Multiple work-related accidents: tracing the role of hearing status and noise exposure
- 1Institut national de santé publique du Québec, Québec (QC), Canada
- 2Ecole d’orthophonie et d’audiologie, Université de Montréal, Montréal (QC), Canada
- 3MRC Hearing and Communication Group, School of Education, University of Manchester, Manchester, UK
- 4Département de médecine sociale et préventive, Université Laval, Québec (QC), Canada
- Serge André Girard, Institut national de santé publique du Québec (INSPQ), 945 Avenue Wolfe, 5e étage, Québec (QC), Canada G1V 5B3;
- Accepted 5 December 2008
- Published Online First 27 January 2009
Objectives: Our main purpose was to investigate any relationship between noise exposure levels in the workplace, degree of hearing loss (HL), and the relative risk of accident (OR of single or multiple events).
Methods: We conducted a retrospective study of 52 982 male workers aged 16–64 years with long-standing exposures to occupational noise over a 5-year period, using “hearing status” and “noise exposure” from the registry held by the Quebec National Institute of Public Health. Information on work-related accidents was obtained from the Quebec Workers’ Compensation Board. Hearing threshold level measurements and noise exposures were regressed on the numbers of accidents after adjusting for age.
Results: Exposure to extremely noisy environments (Leq8h (equivalent noise level for 8 h exposure) ⩾90 dBA) is associated with a higher relative risk of accident. The severity of hearing impairment (average bilateral hearing threshold levels at 3, 4 and 6 kHz) increases the relative risk of single and multiple events when threshold levels exceed 15 dB of hearing loss. The relative risk of multiple events (four or more) is approximately three times higher among severely hearing-impaired workers who are exposed to Leq8h ⩾90 dBA.
Conclusion: Single and multiple events are associated with high noise exposure and hearing status. This suggests that reducing noise exposure contributes to increased safety in noisy industries and prevents hearing loss. Hearing-impaired workers assigned to noisy workstations should be provided with assistive listening devices and efficient communication strategies should be implemented.
There is mounting evidence that industrial noise and hearing loss affect safety at work when individual exposures and hearing status are studied as independent variables or in combination with other factors such as task complexity.1–4 The magnitude of the problem and its consequences for workers are not yet fully understood.
Many reports acknowledge the role of one particular risk factor in work accidents. These studies provide an estimation of the increase in risk attributable to a given factor based on correlation analysis. Other studies have focused on the incidence rate in populations without taking into account the fact that some workers will have been involved in several accidents.
Up to now, the particular case of “noise exposure acting independently or interactively with hearing status to cause a work accident” has not been studied in association with accidents occurring to the same workers during a predefined study period. Gauchard et al have studied multiple events, including all accidents incurred by the same persons within a defined period, but did not specifically focus on noise exposure or hearing status.5
Some studies of multiple or recurring accidents, such as falls among the elderly, are based on the assumption that these events might be related. However, the situation may be different in the workplace where accidents occurring to individuals are more likely to be unrelated. The protective role of experience must be also considered6: older skilled and experienced individuals are likely to avoid risky situations despite aging-related impairments in vision, strength and balance.
In this study, we report the findings from a large retrospective cohort study of workers with permanent hearing loss and long exposure to high levels of occupational noise. The dependent variable is the number of work-related accidents occurring to these workers over 5 years following a standardised audiometric examination.
Population and databases
The initial cohort consisted of 88 320 workers whose hearing was tested by the Quebec National Institute of Public Health (Institut national de santé publique du Québec; INSPQ) as a part of a government program to prevent noise-induced hearing loss among workers exposed to daily Leq (equivalent continuous sound pressure level) ⩾80 dBA. Of these workers, 81 346 (92.1%) were men, mostly in blue-collar occupations (82.7%). They were drawn from the following industries (as defined by the North American Industry Classification System): 14 926 workers from the “metal products manufacturing” sector (10.9%), 8744 from “plastic and rubber products manufacturing” (9.9%), 8214 from “mining extraction” (9.5%), 6536 from “wood products manufacturing” (7.4%), 6006 from “sawmills” (6.8%) and 5917 from “transportation equipment manufacturing” (6.7%).7
All these workers had their hearing and noise exposure levels individually assessed by the INSPQ which conducted on-site individual audiometric tests using a mobile laboratory equipped with soundproof rooms. Workers had no exposure to noise for 14 consecutive hours prior to the test. The quality standards of all hearing examinations complied with ISO 6189, including monitoring of the reproducibility of audiometric measurements.8–10
Individual records of workers who had hearing tests conducted by INSPQ between January 1983 and December 1996 were searched for correspondence regarding work-related accidents reported to the Quebec Workers’ Compensation Board (Commission de la santé et de la sécurité du travail; CSST), which is the state authority commissioned to gather this information. More specifically, individual accident histories reported to the CSST were searched forward for a period of 5 years starting at the date of the most recent audiometric test (the “reference” audiogram).
