). Typical average noise equivalent exposures (based on an 8-hr time-weighted average) for construction workers are 72 to 112 dBA (Fernández, Quintana, Chavarría, & Ballesteros, 2009; Neitzel, Seixas, Camp, & Yost, 1999; Neitzel, Stover, & Seixas, 2011; Seixas et al., 2012), 75 to 98 dBA for classical musicians (McBride et al., 1992; Pawlaczyk-Łuszczyńska, Dudarewicz, Zamojska, & Śliwinska-Kowalska, 2011; Royster, Royster, & Killion, 1991; Schmidt et al., 2011), and 62.6 to 92.1 dBA for agricultural workers (McBride, Firth, & Herbison, 2003; Milz, Wilkins, Ames, & Witherspoon, 2008; Williams et al., 2015). However, while exposure to high levels of noise may be the primary cause of hearing difficulties and tinnitus, demographic, health, and lifestyle factors may also be a contributing factor (Agrawal, Platz, & Niparko, 2008; Cruickshanks et al., 2003; Cruickshanks, Klein, et al., 1998; Cruickshanks, Wiley, et al., 1998; Ferrite & Santana, 2005; Heller, 2003). For example, ageing is a well-documented risk factor for hearing difficulties (Cruickshanks, Wiley, et al., 1998; Davis et al., 2016; Dubno, Dirks, & Morgan, 1984; Lin et al., 2013; Patterson, Nimmo‐Smith, Weber, & Milroy, 1982) and tinnitus (Negrila-Mezei, Enache, & Sarafoleanu, 2011; Rosenhall, 2003; Sanchez, 2004). Other demographic risk factors include socioeconomic status (SES), ethnicity, education level, cognition, genetics, and gender, while modifiable health and lifestyle risk factors include smoking, lack of exercise, poor diet, diabetes, and cardiovascular disease, for both hearing difficulties (e.g., Agrawal et al., 2008; Daniel, 2007; Dawes, Cruickshanks, et al., 2014; Moore et al., 2014) and tinnitus (e.g., Heller, 2003; Hoffman & Reed, 2004; Sanchez, 2004). Noise exposure may also interact with demographic, health, and lifestyle factors (Ferrite & Santana, 2005; Mizoue, Miyamoto, & Shimizu, 2003). For example, Ferrite and Santana (2005) estimated that the combined effects of age, smoking, and noise exposure on hearing loss were greater than the sum of effects from each factor.

Determining the impact of noise exposure in relation to hearing difficulties and tinnitus among high-risk industries is problematic because demographic, health, and lifestyle factors that impact on hearing may also vary between industries. For instance, one might predict a higher proportion of males and higher levels of physical activity in agricultural and construction industries compared with finance. Furthermore, it is unknown whether music noise exposure also contributes to hearing difficulties and tinnitus in high-risk industries, and whether certain industries are also more exposed to music noise than others. For example, Tufts and Skoe (2018) showed that musicians had higher levels of both occupational and recreational music noise exposure than nonmusicians, and that recreational music noise exposure contributed more to some musicians’ total noise exposure than occupational musical activities. Therefore, it is important to consider music noise exposure as well as other high-risk demographic, health, and lifestyle factors when assessing hearing function in high-risk groups. Understanding the contributions of occupational noise, recreational (i.e., music) noise, demographic, health, and lifestyle factors to hearing health could have important implications for developing interventions to promote healthy hearing. For example, hearing conservation program for both occupational (see Verbeek, Kateman, Morata, Dreschler, & Mischke, 2014) and recreational (see Zhao, French, Manchaiah, Liang, & Price, 2012) settings could be combined with interventions that target modifiable health and lifestyle risk factors for hearing health, which may be particularly relevant for certain industries or demographic subgroups.

In this study, we aimed to evaluate whether demographic, health, and lifestyle factors including music noise exposure could account for the excess risk of hearing difficulties and tinnitus in industries at high risk of hearing problems due to occupational noise exposure (construction, agricultural, and music industries) compared with an industry with low levels of occupational noise exposure (finance).

Method

Participants

In this retrospective cross-sectional study, the participant sample was drawn from the U.K. Biobank resource for epidemiology, which contains data for more than 500,000 persons aged 40 to 69 years across England, Scotland, and Wales (Collins, 2012). All data were collected between 2006 and 2010. Individuals registered with the National Health Service and living within a 25-mile radius of one of 22 U.K. Biobank assessment centers were invited to participate (Allen et al., 2012). Participation involved completion of touch-screen questionnaires and nurse-led interviews to collect epidemiological data on demographic, health, environmental, and lifestyle factors. Measures of hearing function and tinnitus questions (described later) were added part way through data collection, and so hearing function and tinnitus data are available for a subset of the U.K. Biobank sample (see Table 1). If participants reported that they were currently employed or self-employed, the interviewer was prompted to ask the participant to describe their current occupation. The occupation title was coded according to the U.K. Standard Occupational Classification (Office for National Statistics, 2000; see Supplementary Materials). Industries of interest were construction (classified as machinery, construction plant, or building work; n = 9,249), agricultural (classified as farming or agriculture; n = 2,081), music (classified as a performing musician or director or conductor; n = 395), and finance (classified as banking or finance; n = 11,211; total: n = 22,936; see Table 1). Note that previous employment history was not captured by the U.K. Biobank questionnaires, and inclusion was based on current employment status; unemployed and retired individuals were not included in the analysis.

Table 1.

Hearing Difficulties, Tinnitus, Demographic, Health, and Lifestyle Factors for the Overall Subsample, and Split by Agricultural, Construction, Music, and Finance Industries.

