Annals of Occupational Hygiene Advance Access originally published online on November 4, 2008
Annals of Occupational Hygiene 2009 53(1):41-54; doi:10.1093/annhyg/men071
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Comparison of Perceived and Quantitative Measures of Occupational Noise Exposure
1 Department of Environmental and Occupational Health Sciences, University of Washington, Box 354695, Seattle, WA 98195, USA
2 Department of Biostatistics, University of Washington, Box 357232, Seattle, WA 98195, USA
3 School of Environmental Health, University of British Columbia,Vancouver, British Columbia, Canada, V6T1Z3
* Author to whom correspondence should be addressed. Tel: 1-206-221-5445; fax: 1-206-616-6240; e-mail: rneitzel{at}u.washington.edu
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Objectives: Characterization of highly variable noise exposures over long periods of time presents a major challenge. Common exposure assessment strategies such as assignment of exposure levels based on job title information may not provide adequate exposure contrast or precision for variable exposures. Subjective exposure data may offer an alternative or complementary exposure assessment strategy. This study evaluated the relationship between perceived and quantitatively measured exposure.
Methods: Twenty subjects were recruited at each of three worksites with different noise environments (continuous, intermittent and highly variable). Full-shift quantitative measurements (n = 206) were made on each subject during four workshifts over 2 weeks. Perceived exposure data were collected via surveys on subjects’ first (n = 58) and last (n = 57) monitored shifts, as well as through timeline logs completed by subjects during each monitored shift. The first survey focused on the first shift only, while the second survey covered the whole 2-week period.
Results: Timeline log data suggested that subjects could differentiate between different noise levels and degrees of noise variability. Survey items on perceived exposure variability and impulsiveness performed well at the continuous and highly variable sites. Analyses of contrast between exposure grouping strategies showed that job title generally did not produce statistically distinct exposure groups and that several survey items provided greater contrast than job title. The precision of exposures predicted from survey items was comparable to, or slightly better than, that of job title for several survey items, and the addition of survey items to prediction models which included job title improved model fit and precision.
Conclusions: Supplemental perceived noise exposure information appears to offer promise for improving exposure estimates, particularly for individuals with highly variable exposures.
Keywords: exposure assessment • hearing loss • noise • quantitative measurement • subjective rating
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Noise is one of the most common occupational exposures in the US and worldwide (NIOSH, 1998). Although regulations (OSHA, 1983) and recommended standards (ACGIH, 2006) have been in place for decades, the prevalence of overexposures and noise-induced hearing loss (NIHL) remains high (Nelson et al., 2005; Daniell et al., 2006). NIHL has a profound social and occupational impact on affected individuals and substantially reduces quality of life (Sataloff and Sataloff, 1987). The risk of NIHL associated with chronic exposure to continuous, stable noise has been well characterized, and models are available that predict hearing impairment expected to result from this type of exposure (ANSI, 1996). However, NIHL risk for workers with exposure levels that vary widely both within and between shifts is less well understood; for example, several studies of construction workers with highly variable noise exposures measured less NIHL than would be expected from the available prediction models (Hessel, 2000; Seixas et al., 2005b). Given that there is evidence to suggest that the prediction models should be appropriate for use in forecasting NIHL from variable and intermittent noise (Passchier-Vermeer, 1973; ANSI, 1996; NIOSH, 1998), the uncertainty surrounding the risk of NIHL from highly variable noise exposures in construction is due in part to challenges in accurately assessing such exposures.
A number of different metrics are available for summarizing variable noise exposures. Most scientific and regulatory agencies rely on the equivalent continuous noise level (Leq) to assess NIHL risk; this metric is generally regarded as the most protective measure of continuous, intermittent and impact noise (NIOSH, 1998). However, US regulations specify a different metric, the Lavg, which incorporates a less protective time–intensity exchange rate (5 dB, as opposed to 3 dB for the Leq). Measurements made with these two metrics can result in divergent conclusions regarding NIHL risk, and the greater the exposure variability, intermittency or impulsiveness, the greater the divergence (Petrick et al., 1996). Metrics have been developed to assess exposure intermittency by examining the ratio of simultaneously measured Leq and Lavg levels, as well as to assess exposure ‘peakiness’ (i.e. the presence of impact or impulse-type noise) as the ratio of the maximum level (Lmax) to the Leq level over the same period (Seixas et al., 2005a).