Because these data were drawn from two independent databases based on the same initial population of noise-exposed workers, the construction of the cohort was entirely retrospective, having more in common with occupational cohort and case–control studies than with prospective cohort studies.11 Cross-tabulation of databases was carried out for the entire period beginning from the reference audiogram, with the two following exceptions: (i) participation of a worker ceased when he reached the age of 65; (ii) when the monitoring period was shorter than 5 years, 30 April 1998 was defined as the end point of the observation period. Once the INSPQ and the CSST databases were merged, the level of participation was 100%. There is a risk of under-reporting of accidents, but it may be possible to partly compensate for this shortcoming by examining large data sets. The likely result is an under-estimation of the real numbers of accidents leading to a conservative bias of estimates and correlates.
Selection of participants
Hearing status was assessed by a standardised analysis of auditory thresholds. Workers identified professionally or through personal case history as having a “normal” hearing profile or a hearing impairment compatible with significant noise exposure were included in the study. These workers were defined after Picard et al as being “otologically normal except for a history of occupational noise exposure” if they satisfied the criterion of having a normal low-frequency tympanogram (226 Hz) at the time of their hearing test.12
Based on these criteria, 30 588 workers were excluded from the study because of hearing deficits attributable to other causes (eg, chronic middle-ear disease including surgery and/or a hearing loss at 0.5, 1 or 2 kHz exceeding the 90th percentile for age, as per ISO 7029).13 Only male participants were studied as just 4750 noise-exposed women were included in the INSPQ database. This resulted in a final cohort of 52 982 male participants aged 16–64 years.
Independent variables included noise and hearing assessments, both recorded in the INSPQ database. The number of accidents per worker was derived from the prevalence of accidents reported to the CSST (the dependent variable) assuming that each claim referred to a new and distinct event regardless of its nature, the severity of its consequences or the circumstances. This was an essential requirement for events to be considered independent of one another.
The age of workers at the time of the hearing examination was used as an adjustment variable, given that presbyacusis (ISO 7029) can lessen the impact of work experience.13 Presbyacusis and experience predictably act in opposite directions: more hearing loss may result from aging, thus increasing the risk of accidents; however, as age increases, most individuals have greater experience and this improves work safety.14 Failing to disentangle these two effects may falsely suggest that slowly progressing hearing impairment, such as presbyacusis and noise-induced hearing loss, results in a lesser risk of accidents as the positive effect of experience outweighs the negative effect of lower sensory levels, or else causes a possible null effect. This serious confounding factor must be considered in the analysis. The “age” variable was stratified as follows: 16–24, 25–34, 35–44, 45–54 and 55–64 years.
Occupational ambient noise levels were measured at the workstation occupied by every individual in the months prior to the reference audiometric test. These exposure levels were grouped into two categories: Leq8h (equivalent noise level for 8 h exposure) <90 dBA and ⩾90 dBA. Because of point measurement, exposure levels were not expected to have the same level of precision as a cumulative index of exposure in a career. Participants from industrial sectors with high staff turnover were not retained in the study.
As with the hearing assessment data, audiometric data were analysed quantitatively and qualitatively.12 Given the similarity of findings between the scaling methods, only the analysis completed on categorical data is reported here. The hearing categories based on the degree of impairment proposed by Yantis are the same as those of Picard et al.12 15 16
For the accident count, up to three events were specified in discrete units. Since less than 1.5% of participants (n = 790) experienced more than four accidents during the observation period, workers with four or more events were grouped into a fourth residual category.
Estimation of the risk of being involved in one or more accidents was computed as a function of levels of occupational noise exposure and degrees of hearing loss adjusted for age. The dependent variable (number of accidents) is displayed on an ordinal scale (0, 1, 2, 3, 4 or more accidents). Since the proportional odds assumption for the cumulative logit model could not be satisfied, a polytomous regression (PLR) model was used. In a first PLR analysis, all data requiring adjustment were integrated into a single model in which the adjustments were ordered as a function of excess accident counts by comparison with accidents occurring to workers with normal hearing, who were used as baseline controls. A second PLR analysis was carried out to study the combined effects of hearing loss and increases in levels of noise exposure. For this analysis, baseline controls were individuals with normal hearing who were also exposed to Leq8h <90 dBA.