Factors	N participants	Overall (N = 22,936)	Agricultural (N = 2,081)	Construction (N = 9,249)	Music (N = 395)	Finance (N = 11,211)	p
Hearing difficulties[a]	9,035	10.70%	14.36%	13.40%	4.88%	8.26%	<.001
Tinnitus, yes[a]	9,312	17.29%	18.59%	20.60%	23.95%	14.26%	<.001
Age[b]	22,936	53.94 (7.87)	55.10 (7.98)	54.47 (7.95)	54.88 (7.89)	53.25 (7.71)	<.001
Gender, male[a]	22,936	72.29%	72.47%	97.09%	59.24%	52.26%	<.001
Ethnicity, White[a]	22,832	95.34%	98.60%	96.12%	97.96%	93.99%	<.001
Townsend quartile[a]	22,909						<.001
Most affluent		31.98%	28.01%	29.06%	22.03%	35.48%
Second quartile		26.61%	30.13%	26.25%	18.23%	26.55%
Third quartile		22.27%	22.52%	22.86%	28.35%	21.52%
Most deprived		19.14%	19.35%	21.83%	31.39%	16.45%
Work noise exposure, high[a]	9,390	27.53%	36.29%	50.12%	4.96%	8.10%	<.001
Music noise exposure, high[a]	9,384	14.79%	15.03%	18.64%	46.75%	10.61%	<.001
Smoking status[a]	22,843						<.001
Current smoker		11.84%	12.31%	14.87%	12.47%	9.24%
Ex-smoker		33.32%	32.38%	36.43%	32.06%	30.97%
Nonsmoker		54.84%	55.31%	48.70%	55.47%	59.79%
Alcohol consumption, yes[a]	22,895	95.12%	93.50%	95.57%	93.15%	95.12%	<.001
Ototoxic medication, yes[a]	22,936	36.75%	36.47%	37.03%	30.13%	36.79%	.05
Physical activity, inactive[a]	10,113	28.23%	14.60%	26.96%	25.11%	30.83%	<.001
Body mass index[b]	22,618	27.54 (4.36)	26.81 (4.24)	28.07 (4.09)	25.94 (4.40)	27.29 (4.54)	<.001
Diabetes, yes[a]	22,936	4.11%	3.27%	4.52%	3.04%	3.97%	.02
Cardiovascular disease, yes[a]	22,936	7.39%	8.31%	8.58%	4.81%	6.32%	<.001
Hypertension, yes[a]	22,936	54.36%	53.87%	60.64%	43.04%	49.67%	<.001
High cholesterol, yes[a]	22,936	15.21%	13.41%	17.19%	9.37%	14.12%	<.001

Note. Comparisons between job categories were conducted. Pearson’s χ2 tests for categorical variables and one-way analysis of variance for continuous variables (α = .001). SD = standard deviation.

aDescriptive statistics for all categorical (%) variables.

bDescriptive statistics for all continuous (mean ± SD) variables.

U.K. Biobank procedures were approved by the North West Multi-centre Research Ethics Committee, and all participants provided written informed consent.

Procedures

Hearing function

Participants completed a shortened version of the digits-in-noise (DIN) test; a test of speech recognition in background noise, which correlates strongly with pure-tone audiometric thresholds (Smits, Kapteyn, & Houtgast, 2004). For the U.K. Biobank version of the DIN test, 15 sets of English monosyllabic digits were presented. Each ear was tested separately, with the order of testing randomized between participants. Stimuli were presented via circumaural headphones (Sennheiser HD-25) at a comfortable level set by participants. Digit triplets (e.g., 2–5–3) were presented against a background of speech-shaped noise matched to the complete set of the nine monosyllabic digits (0–9, “0” was/oh/, and 7 was excluded). Participants were required to identify and key in the three digits via a touch-screen interface. The sound level of the background noise varied adaptively after each presentation depending on whether all three digits were correctly identified. The speech reception threshold was estimated based on the average signal-to-noise ratio (SNR) from the last eight triplets. The SNR varied between −12 and +8 dB, with more positive scores indicating poorer speech hearing ability. Hearing difficulties were identified based on the performance of the better ear relative to a reference group of normal hearing participants aged 18 to 29 years (Dawes, Fortnum, et al., 2014), such that participants were classified as having normal hearing for SNR values below −5.5 dB or having hearing difficulties for SNR values above −5.5 dB. Difficulty hearing in background noise is the most common hearing complaint (Pienkowski, 2017); therefore, the assessment of hearing function using a test of number recognition in background noise represents an ecologically valid and objective index of hearing.

Tinnitus

Participants were asked “Do you get or have you had noises (such as ringing or buzzing) in your head or in one or both ears that lasts more than 5 minutes at a time?” with a choice of responses: (a) Yes, now most or all of the time; (b) Yes, now a lot of the time; (c) Yes, now some of the time; (d) Yes, but not now, but have in the past; (e) No, never; (f) Do not know; or (g) Prefer not to answer. The presence of tinnitus was characterized by participants currently having symptoms at least “now some of the time.”

Demographic factors

Demographic factors included age, sex, SES, and ethnicity. Ethnicity was coded according to White or non-White ethnic groups based on U.K. 2001 census categories (see U.K. Biobank protocol; http://www.ukbiobank.ac.uk/). The Townsend index was used as a proxy measure of SES; a composite measure of unemployment, non car ownership, non home ownership, and household overcrowding based on the area of residence (Townsend, Phillimore, & Beattie, 1988). The Townsend index is expressed in terms of a z-score. These scores were categorized into quartiles, with the lowest scores or quartile representing more affluent status.

Health and lifestyle factors.