Although exposures may be estimated for variable noise using one or more of the metrics described above, the degree to which estimates made with these metrics represents risk of NIHL is unclear. Accurate exposure assignment for epidemiological studies is difficult in industries like construction, which feature large temporal and spatial exposure variability. In these, repeated measurements on individuals, the approach often considered the ‘gold standard’ in epidemiological studies (Checkoway et al., 2004), may not be feasible. For some types of studies, such as retrospective cohort studies, exposure measurements may be impossible to obtain. Measurement error or misclassification in the exposure assessment process—regardless of the specific metric used—leads to bias in the estimated dose–response relationship (White et al., 2008). Strategies for creating homogenous exposure groups (Rappaport et al., 1993) often focus on job title. However, in industries like construction, where individuals in the same job title may perform a variety of activities, large within-job variability may mean that job title is not an optimal grouping strategy. For exposure assessment purposes, minimal within-group and maximal between-group exposure variability is desirable (Kromhout and Heederik, 1995). Within- and between-group variability may be summarized using the concept of contrast, calculated as the ratio of between-group variance to the sum of between- and within-group variance (Kromhout and Heederik, 1995). Contrast values near one indicate groups with differing mean exposure, while values near zero indicate groups with indistinguishable means.
To address questions surrounding the risk of NIHL from highly variable noise, noise exposures and hearing ability have been evaluated in a prospective cohort of construction workers in Seattle, WA. The initial cohort was evaluated from 1999–2004 (Seixas et al., 2004; Seixas et al., 2005b). Cohort members were then approached to participate in the second stage of the study (2005–2010). Subjects received annual hearing assessments during both stages. However, exposure measurements on cohort members have not been possible due to logistical challenges, which have prevented the estimation of Leq exposure levels for individual subjects from measurements made on those subjects. As an alternative to direct measurements, job title- and task-based (e.g. framing, demolition, grinding, etc.) noise exposure levels have been developed from a convenience sample of Seattle construction workers collected since 1999 (Neitzel et al., 1999; Seixas et al., 2001; Neitzel and Seixas, 2005). Job title-based exposure levels offer potentially unbiased estimates of group-mean exposures since the errors associated with each individual's measured exposure average out to approximately zero (Seixas and Sheppard, 1996), but as the variability between individuals within a group grows large, as is the case in construction, the precision of group-level estimates worsens since individual exposures vary around their group means (Seixas and Sheppard, 1996). Additionally, average exposure levels are similar among construction job titles (Neitzel et al., 1999; Neitzel and Seixas, 2005), resulting in minimal contrast between groups. Task-based estimates based on individual time-at-task information and measured task-specific noise levels can offer increased precision compared to job title-based estimates, but may include substantial bias for logarithmic exposure metrics such as those used for noise if within-task variability is not properly considered (Seixas et al., 2003). While task-based evaluations provide detail about individual exposure variability not explained by job title, neither approach offers information about additional sources of residual variability, e.g. site- or individual-specific factors. These limitations in job title- and task-based evaluations suggest a need for another source of individual-level exposure information.
Qualitative exposure assessment represents an additional pathway to obtain noise exposure information for members of the Seattle cohort. Although qualitative approaches are generally considered less accurate than quantitative methods (Teschke et al., 2002), for well-perceived exposures such as noise and highly variable exposures such as those found in construction, this method might play a useful role in exposure assessment. Perceived exposure intensity and duration have been assessed via self-report in a number of epidemiological studies, including several which have focused on noise (Ising et al., 1997; Palmer et al., 2002; Ahmed et al., 2004; Koushki et al., 2004; Jokitulppo et al., 2006). Whenever self-reported information is used, recall bias is a consideration; however, self-reported data on exposure durations and work environments (Palmer et al., 2000; Reeb-Whitaker et al., 2004) and on specific work activities (Reeb-Whitaker et al., 2004) may be accurate enough for use in exposure assessment. Previous studies have demonstrated that workers’ subjective perceptions of the intensity of their occupational noise exposures correlated well with brief sound level measurements (Ising et al., 1997) and that survey items relating to perceived exposure intensity exposure could be used to identify workers with full-shift levels over 85 dBA (Ahmed et al., 2004). However, no studies appear to have evaluated the potential use of subjective perceptions of noise intensity to predict average occupational exposure levels.
