The present study introduces a new occupational noise exposure assessment tool using the O*NET database, evaluating its applicability to an examination of noise-related adverse health effects in the general population using hearing loss, a well-established noise-induced health outcome.
Our findings suggest that the use of O*NET scores may provide enough variation in the proxy measure of occupational noise exposure so that it can be applied for the general population with a wide range of occupation groups. It should be noted that this study did not attempt to validate the O*NET scores as a surrogate for personal occupational noise exposure levels. Rather, we evaluated an applicability of the O*NET scores as a proxy measure in association with occupational noise-related health effects in the general population, given available job title information. We found a significant dose–response relationship of O*NET noise scores with hearing loss and noise notch in NHANES, confirming that O*NET scores would be useful for examining noise-related health effects in the absence of personal occupational noise exposure data. Our results also extend evidence of noise-induced hearing loss in workers with extremely high noise exposure to the general population with low noise exposure, reinforcing occupational noise as an important risk factor for hearing loss. We found that men, non-Hispanic white and the age group of 40–59 years were more susceptible to occupational noise-associated hearing loss than other groups.
In fact, we ran regression analyses dealing with O*NET scores as a continuous variable and estimated the β coefficients corresponding to a one-unit increase in O*NET scores. The OR for risk of hearing loss corresponding to a one-unit increase in O*NET scores (range between 1 and 5) was 1.65 (95% CI 1.28 to 2.13) in a multivariable-adjusted model. A significant dose-dependent relationship with O*NET scores was retained in sequential models after adjusting for socioeconomic factors, non-occupational noise exposures and other potential risk factors. This suggests that the association between occupational noise exposure and hearing loss is independent of such risk factors. This increased risk is roughly equivalent to 20 or more pack-years of smoking (OR=1.54), diabetes (OR=1.66) and recreational noise exposure (OR=1.62) (see supplemental table 8). The estimated effect size of O*NET score is also similar to the effects of 5 years of ageing (OR=1.69) when age is fit linearly.
It is difficult to compare our findings to other studies because there are no studies of dose–response relationship between occupational noise exposure and hearing loss in the general population with low to high exposure as a continuous variable. A few previous investigations of noise and hearing loss have been made across crude occupational groups in the general population. In one such study, over 3500 older adults in Beaver Dam, Wisconsin, were examined for hearing loss in six occupation categories. A statistically significant increased risk of hearing loss was found in service (OR=1.85, 95% CI 1.40 to 2.43), operations/fabricators (OR=1.99, 95% CI 1.53 to 2.59) and production (OR=3.48, 95% CI 2.53 to 4.79) compared to management as a reference group.6
Another study mailed a questionnaire over 22 000 adults of working age across Britain and examined the association between years worked in a noisy job and self-reported hearing difficulty. That study found an increase in hearing difficulty by years worked in a noisy job and a statistically significant increased risk of hearing loss in 5–10 years (prevalence ratio=3.0, 95% CI 1.5 to 6.1) and over 10 years (prevalence ratio =3.8, 95% CI, 2.4 to 6.2) compared to non-exposed group.16
Our findings are broadly compatible with these studies and confirm the evidence of increased risk of hearing loss with increase in occupational noise exposure in the general population.
In addition to its relationship with hearing loss, the O*NET occupational noise score had a strong dose-dependent relationship with noise notch. The presence of noise notch is one diagnostic in determining that hearing loss is noise induced rather than the effect by other factors such as ageing.11
The ORs of the risk of noise notch increased also gradually across the O*NET noise score quintiles (ORs 1 (Reference), 0.96, 0.79, 1.35 and 1.51). This suggests that the O*NET occupational noise could be a good proxy for occupational noise exposure.
The main strengths of this study include (1) the use of representative samples of the US general population, including over sampled minority populations, which enables the observed results to be generalisable; (2) the adjustment for various potential confounding factors of the association between occupational noise and hearing loss, especially noise exposure other than workplace noise, such as firearm and recreational noise, and use of ototoxic medication and (3) the use of NHANES data conducted with strict quality control procedures.
This study has several limitations that should be considered. Because the O*NET database we used is based on the frequency of exposure to sounds and noise levels considered distracting and uncomfortable rather than on actual noise measurements, exposure misclassification may exist. Moreover, the O*NET data are classified only by occupation groups and do not account for variations in noise exposure from different industry groups or different job task groups within the same occupation classification. The assumption that jobs with the same title have similar occupational noise exposure could also lead to misclassification of exposure. Misclassification might also have occurred when 801 O*NET occupation groups were combined into 40 NHANES occupation groups. Because the O*NET survey is totally independent of the audiometry tests in NHANES, however, such exposure misclassification is likely to be non-differential and leads to a true association towards the null.
Although our study showed that as an exposure proxy, longest job is better than current job in predicting occupational noise-induced hearing loss, we could not account for the job history nor the duration of each job. Because the reported longest job is more likely to be related to hearing loss, however, the bias would be non-differential. Collecting information on full job history and duration would improve validity and reliability of any noise exposure assessment using O*NET.
Although we examined three cycles of the NHANES data, which offers significant power, causal inferences may not be made because of the cross-sectional nature of the NHANES data. Nevertheless, use of the longest job may be temporally relevant to current audiometry test results.
One might argue that there is selection bias in that the association between occupational noise and hearing loss is different for subjects included in our analysis who provided information on their longest job and those excluded due to no longest job information. We found that the prevalence of noise notch for included subjects was significantly different from the prevalence for excluded subjects and that included subjects were more likely than excluded subjects to have been exposed to loud job noise for at least 3 months on all previous jobs (supplemental table 2). Most of the excluded subjects had never worked (75%), are currently housewives (67%, all women), disabled people with no job history (10%) and students (8%). Although our results cannot be generalised to the non-included people (housewives, students and the disabled), we believe that the observed associations are valid to conclude noise exposure at workplaces as an important risk factor for hearing loss and that the selection bias is unlikely.
In summary, the present study supports the hypothesis that occupational noise exposure increases the risk of hearing loss across various occupations. Utilisation of the O*NET noise exposure data would allow us to perform epidemiological studies of occupational noise exposure in the general population and to better understand the health effects of occupational noise exposure.