Body mass index and physical activity

Body mass index (BMI) was calculated as weight (kg) divided by height (m2). Physical activity was measured using the question “Yesterday, about how long did you spend doing activities that needed moderate effort, making you somewhat out of breath? For example, walking upstairs, going to the gym, jogging, energetic dancing, aerobics, most sports, using heavy power tools, and other physically demanding DIY and gardening.” Participants were classified as inactive if they reported doing less than 10 minutes moderate activity, and active for any other lengths of moderate activity over 10 minutes.

Smoking and alcohol consumption

Smoking status was based on the responses to two questions “Do you smoke tobacco now?” and “In the past, how often have you smoked tobacco?,” with participants classified as current smoker (currently smoking occasionally or on most or all days), ex-smoker (previously smoking occasionally or on most or all days), or nonsmoker (never smoked or reported just having tried smoking once or twice). Alcohol consumption was classified on the basis of participants responses to the question “About how often do you drink alcohol?,” with nondrinkers identified as having “never” drank alcohol, and drinkers identified on the basis of the remaining options (“special occasions only,” “one to three times a month,” “once or twice a week,” “three or four times a week,” “daily or almost daily”).

Diabetes, cardiovascular disease, high cholesterol, hypertension, and ototoxic medication

Diabetes was identified if the participant reporting having either Type 1 or Type 2 diabetes or reported taking medication for diabetes. Cardiovascular disease included reporting having angina, heart attack, heart failure, stroke, transient ischemic attack, intermittent claudication, arterial embolism, or deep venous thrombosis. High cholesterol was based on the participant self-reporting that they had high cholesterol or if they were taking medication for high cholesterol. Hypertension was identified if the participant self-reported having high blood pressure, took medication for high blood pressure, or if measured systolic blood pressure exceeded 140 mm Hg or diastolic pressure greater than 90 mm Hg. Ototoxic medication use was identified based on currently taking on a regular basis (i.e., daily, weekly, or monthly), but not a course of (i.e., 1 week only), of any medication with a known ototoxic effect, including loop diuretics, aminoglycoside antibiotics, quinine derivatives, nonsteroidal anti-inflammatories, and salicylates.

Noise exposure

Occupational noise exposure was assessed by the question “Have you ever worked in a noisy place where you had to shout to be heard?” Music noise exposure was used as a proxy measure of recreational noise exposure and was assessed by the question “Have you ever listened to music for more than 3 hours per week at a volume which you would need to shout to be heard or, if wearing headphones, someone else would need to shout for you to hear them?” Both questions had the following response options: (a) Yes, for more than 5 years; (b) Yes, for around 1 to 5 years; (c) Yes, for less than 1 year; (d) No; (e) Do not know; or (f) Prefer not to answer. Participants were categorized as high exposure if they answered “Yes,” irrespective of how long they were exposed, and low if they answered “No,” for both occupational and music noise exposure separately. This minimum criterion for both occupational- and music-related noise corresponds to exposure exceeding 85 dBA for 8 hr/day (Control of Noise at Work Regulations, 2005), thus ensuring the we were capturing individuals with high levels of noise exposure.

Further information about the assessment procedure can be found at that U.K. Biobank website (http://www.ukbiobank.ac.uk/).

Data Analysis

Analyses were performed using IBM SPSS version 22. The first part of the analysis compared between the industries of interest (Table 1). One-way analysis of variance were used to compare the effect of age and BMI, with industry type as a single between-subjects factor, and Pearson χ2 tests were used to compare all other categorical variables, including levels of hearing difficulties and tinnitus. Note that there were missing data for some measures (Table 1), most likely due to these measures being added into the U.K. Biobank protocol after data collection for hearing difficulties and tinnitus measures had started, or participants not completing certain questions.

Logistic regression analyses were conducted separately for hearing difficulties and tinnitus to model the effects of industry type, demographic factors (age, sex, SES, and ethnicity), health, and lifestyle factors (BMI, smoking status, alcohol consumption, diabetes, cardiovascular disease, hypertension, cholesterol, physical activity, and ototoxic medication) and noise exposure (occupational and music). In the first iteration of each of these regression models, only industry type was included in the analysis. Demographic factors were then added into the second iteration of the models, and health and lifestyle factors into the third iteration. Occupational and music noise exposures were then factored into the fourth and final iterations of each of the models. With each new iteration, the step change in the omnibus test of model coefficients (χ2) determined whether the additional factors led to a significant improvement in the model prediction.

For each iteration of the logistic regression model, an overall measure of the percentage accuracy in classification for hearing difficulties and tinnitus is provided. The amount of variance explained by each model iteration (equivalent to R2 in multiple regression) is given by a pseudo-R2 value (Nagelkerke R2), with higher values indicating that a larger proportion of the variance of the dependent variable can be accounted for by the independent (predictor) variables.

Logistic regression models provide odds ratios (ORs), which are measures of the likelihood of the dependent variable (e.g., hearing difficulties) occurring for a given independent variable (e.g., industry type). An OR greater than 1 indicates an increased likelihood of the dependent variable occurring for a given independent variable, whereas an OR less than 1 indicates a decreased likelihood of the dependent variable occurring. For categorical variables, one group within each category is chosen as the control group with which to compare all other categories. For example, for the analysis of industry type, the finance industry was selected as the control group to compare with the other high-risk industries. Wald χ2 tests provide a measure of whether the OR is significantly greater than—or less than—1 at the level of α = .05. If the 95% confidence interval (CI) for the OR includes the value of 1, then the Wald χ2 statistic will be nonsignificant. For each iteration of the regression model, a convergence of the OR toward 1 (i.e., toward nonsignificance) for high-risk industries versus finance indicates that other predictor variables (i.e., demographic, health and lifestyle, noise exposure) are accounting for the levels of hearing difficulties or tinnitus between these industries.