The goal of the current study was to evaluate the relationship between perceived and measured noise exposures assessed over a short period among workers in three different noise environments and to determine the potential utility of the perceived measures to predict exposure levels. Perceptions of exposure intensity, duration, variability and peakiness were evaluated. Factors relating to use of hearing protection were evaluated on the same population and will be reported elsewhere. The first hypothesis tested in this study was that subjective survey items would create groups with greater exposure contrast than job title. The second hypothesis was that exposures predicted from survey items would have better precision than that associated with job title.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Sites and subject recruitment
Three facilities participated in this study. The first site was a warehouse with continuous, stable noise levels (the ‘continuous’ site). The second site was a construction site with highly variable noise (the ‘highly variable’ site). The third was a sheet metal shop with intermittent noise (the ‘intermittent’ site). Twenty subjects from each site were recruited to participate in the study for a period of 2 weeks each. The exclusion criteria used during recruitment were non-literacy in English and non-availability over the entire study period. Potential subjects were given an overview of study purposes and procedures; interested individuals signed a consent form and were enrolled in the study. Subjects received a $20 incentive each day they participated. Data collection occurred between November 2006 and March 2007. All procedures were approved by the University of Washington Institutional Review Board, and all subjects were informed of their exposure levels at the conclusion of the study.
Noise measurements
Subjects wore noise dosimeters (model Q-300 or NoisePro DLX, Quest Technologies, Oconomowoc, WI, USA) on four randomly selected workdays over 2 weeks. Researchers fit subjects with dosimeters at the start of their shift and removed the units at the end of the shift. All dosimeters were calibrated before and after each shift. The dosimeters measured noise according to both the Leq criteria recommended by the US National Institute for Occupational Safety and Health (NIOSH, 1998) and the regulatory Lavg criteria of the US Occupational Safety and Health Administration (OSHA, 1983). The dosimeters logged Leq, Lavg and Lmax levels for each 1-min interval of the monitored shifts. During all four monitored shifts, subjects completed a 15-min resolution timeline log describing the intensity of their perceived exposure (‘timeline log perceived intensity’ item) on a five-point scale and exposure variability (‘timeline log perceived variability’ item) on a four-point scale. The text of the timeline log items is shown in Appendix 1.
Survey measurements
At the end of the first and fourth monitored shifts, subjects completed a brief (approximately 10 minute) self-administered survey written in English at the sixth-grade level (Flesch-Kincaid Grade Level, MS Word, Microsoft, Redmond, WA, USA). Both surveys contained 30 items related to perceived noise exposure, use of hearing protection, noise sensitivity and annoyance, perceived hearing ability and noise predictability. Additional demographic items were asked on the first survey only. The first survey focused on the first measured shift, while the second survey, delivered 9–11 days later, covered the entire 2-week study period. Both surveys included six perceived exposure intensity items, one exposure variability item and one item related to exposure peakiness. Survey 2 contained two additional items relating to noise-related changes in hearing. The full text of each noise-related survey item is shown in Appendix 1. All but one of the noise-related items used a five-point ordinal response scale. Items were framed in reference to absolute loudness (items ‘usual noise’, ‘noise frequency’ and ‘percent time in noise’), relative loudness (‘relative noise’), exposure-related behaviors (‘raise voice’ and ‘leave work’), peakiness (‘noise bursts’) and variability (‘noise steady’). The usual noise item was adapted from Ising et al. (1997), and the noise frequency item was adapted from Ahmed et al. (2004). The raise voice item was similar to one used by Ahmed et al. and was based on speech intelligibility data suggesting that if a speaker must raise their voice to be heard at arm's length, the noise level is likely >90 dBA (Robinson and Casali, 2000). The leave work item was based on the same data, which indicate that communication is impossible even at close range >100 dBA (Robinson and Casali, 2000). The percent time in noise item asked subjects to rate the percent of time spent in each of five response categories. Finally, the relative noise item was designed to complement the timeline log perceived intensity item.