Sample principal characteristics
Table 1 shows the characteristics of the sample regarding history of occupational noise exposure (daily levels and duration in career) and degree of bilateral high-frequency hearing loss. Permanent hearing losses appear as a consequence of extended exposure. This is indicated by an almost threefold increase in the prevalence of high-frequency hearing loss when noise exposure has lasted for 8–16 years and, similarly, when it has lasted for 16–25 years. However, part of this effect may be attributable to more intense exposure levels given that those exposed to the highest levels may also be those with more severe hearing loss (see in table 1 the slightly different audiometric profile of workers exposed to levels above or below 90 dBA).
The majority (31 570; 59.6%) of the 52 982 participants worked in ambient noise levels ⩾90 dBA at the time of the hearing tests. The distribution of these highly-exposed workers by auditory status (table 1) is as follows: 58.65% of the 31 347 workers with normal hearing (n = 18 386) were exposed to levels ⩾90 dBA (Leq8h), as were 58.82% of the 10 585 workers with very slight hearing loss (n = 6226), 63.18% of the 3053 workers with moderate hearing loss (n = 1929) and 65.37% of the 3809 workers with severe hearing loss (n = 2598). A χ2 Cochran-Mantel-Haenzel test shows that the proportion of workers with mild-to-severe hearing loss is higher among those exposed to the higher noise level of ⩾90 dBA (p<0.001). This suggests that high-frequency hearing impairment increases with noise exposure levels as expected for the acquisition of noise-induced hearing loss.
The distribution of participants across age groups is displayed in table 2. Overall, 59.17% of participants (n = 31 347) had normal hearing, although 59.59% had been exposed to Leq8h ⩾90 dBA, the remainder being exposed to 80–89 dBA. Participants were relatively young at 36.1 years of age, with a mean duration of exposure of 13.1 years. Table 2 also shows that a proportionally greater number of younger workers have normal hearing compared to their older peers who have been exposed to noise over longer periods of time (table 1).
Table 3 shows the distribution of workers (numbers and proportions) by hearing status and number of accidents. Of particular significance is the reduction in worker numbers as accident counts per worker increase. For instance, 22 566 of the 52 982 noise-exposed participants (42.59%) experienced at least one accident, totalling 43 250 events. By contrast, the upper end of the distribution shows that a mere 87 individuals (0.38%) experienced 10 or more accidents, the maximum value being 25 for one person, for a total of 1017 separate accidents, representing 2.35% of all events. Within this context, the particular contribution of hearing loss may be lost due to the lack of controlling for the protective effect of age. Table 3 further indicates that among the 31 347 workers reported in table 2 as having normal hearing, 44.38% (n = 13 911) had at least one accident during the 5-year follow-up period: 23.47% of workers with normal hearing had one accident (n = 7358), while 5.33% (n = 1672) experienced four or more.
Table 4 shows that the relative risk of accident (odds ratio, OR) increases with deteriorating hearing for all accident counts. When only one accident is reported, ORs increase from 1.12 when hearing loss is very slight to 1.40 when it is severe. When two accidents are reported during the 5-year follow-up period, the relative risk increases from 1.25 to 1.69 for hearing losses progressing from very slight to severe, and this pattern applies also to the categories of three and four or more accidents. Of particular interest is the large increase in the OR of having four or more accidents when hearing loss deteriorates from very slight to severe (OR values increasing from 1.33 to 2.30). The PLR analysis using normal hearers as controls further indicates that there is no significant interaction between hearing loss and noise exposure level, thus confirming the particular contribution of hearing loss to the risk of consecutive accidents.
Table 4 also reports the ORs of being involved in one, two, three, four or more accidents when Leq8h ⩾90 dBA. Working in very noisy environments increases the ORs of all accident counts from 1.10 for a single event to 1.30 for four or more.
Table 5 indicates that the increase in accident risk is associated with higher daily levels of noise and degrees of hearing loss when baseline controls are restricted to minimally noise-exposed workers with normal hearing. This is true for all accident counts except in a few cases where exposure is <90 dBA. Of particular significance is the almost threefold increase in the relative risk of having four or more accidents (OR 2.79) when workers who are exposed to noise levels ⩾90 dBA also have severely impaired hearing. This particular analysis suggests a synergistic interaction between these two independent variables.
Finally, the ORs displayed in table 5 suggest that the effect of noise exposure is equal to or somewhat larger than that of hearing loss. For instance, some moderately hearing-impaired workers have had only one accident despite exposure to Leq8h⩾90 dBA. The OR of this group is accrued to 1.53 when compared to baseline controls. This is a larger risk than that experienced by individuals with the same hearing loss who are exposed to <90 dBA (OR = 1.29) and a somewhat greater risk than that of severely hearing-impaired workers who are exposed to Leq8h<90 dBA (OR = 1.34).