Results

Descriptive Statistics

For the comparison between industries, there were significant differences for all variables (p < .001; Table 1), except ototoxic medications (p = .05) and diabetes (p = .02; Bonferroni-corrected α level for multiple corrections = .001). For example, the agricultural industry had the oldest workers on average, had the highest proportion of people reporting White ethnic background, and had the highest levels of hearing difficulties. The construction industry had the highest proportion of males, the highest levels of chronic health conditions, and the highest levels of occupational noise exposure. Hearing difficulties were relatively low in the music industry, but also had the lowest SES (i.e., least affluent), highest levels of music noise exposure, and the highest levels of tinnitus. The finance industry had the highest proportion of females, the highest SES (i.e., most affluent), and the lowest levels of noise exposure.

Hearing Function

In the first iteration of the logistic regression model (Table 2), the overall model prediction was significant, χ2(3) = 27.83, p < .001. The model explained 1.11% of the variance (Nagelkerke R2) in hearing difficulties, and the overall correct classification for hearing difficulties was 91.24%. Industry type was a significant correlate of hearing difficulties (p < .001). Agricultural (OR: 1.63, 95% CI [1.18, 2.27]) and construction industry workers (OR: 1.60, 95% CI [1.31, 1.94]) were more likely to have hearing difficulties compared with finance. There was no significant difference between music and finance industries (OR: 0.59, 95% CI [0.24, 1.46]).

Table 2.

Logistic Regression Model for Hearing Difficulties With Industry Type, Demographic, Health, and Lifestyles Factors, and Noise Exposure.

Factors	Step 1				Step 2				Step 3				Step 4
	χ2(3) = 27.83, p < .001; R2 = .01				χ2(6) = 251.51, p < .001; R2= .11				χ2(10) = 11.99, p =  .29; R2 = .11				χ2(2) = 6.55, p = .04; R2 = .12
	Wald χ2	OR	95% CI		Wald χ2	OR	95% CI		Wald χ2	OR	95% CI		Wald χ2	OR	95% CI
			Lower	Upper			Lower	Upper			Lower	Upper			Lower	Upper
Industry type
Industry (vs. finance)	27.31***	–	–	–	24.17***	–	–	–	23.67***	–	–	–	16.26**	–	–	–
Agricultural	8.53**	1.63	1.18	2.27	11.23***	1.79	1.27	2.52	11.38***	1.81	1.28	2.55	8.57**	1.68	1.19	2.38
Construction	21.59***	1.60	1.31	1.94	15.78***	1.58	1.26	1.97	15.04***	1.56	1.25	1.96	8.03**	1.42	1.11	1.80
Music
	1.32	.059	0.24	1.46	1.44	0.57	0.23	1.43	1.44	0.57	0.23	1.43	2.05	0.51	0.20	1.28
Demographic factors
Age					173.46***	1.10	1.08	1.11	147.30***	1.10	1.08	1.11	148.03***	1.10	1.08	1.12
Gender (vs. female)					0.44	0.92	0.72	1.18	0.38	0.93	0.72	1.19	0.94	0.88	0.69	1.14
Ethnicity (vs. White)					53.65***	3.32	2.41	4.58	50.59***	3.35	2.40	4.68	51.21***	3.38	2.42	4.72
Townsend (vs. most affluent)					17.33***	–	–	–	15.79**	–	–	–	14.87**	–	–	–
Second quartile					2.95	1.27	0.97	1.67	2.90	1.27	0.97	1.67	2.80	1.26	0.96	1.66
Third quartile					11.28***	1.59	1.21	2.08	10.55**	1.57	1.19	2.05	9.87**	1.54	1.18	2.02
Most deprived					13.88***	1.72	1.29	2.29	12.60***	1.69	1.26	2.25	12.00***	1.67	1.25	2.23
Lifestyle factors
Smoking status (vs. nonsmoker)									4.74	–	–	–	3.89	–	–
Current smoker									4.28*	1.42	1.02	1.97	3.57	1.38	0.99	1.90
Previous smoker									1.54	1.14	.93	1.41	1.18	1.12	0.91	1.39
Alcohol drinker (vs. nondrinker)									0.03	0.97	0.66	1.41	0.03	0.97	0.66	1.42
BMI									1.71	0.98	0.96	1.01	2.01	0.98	0.96	1.01
Diabetes (vs. no)									0.14	0.92	0.60	1.41	0.11	0.93	0.61	1.43
Cardiovascular disease (vs. no)									0.01	1.02	0.73	1.42	0.03	1.03	.74	1.44
High cholesterol (vs. low)									0.50	1.10	0.85	1.42	0.44	1.09	0.84	1.41
Hypertension (vs. no)									0.69	0.91	0.74	1.13	0.69	0.91	0.74	1.13
Physically inactive (vs. active)									0.65	1.09	0.88	1.34	0.76	1.10	0.89	1.35
Ototoxic medication (vs. no)									2.53	1.19	0.96	1.46	2.30	1.18	0.95	1.45
Noise exposure
Work noise exposure (vs. low)													.5.58*	1.33	1.05	1.68
Music noise Exposure (vs. low)													0.53	1.17	0.83	1.50
Constant	1,394.72***	0.08	–	–	354.53***	2.50 × 10−4	–	–	166.13***	3.63 ×10−4	–	–	165.59***	3.40 ×10−4	–	–

Note. Total: n = 5,603; agricultural: n = 425, construction: n = 1,871, music: n = 114, finance: n = 3,193. BMI = body mass index; CI = confidence interval.