Data analysis
Survey and dosimetry timeline log data were entered into an MS Access database. Dosimetry data were evaluated, and frank instrument measurement errors resulting in untenable data corrected, according to the following protocol (Seixas et al., 2005a). All Leq and Lavg measures were assigned a minimum value of 69.9 dB, the lowest level the dosimeters accurately record. Every 1-min interval was examined to ensure that Lavg < Leq < Lmax. If any of the levels violated these assumptions, the Leq was assumed correct and measures were adjusted down (for Lavg) or up (for Lmax) to the Leq (Seixas et al., 2005a). Full-shift measurements with errors in >10% of 1-min observations were removed from the database, as were full-shift measurements with a Leq/Lavg ratio >50 or a Lmax/Leq ratio >1000. After cleaning, dosimetry data were merged with survey and timeline log data and exported to Stata 9.1 (StataCorp, College Station, TX, USA) for analysis.
Arithmetic mean and median levels and standard deviations (SDs) were computed for 1-min Leq levels within and across shifts for the five timeline log perceived intensity categories and the four timeline log perceived variability categories, respectively. The means of the 1-min Leq/Lavg and Lmax/Leq ratios (Seixas et al., 2005a) were computed within each shift, as was the percentage of 1-min intervals within each shift where the Leq exceeded 85 dBA. Full-shift Leq levels were computed for individual i on shift j using equation (1):
 | (1) |
Categorical survey data were analyzed using Pearson's chi-square or Fisher's exact test; continuous survey data were analyzed using one-way ANOVA. Mean first shift and two-week Leq levels and percent of time spent >85 dBA were computed by response category for the perceived intensity survey items. Mean Lmax/Leq and Leq/Lavg ratios were computed by response categories for the peakiness and noise variability survey items, respectively. Where mean noise levels associated with survey item response categories did not increase in the expected direction, categories were collapsed in a systematic fashion. Items with four original categories were collapsed into two levels (two higher versus two lower categories). Items with five original categories were collapsed into three levels by combining the middle three categories. Spearman correlation coefficients were used to summarize the relationship between survey item responses and dosimetry noise levels. Variables that could potentially influence perceived exposure (e.g. hearing protection use, perceived hearing sensitivity, noise annoyance, sensitivity to noise and feelings toward noise exposure) were explored through the use of two-way scatter plots and Spearman correlations of measured levels associated with item responses, as well as descriptive statistics of measured noise levels stratified by item responses.
To assess the ability of the perceived intensity survey items to correctly identify workers with Leq levels 
85 dBA, item sensitivity and specificity were computed. Measured Leq levels were treated as the true exposure (85 or <85 dBA), and survey items were collapsed into binary categories using two strategies (lowest category versus all higher categories and highest category versus all lower categories). Sensitivity was computed as 
, where TP is the number of exposures >85 dBA (true positives) correctly identified by survey binary categories and FN is the number of incorrectly identified exposures >85 dBA (false negatives). Specificity was computed as , where TN is the number of exposures <85 dBA (true negatives) correctly identified by survey binary categories and FP is the number of incorrectly identified exposures <85 dBA (false positives). Sensitivity and specificity values range from zero to one; larger values indicate a higher probability of correctly identifying an overexposed individual (sensitivity) or correctly identifying an individual exposed <85 dBA (specificity).