These results confirm the association between exposure to occupational noise, noise-induced hearing loss, and the risk of single or multiple accidents at work. Our computation of ORs for multiple events adds significantly to previous reports by Girard et al and Picard et al1 12 17 and augments the list of independent variables important for occupational safety. These findings show that noise exposure and hearing impairment affect work safety. Furthermore, the size of the databases used in this investigation and the duration of the observation period provide about 240 000 person-years of experience, strengthening the hypothesis that the investigated independent variables are associated with industrial accidents.
This study concludes that noise exposure in the workplace adversely effects hearing and compromises safety by significantly increasing the risk of accident.
This study significantly contributes to current knowledge on safety and health by linking accident prevention to the control of noise in industry.
Efforts to diminish noise levels in the workplace would contribute to better safety performance.
ORs of single or recurrent events of 1.10–1.30 in high noise levels (table 4) may appear modest. However, combinations of the independent variables that are especially hazardous, such workers with severe hearing loss being exposed to noise levels ⩾90 dBA (table 5), should be taken into account. Assigning these workers to extremely noisy workstations is a serious safety hazard. Current indications that hearing loss and noise exposure ⩾90 dBA contribute equally to the relative risk of multiple accidents, are also a reminder of the pervasive influence of noise exposure. Despite the power and consistency of our findings, the fact that they are population-based means they cannot be used to predict the individual risk of accident.
Noise exposure and hearing loss as underlying mechanisms of accident occurrence
The fact that daily exposures to noise ⩾90 dBA increase the risk of single and multiple accidents even in the presence of normal hearing indicates the importance of noise levels in the workplace; the masking effect of occupational noise could explain its negative effect on safety. Our study includes manual workers from industrial sectors known to contain hazardous environments. However, this particular subset of workers with normal hearing held jobs that were not significantly different in terms of stress, strain and degree of complexity to those held by other workers, nor are there any indications that these individuals worked unusual shifts, schedules or much in excess of our reference of 40 h/week. Therefore, we do not consider the potentially confounding effect of these critical variables to be greater than statistical random error. We also consider that lack of information about some structural components of safety problems, such as workload, unusual stress or work schedules and workplace context, also applied to the workers in the baseline control comparison group.
One of the study limitations relates to the assumption of independence between reported accidents, although further insights into accident circumstances may challenge this. Investigations capable of modelling the length of time between the different accidents in relation to noise exposure level and hearing status may lead to different conclusions.
The definition of noise exposure levels deserves further attention. In this study, the current estimate is a point measurement that may differ with the total dose throughout a career. Our method may be imprecise insofar as significant numbers of participants may have been exposed to varying noise levels because of different jobs. As it is not possible to ascertain how many of the 52 982 workers had changed jobs, there is no way to assess the effect of this confounding variable or determine whether it may have compromised the statistical validity of the analysis.
The available data did not consider the use of personal hearing protective devices. This could lead to some over-estimation of noise exposure especially at low frequencies (<1 kHz). However, the contribution of these devices to improving safety in the workplace is unclear in light their effects on hearing (decreased detection and salience, effects on localisation and recognition of noise) and vocal production (reduction of vocal effort due to the Lombard effect). The systematic use of ear protectors in noisy workplaces is not mandatory in Quebec so assuming that many workers wore hearing protection could lead to incorrect estimates of noise exposure levels.
Moreover, this study only includes work-related accidents reported to the CSST. Failure to report accidents is a real possibility as mentioned by Rosenman et al in their study of declarations to the Bureau of Labor Statistics (BLS) in the state of Michigan.18 The two situations differ, however, as the CSST has full authority to inspect industrial plants and investigate work-related accidents, while the BLS is limited to record keeping. However, there may be some under-reporting of minor accidents in view of the financial consequences to workers and employers.
Finally, it must be emphasised that the cohort used in this investigation is not representative of the total population of workers occupationally exposed to noise in Quebec. In particular, the database maintained by the INSPQ does not contain large numbers of workers from the particularly noisy and hazardous mining, logging and construction industries. Again, this exclusion introduces a bias promoting under-estimation in our figures. As a whole, the sample investigated represents about 13% of all Quebec workers exposed to potentially hazardous noise levels.
These results suggest that occupational noise exposure adversely effects work safety, significantly increasing the risk of single and multiple accidents, in addition to the known effects of noise exposure on hearing.19 20 In view of the increased relative risk of single and multiple accidents associated with occupational noise exposure and the degree of hearing loss, these findings call for an increased and sustained commitment to the prevention of accidents by reducing noise levels in industry. Also, a new set of variables should be introduced in the analysis of work-related accidents where noise is a factor, especially where some hearing loss is also present.
Competing interests: None.
Funding: Funding was provided by the Institut national de santé publique du Québec (INSPQ), Québec, Canada.