*p < .05. **p < .01. ***p < .001.

In the second iteration, the addition of demographic factors (age, gender, ethnicity, and SES) produced a significant increase in the overall model prediction compared with the first iteration, χ2(6) = 251.51, p < .001. The model explained 10.86% of the variance in hearing difficulties, and the overall correct classification for hearing difficulties was 91.25%. With these additional demographic factors, industry type remained a significant correlate of hearing difficulties (p < .001). The odds of having a hearing difficulty in the agricultural industry compared with finance increased (OR: 1.79, 95% CI [1.27, 2.52]), whereas the odds for the construction industry workers compared with finance slightly decreased (OR: 1.58, 95% CI [1.26, 1.97]). The odds of hearing difficulties in the music industry compared with finance reduced further, but there was no significant difference between these industries (OR: 0.57, 95% CI [0.23, 1.43]). Ethnicity was associated with hearing difficulties, with those reporting non-White ethnic background having a much higher probability of having hearing difficulties compared with those who reported White ethnic background (OR: 3.32, 95% CI [2.41, 4.58]), when accounting for all variables.

Health and lifestyle factors (BMI, smoking status, alcohol consumption, diabetes, cardiovascular disease, hypertension, cholesterol, physical activity, and ototoxic medication) were added in the third iteration of the model. The addition of these factors did not improve the overall model prediction, χ2(10) = 11.99, p = .29, and none of these individual lifestyle factors were significant correlates of hearing difficulties. Industry type remained a significant correlate of hearing difficulties (p < .001), and the addition of these factors caused a minor increase in the odds of having hearing difficulties in the agricultural industry (OR: 1.81, 95% CI [1.28, 2.55]), a minor decrease in the construction industry (OR: 1.56, 95% CI [1.25, 1.96]), and no change in the music industry compared with the finance industry.

In the final iteration of the regression model, the addition of occupational and music noise exposure produced a significant increase in the overall model prediction, χ2(2) = 6.55, p = .04. The model explained 11.56% of the variance in hearing difficulties, and the overall correct classification for hearing difficulties was 91.20%. This final model iteration caused a decrease in the odds of having hearing difficulties in the agricultural industry (OR: 1.68, 95% CI [1.19, 2.38]) and construction industry (OR: 1.42, 95% CI [1.11, 1.80]) compared with finance. There was also a further decrease in the odds of having hearing difficulties in the music industry compared with finance (OR: 0.51, 95% CI [0.20, 1.28]), but this difference between these industries was not significant. Industry type remained a significant correlate of hearing difficulties (p = .001), even when accounting for all of these additional factors. Occupational noise exposure was a significant correlate of hearing difficulties, with exposed individuals being more likely to have hearing difficulties than those who were not exposed (OR: 1.33, 95% CI [1.05, 1.68]). However, music noise exposure was not a significant correlate of hearing difficulties (OR: 1.17, 95% CI [0.83, 1.50]).

Tinnitus

In the first iteration of the logistic regression model (Table 3), the overall model prediction was significant, χ2(3) = 35.97, p < .001. The model explained 1.11% of the variance (Nagelkerke R) in tinnitus prevalence, and the overall correct classification for tinnitus was 83.53%. Industry type was a significant correlate of levels of reported tinnitus (p < .001). The odds of reported tinnitus were significantly greater in the music (OR: 1.99, 95% CI [1.29, 3.05]) and construction industries (OR: 1.53, 95% CI [1.32, 1.78]) compared with the finance industry. The agricultural industry showed slightly higher odds of reported tinnitus than finance, but this difference was not significant (OR: 1.18, 95% CI [0.90, 1.54]).

Table 3.

Logistic Regression Model for Tinnitus With Industry Type, Demographic, Health, and Lifestyles Factors, and Noise Exposure.