To evaluate the first hypothesis concerning contrast between exposure groups, a one-way ANOVA (Stata ‘loneway’) was used to compute contrast values for first shift and 2-week Leq levels for exposure groups defined by job title and survey item. Contrast is computed as 
, where 
and 
represent between- and within-group variance, respectively (Kromhout and Heederik, 1995). To evaluate the second hypothesis, which concerned the precision of exposure estimates, first shift and 2-week exposure predictions were created for each subject. Using job title and the three survey items demonstrating the greatest exposure contrast, exposures were predicted via linear regression (Stata ‘regress’ and ‘predict’ commands), with measured Leq levels as the modeled outcome. Predictions for the percent time in noise item were made using the response category with the highest reported percentage of time. Bias was computed as the mean difference between the predicted and measured levels, and precision was computed as the SD of the differences between predicted and measured levels (Reeb-Whitaker et al., 2004); smaller bias and precision values therefore indicate better performance. Interactions between job title and survey items were explored through all-site analyses to evaluate any potential differences in perceived noise levels by job title.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
The majority of subjects in the study were white (68%), male (90%) and spoke English as their primary language (97%); they had an average age of 38 ± 11 (SD) years and had worked 10 ± 9.3 years in the job title. Subjects at the three sites were demographically similar, with the exception of race; the continuous site had more non-white participants (60%) than the other sites (20%, Fisher's exact test, P < 0.001). Fifteen subjects at the highly variable site were classified as carpenters and five as laborers. All 20 subjects at the intermittent site were sheet metal workers. Eighteen of the continuous site subjects were warehouse operators and two were maintenance staff.
Noise measurements
A total of 206 valid full-shift noise measurements were made over 16 monitoring days. An additional 17 measurements were unsuccessful due to equipment malfunction, for a failure rate of 8%, consistent with previous research using similar instruments (Neitzel et al., 1999; Neitzel et al., 2004). Data cleaning resulted in corrections to 
0.5% of 1-min observations, and no full-shift samples were eliminated through this process. The number of valid dosimetry measurements per subject was 3.4 ± 0.8. Thirty-six of 60 subjects (60%) completed all 4 days of monitoring and 52 (87%) completed 3 days.
The three sites had significantly different first shift and 2-week Leq levels and Leq/Lavg and Lmax/Leq ratios (Table 1, one-way ANOVA, P < 0.05). The percentage of time spent >85 dBA did not vary significantly between sites. All sites had a large fraction of first shift and 2-week Leq levels that exceeded the 85 dBA exposure limit used by most regulatory agencies, and these fractions were significantly different between sites (Kruskal–Wallis, P < 0.05). The highly variable site had the highest mean first shift and 2-week Leq levels and mean Lmax/Leq ratio values. The continuous site had the lowest mean Leq/Lavg and Lmax/Leq ratios. The ratio metrics correctly identified the continuous site as having the most steady and least ‘peaky’ noise, the highly variable site as having the most peaky noise and the intermittent site as having the most intermittent noise (i.e. highest mean Leq/Lavg ratio).
|
Median 1-min Leq dosimetry levels for the five response categories of the timeline log perceived intensity item generally increased in the expected direction (Fig. 1), although several categories at the highly variable and intermittent sites had similar or slightly reversed levels. There was also overlap between the levels of the response categories within site. Mean 1-min Leq levels associated with the ‘normal’ and ‘much louder’ categories differed significantly by site (one-way ANOVA, P < 0.05). The SDs of the 1-min Leq levels also generally increased in the expected direction across the five timeline log perceived variability categories. Mean levels for the ‘always variable, not at all constant’ category differed significantly by site (one-way ANOVA, P = 0.01). Overall, the results suggest that subjects can to some degree identify changes in noise intensity and variability within a shift.
|
Mean first shift and 2-week Leq/Lavg and Lmax/Leq ratios were computed by response category for the variability and peakiness survey items, respectively, and trends across categories were analyzed via one-way ANOVA. Noise levels did not increase in a consistent fashion across the original scales for these items, so the variability and peakiness scales were collapsed to two- and three-point scales, respectively. After collapsing, mean levels generally increased in the anticipated direction (Table 2). Significant within-site associations were found between the noise variability item and mean Leq/Lavg ratio at the continuous and intermittent sites for the first shift measures. For the 2-week measures, significant within-site associations were found between the noise variability item and the mean Leq/Lavg ratio at the continuous and highly variable sites. Significant associations were found between the peakiness item and the mean Lmax/Leq ratio for first shift and 2-week measures at the highly variable site. Overall, Leq/Lavg ratios were best explained by the noise variability survey item at the continuous site, and the peakiness survey item best explained Lmax/Leq ratios at the highly variable site.
|
Survey items
Examination of variables that could potentially influence perceived noise intensity (e.g. use of hearing protection, perceived hearing sensitivity, noise annoyance and sensitivity and feelings toward noise) did not identify any relationships between these variables and perceived intensity. These variables were therefore not controlled for in the analyses presented here.