Factors	Step 1				Step 2				Step 3				Step 4
	χ2(3) = 35.97, p < .001; R2= .01				χ2(6) = 91.77, p < .001; R2= .04				χ2(10) =9.10, p = .52; R2= .04				χ2(2) = 56.17, p < .001; R2 =.06
	Wald χ2	OR	95% CI		Wald χ2	OR	95% CI		Wald χ2	OR	95% CI		Wald χ2	OR	95% CI
			Lower	Upper			Lower	Upper			Lower	Upper			Lower	Upper
Industry type
Industry (vs. finance)	39.54***	–	–	–	21.55***	–	–	–	20.54***	–	–	–	4.10	–	–	–
Agricultural	1.35	1.18	0.90	1.54	0.75	1.13	0.86	1.49	0.61	1.12	0.85	1.47	0.05	0.97	0.73	1.28
Construction	31.16***	1.53	1.32	1.78	15.75***	1.39	1.18	1.64	14.81***	1.38	1.17	1.63	2.15	1.14	.96	1.37
Music	9.75**	1.99	1.29	3.05	8.86**	1.94	1.26	3.01	8.60**	1.93	1.24	2.99	2.06	1.39	.89	2.19
Demographic factors
Age					71.12***	1.04	1.03	1.05	60.87***	1.04	1.03	1.05	76.48***	1.05	1.04	1.06
Gender (vs. female)					3.14	1.18	.98	1.41	4.18*	1.21	1.01	1.45	.78	1.09	.90	1.31
Ethnicity (vs. White)					1.13	0.83	0.59	1.17	0.92	.084	0.60	1.19	0.86	0.85	0.60	1.20
Townsend (vs. most affluent)					4.48	–	–	–	4.76	–	–	–	5.2	–	–	–
Second quartile					0.01	1.01	0.84	1.22	1.70 × 10−3	1.00	0.83	1.21	1.65 × 10−4	1.00	.83	1.21
Third quartile					1.30	1.12	0.92	1.35	1.18	1.11	0.92	1.34	0.70	1.09	0.90	1.32
Most deprived					1.26	0.88	0.71	1.10	1.53	0.87	0.70	1.08	2.39	.84	0.67	1.05
Lifestyle factors
Smoking status (vs. nonsmoker)									4.88	–	–	–	2.98	–	–	–
Current smoker									0.20	1.06	0.82	1.37	2.66 × 10−3	0.99	0.77	1.29
Previous smoker									4.87*	1.19	1.02	1.39	2.73	1.14	0.98	1.33
Alcohol drinker (vs. nondrinker)									1.09	0.85	0.62	1.15	1.22	0.84	0.62	1.14
BMI									1.13	0.99	0.97	1.01	1.75	0.99	0.97	1.01
Diabetes (vs. no)									0.18	0.93	0.65	1.38	0.12	0.94	0.65	1.35
Cardiovascular disease (vs. no)									0.12	0.95	0.72	1.26	0.11	0.95	0.72	1.26
High cholesterol (vs. low)									0.60	0.92	0.75	1.13	0.60	0.92	0.75	1.14
Hypertension (vs. no)									0.38	0.95	0.81	1.11	0.39	0.95	0.81	1.11
Physically inactive (vs. active)									5.27 × 10−3	1.01	0.86	1.18	0.07	1.02	0.87	1.20
Ototoxic medication (vs. no)									0.78	1.07	0.92	1.25	0.42	1.05	0.90	1.23
Noise exposure
Work noise exposure (vs. low)													27.07***	1.60	1.34	1.91
Music noise exposure (vs. low)													23.12***	1.62	1.33	1.98
Constant	1,287.83***	.16	–	–	219.26***	.02	–	–	80.89***	.02	–	–	90.01***	.02	–	–

Note. Total: n = 5,727; agricultural: n = 439, construction: n = 1,914, music: n = 118, finance, n = 3,256. BMI = body mass index; CI = confidence interval.

*p < .05. **p < .01. ***p < .001.

The addition of demographic factors in the second iteration of the model significantly improved the overall model prediction of tinnitus prevalence, χ2(6) = 91.77, p < .001, and accounted for 3.73% of the overall variance. The overall correct classification of tinnitus stayed at 83.53%. Industry type remained a significant correlate of levels of tinnitus after including demographic factors (p < .001). The music industry (OR: 1.94, 95% CI [1.26, 3.01]) and construction industry (OR: 1.39, 95% CI [1.18, 1.64]) were more likely to report tinnitus than the finance industry, but the odds were reduced with the addition of these demographic factors. There was no significant difference between agricultural and finance industries, but the odds of reported tinnitus were also slightly reduced in the agricultural industry (OR: 1.13, 95% CI [0.86, 1.49]). Age was a significant correlate of tinnitus, with a 4% increase in the probability of reported tinnitus with each year of age (OR: 1.04, 95% CI [1.03, 1.05]).

The addition of health and lifestyle factors in the third iteration did not improve the overall model prediction, χ2(10) = 9.10, p = .52. The correct classification of tinnitus was 8.45%, and accounted for 3.99% of the variance. None of the health and lifestyle factors were significant correlates of tinnitus (all p > .05). Industry type remained a significant correlate of tinnitus (p < .001). There was a negligible change in the odds of reported tinnitus in music, construction and agricultural industries compared with finance with the addition of these factors. Age remained a significant correlate of tinnitus, and gender was also a significant predictor with the addition of these health and lifestyle factors, with males having greater odds of reported tinnitus compared with females (OR: 1.21, 95% CI [1.01, 1.45]).

In the final iteration, noise exposure increased the overall model prediction, χ2(2) = 56.17, p < .001, and accounted for 5.60% of the variance in tinnitus. The overall correct classification of tinnitus was 83.53%. Industry type was no longer a significant correlate of tinnitus (p = .25). There was a reduction in the odds of reported tinnitus in music (OR: 1.39, 95% CI [0.89, 2.19]), construction (OR: 1.14, 95% CI [0.96, 1.37]), and agricultural (OR: 0.97, 95% CI [0.73, 1.28]) industries compared with finance, and there was no significant differences between high- and low-risk industries (all p > .05). Age remained a significant correlate of tinnitus, while both occupational (OR: 1.60, 95% CI [1.34, 1.91]) and music noise exposure (OR: 1.62, 95% CI [1.33, 1.98]) increased the probability of reported tinnitus.

Discussion

We aimed to determine the extent to which occupational noise exposure versus demographic, health, and lifestyle factors might account for higher levels of hearing difficulties and tinnitus in high noise exposure industries (agricultural, construction, and music) compared with an industry with low noise exposure (finance). Data for a large participant sample were obtained from the U.K. Biobank resource. Hearing difficulties were more common in the agricultural and construction industries compared with the finance industry, and tinnitus was more common in construction and music industries compared with finance. Occupational noise exposure, older age, low SES, and non-White ethnic background partially accounted for higher levels of hearing difficulties in the construction industry compared with the finance industry. Occupational noise exposure alone partially accounted for higher levels of hearing difficulties in the agricultural industry compared with finance. However, the predictors of hearing difficulties included in the model did not fully account for the increased likelihood of hearing difficulties in these high-risk industries. Higher levels of tinnitus in music and construction industries compared with finance were accounted for by occupational and music noise exposure, as well as older age. These findings are discussed in detail below.

Prior to accounting for demographic, health, lifestyle, and noise exposure factors, construction and agricultural industries were found to have higher odds of hearing difficulties than the finance industry; in accordance with previous research (Nelson et al., 2005; Sliwinska-Kowalska & Davis, 2012; Stucken & Hong, 2014; Tak et al., 2009). Farming is one of the highest ranked occupations at a risk of hearing loss due to the high levels of noise exposure (Milz et al., 2008), exposure to farming-related noise from an early age (Ehlers & Graydon, 2011; Renick, Crawford, & Wilkins, 2009), and relatively low uptake of hearing protection devices (Depczynski, Challinor, & Fragar, 2011; McCullagh, Banerjee, Cohen, & Yang, 2016; McCullagh, Lusk, & Ronis, 2002). Hearing protection use in the construction industry has become more ubiquitous in the past 30 to 40 years (NIOSH, 1998; Rabinowitz, Galusha, Dixon-Ernst, Clougherty, & Neitzel, 2013), which could explain why the levels of hearing difficulties were found to be slightly lower than the agricultural industry in the current study.

For the construction industry, the addition of demographic factors in the second step of the regression model lead to a slight reduction in the likelihood of hearing impairment compared with the first step. This suggests that demographic factors such as increasing age, non-White ethnic background, and lower SES may also contribute to higher levels of hearing impairment in construction compared with the finance industry. For the agricultural industry, the addition of demographic factors caused an increase in the likelihood of hearing difficulties, therefore demographic factors were less able to account for differences in hearing difficulties compared with the finance industry. This could be due to the agricultural industry having a higher proportion of people reporting White ethnic background than other professions (Table 1), with those reporting White ethnic background being less likely to have a hearing difficulty compared with non-White ethnic background in the current analysis. The finding of lower levels of hearing difficulties among those reporting majority White British ethnic background than in those reporting non-White ethnic background is in accordance with previous U.K. Biobank studies (e.g., Dawes, Fortnum, et al., 2014). In these previous studies, high levels of hearing problems were observed for particular non-White ethnic groups that tend to experience the highest levels of deprivation (Dawes, Fortnum, et al., 2014). These particular U.K. ethnic groups also tend to have the poorest health outcomes more generally (Department of Health, 2001). Previous research in the United States reported that non-White ethnicity was associated with reduced risk of hearing loss (Agrawal et al., 2008). The explanation offered for better hearing among U.S. non-White ethnic groups was related to protective effects of melanin in the cochlea (Barrenäs & Lindgren, 1991). It may be that in the United Kingdom, factors associated with deprivation in particular ethnic minority groups might outweigh any biological resilience to NIHL conveyed by high levels of cochlear melanin.

Construction and agricultural industry workers remained at higher risk of hearing difficulties compared with finance after accounting for noise exposure and all demographic, health, and lifestyle factors. Occupational noise exposure, increasing age, ethnicity, and low SES may therefore only partially account for the higher risk of hearing difficulties in the construction industry compared with finance, and occupational noise exposure alone partially accounts for the higher risk in the agricultural industry. This could mean that (a) there are other unknown or unmeasured factors that could explain differences in hearing difficulties between high- and low-risk industries (e.g., occupation-related exposure to toxic substances that impact hearing; Rybak, 1992) and/or (b) measures included in the current analysis were not fully capturing the impact on hearing (e.g., noise exposure; see section Limitations).

The music industry workers did not have a greater likelihood of having hearing difficulties compared with the finance industry. Instead, this study showed a trend for lower levels of hearing difficulties in the music industry compared with finance. These findings are contrary to predictions, given that NIHL in musicians is well-documented (Jansen et al., 2009; Phillips, Henrich, & Mace, 2010; Pouryaghoub, Mehrdad, & Pourhosein, 2017; Sataloff, 1991). Nevertheless, there is also evidence to suggest that musicians may be less susceptible to the effects of noise damage on hearing as nonmusicians (Parbery-Clark, Anderson, & Kraus, 2013). This could be due to musicians having better listening abilities compared with nonmusicians, which could counteract the effects of noise exposure. For example, musical training could lead to enhanced auditory perception, working memory, and attention, which may improve performance on speech-in-noise tasks (for recent review, see Coffey, Mogilever, & Zatorre, 2017). This could have been true for musicians undertaking the DIN test in the U.K. Biobank protocol.

Industry type was also found to be a significant correlate of tinnitus, with higher levels of tinnitus in music and construction industries compared with finance. The addition of demographic factors in the second step of the regression model caused a decrease in the likelihood of tinnitus in high-risk industries. In particular, older age may account for some of the differences in likelihood of tinnitus between high- and low-risk industries. Crucially, after accounting for occupational and music noise exposure, as well as all demographic, and health and lifestyle factors in the regression model, industry type was not a significant correlate of tinnitus, and there were no significant differences in the likelihood of tinnitus between high- and low-risk industries. This suggests that both occupational and music noise exposure may account for the differences in the greater odds of having tinnitus in music and construction industries compared with finance, with older age in music and construction industry workers also partially accounting for this finding.

It is unclear why the agricultural industry workers did not have greater odds of tinnitus compared with the finance industry, especially given that agricultural workers had a greater probability of having hearing difficulties. This finding could be due to slightly lower levels of occupational and music noise exposure among agricultural workers compared with music and construction industries. This emphasizes the contribution of factors other than noise exposure (i.e., demographic factors) in explaining the differences in hearing difficulties between industries.

Health and lifestyle factors including smoking, hypertension, and cardiovascular disease were highest in agricultural and construction industries, while high BMI and physical inactivity were greatest in the finance industry. One might expect that these differences in health and lifestyle factors between industries may have had opposing effects on hearing across industries. But these health and lifestyle factors were not significant correlates of hearing difficulties and tinnitus, and their inclusion did not significantly improve the overall model predictions, and so they may not be able to explain differences in hearing difficulties and tinnitus between industries.

Limitations

The response rate to the U.K. Biobank invitation was 5.5%, and the resulting participant sample was generally healthier and of higher SES than the general population; a “healthy volunteer” selection bias (Fry et al., 2017). Therefore, the subsample of U.K. Biobank participants included in this study may not be representative of the general population and may not provide representative hearing impairment and tinnitus rates. Selection bias (or collider bias) can also induce a distorted association between predictor and outcome variables (e.g., lifestyle and hearing impairment), and thus exposure–disease relationships may not be generalizable and should be treated with caution (Cole et al., 2010; Elwert & Winship, 2014).

In the U.K. Biobank protocol, noise exposure was measured using nonstandardized self-reports that do not provide accurate estimates of noise intensity, noise type, and length of exposure. For example, the “shout to be heard” criterion is more likely to correspond to ≥99 dB (see Lutman & Spencer, 1991), so there may be cases of occupational noise exposure that are not captured by this criterion (i.e., 85–98 dB), and only workers with the highest levels of noise exposure may have been identified. As such, it is difficult to precisely determine the effects of noise exposure on hearing difficulties across different industries. It is possible that the similar levels of hearing difficulties between music and finance industries is because the majority of musicians’ occupational noise exposure contains less harmful music-related activities (Grinn, Wiseman, Baker, & Le Prell, 2017; le Clercq, van Ingen, Ruytjens, & van der Schroeff, 2016; Valderrama et al., 2018; Yeend, Beach, Sharma, & Dillon, 2017). In contrast, construction and agricultural industries have more damaging impulse sounds (Bramhall, Konrad-Martin, McMillan, & Griest, 2017; Depczynski, Franklin, Challinor, Williams, & Fragar, 2005; Starck, Toppila, & Pyykkö, 2003; Suvorov et al., 2001), and fewer silent periods and less variation in sound levels (Rawool, 2011). Furthermore, music noise exposure may not fully account for the level of recreational noise exposure more generally. Although it is difficult to get a precise measure of lifetime noise exposure, there have been recent developments in self-report measures. The Noise Exposure Structured Interview (Guest et al., 2018) gathers comprehensive estimates of occupational and recreational noise exposure, including a broader estimate of noise intensity using a speech communication table ranging from “Normal voice” (<80 dB) to “Shout in listeners ear” (>110 dB), as well as estimating the length of exposure. Future prospective studies would benefit from a more inclusive and accurate measure of occupational and recreational noise exposure, with the latter not limited to music activities.

A further limitation of the U.K. Biobank protocol is that information relating to occupation title was only collected for individuals who were currently employed or self-employed, and previous job history was not recorded. Therefore, participants may have reported occupational noise exposure relating to a previous occupation in a high-risk industry (e.g., construction) but not related to their current occupation which may be in a low-risk industry (e.g., finance). It was not possible to test the extent to which noise exposure from previous occupations relates to current hearing status. Nevertheless, the noise exposure questions do enquire about the history of noise exposure in both occupational and music settings, and so if a participant had previous experience of a noisy occupation or exposure to loud music, then these data would be captured and included in the analysis.

There could also be other factors linked with hearing difficulties and tinnitus which could explain differences between industries, but were not available to be included in the current analysis, such as exposure to chemical contaminants (Rybak, 1992). These additional factors may account for more of the variance in hearing difficulties and tinnitus, where the factors included in the current analysis only explained a small proportion of the variance.

Implications

These findings emphasize the importance of limiting occupational and recreational noise exposure for high-risk occupations such as in agricultural, construction, and music industries. Hearing conservation programs and legislative changes may be effective in isolating or reducing the noise source, increasing the use of hearing protection, and reducing hearing difficulties in high-risk industries (NIOSH, 1998; Rabinowitz et al., 2013; Reddy, Welch, Ameratunga, & Thorne, 2017; Seixas et al., 2011), but there may still be some way to go in improving these hearing conservation programs (Barlow & Castilla-Sanchez, 2012; Daniell et al., 2006; El Dib, Mathew, & Martins, 2012; Lusk et al., 1999; Neitzel & Seixas, 2005; O‘Brien, Ackermann, & Driscoll, 2014; Zhao et al., 2012). Based on the current findings, one possibility to improve the efficiency of interventions to promote hearing health is to focus on high-risk demographic subgroups (e.g., lower SES, older individuals, and certain ethnic groups) within industries which have a higher risk of hearing difficulties or tinnitus.

Conclusions

The likelihood of having a hearing difficulty as assessed by a speech-in-noise test was greater for construction and agricultural industries compared with finance. A combination of occupational noise exposure, older age, low SES, and non-White ethnic background partially explain differences in the risk of hearing difficulties between industries. Tinnitus was more commonly reported in music and construction industries compared with finance, and these differences were explained by occupational and music noise exposure, and older age. These findings highlight the importance of legislation to regulate noise exposure limits for professions with high levels of occupational noise exposure. There is a need for interventions to promote healthy hearing in both occupational and recreational settings with a focus on particular demographic subgroups who might be more susceptible to hearing difficulties and tinnitus.

Click here for additional data file.

Supplemental material, TIA885571 Supplemetal Material for Hearing Difficulties and Tinnitus in Construction, Agricultural, Music, and Finance Industries: Contributions of Demographic, Health, and Lifestyle Factors by Samuel Couth, Naadia Mazlan, David R. Moore, Kevin J. Munro and Piers Dawes in Trends in Hearing