When perceived intensity survey item responses were collapsed into lowest category versus higher categories, items had high sensitivity (0.81–0.92 for first shift and 0.85–0.98 for 2-week measures) in correctly identifying Leq levels of 85 dBA, but much lower specificity (first shift values of 0.08 to 0.23 and two week values of 0.14 to 0.43) in correctly identifying Leq levels <85 dBA. When items were collapsed into highest category versus lower categories, specificity was generally high (0.77–0.85 for first shift and 0.43–1.0 for 2-week measures) but sensitivity was much lower (0.08–0.35 for first shift and 0.10–0.44 for 2-week measures). The item with the highest sensitivity (using lowest category versus higher categories) was raise voice, with first shift and 2-week values of 0.81 and 0.94, respectively. The item with the greatest specificity (using highest category versus lower categories) was noise frequency, with first shift and 2-week values of 0.85 and 1.0, respectively.
Mean first shift and 2-week Leq levels associated with the lowest and highest survey item response categories generally showed appreciable differences, but mean levels did not increase consistently across the middle response categories. After items were collapsed from five to three categories, mean Leq levels generally increased in the correct direction (Table 3). The greatest ranges in mean Leq levels across categories were found for the items percent time in noise (3–5 dBA), usual noise (3–4 dBA) and noise frequency (1–3 dBA). The range across job titles was 9–10 dBA. One-way ANOVA analyses identified one item with a significant trend across response categories (percent time in noise, P = 0.01) for the first shift measures and none for the 2-week measures. The difference in job title mean levels was significant for both first shift and 2-week measures (P < 0.0001).
|
When the mean percentage of time spent >85 dBA by job title and survey item was evaluated (data not shown), significant trends across response categories were found for the raise voice and percent time in noise survey items for the first shift measures (one-way ANOVA, P < 0.05). For the 2-week measures, the noise frequency item had a significant trend (P = 0.01). Differences between job titles were never significant.
Table 4 presents the results of the contrast analysis. Overall, the survey items performed best at identifying distinct exposure groups at the highly variable site, where three of six first shift and five of eight 2-week items provided contrast greater than was indicated by job title. At the intermittent site (where only a single job title was evaluated), three of the six first shift items and one of the 2-week items provided non-negligible contrast. No subjective items at the continuous site provided greater contrast than did job title. The three survey items that generally showed the greatest contrast were percent time in noise, noise frequency and raise voice. Only one contrast value, job title for the first shift measures at the continuous site, reached significance.
|
Table 5 presents R2, bias and precision values from the prediction models of measured Leq levels by job title and survey item. Bias was negligible since the predictions were compared to the original data set from which they were estimated. Greater bias would be expected from a validation data set. The precision of survey item predictions was essentially the same as, or slightly smaller (better) than, those of job title. As with bias, the precision values are likely smaller than would be found using a validation data set. Where survey items showed better precision than job title, the improvement was a fraction of a decibel. The amount of variance explained by the predictions (R2) was similar to or better than that of job title at the highly variable and intermittent sites. At the continuous site and across all sites, job title had the highest R2 values. Interaction terms between job title and survey items were explored, but never reached statistical significance, and were not retained. The addition of survey items to job title in the prediction models improved R2 values for the continuous and highly variable sites by a considerable amount. The utility of the survey items at the intermittent site was less than at the other sites.
|
Table 6 presents regression coefficients for job title, the three survey items with the highest contrast and combinations of job title and survey item. These coefficients can be used to predict exposures given information about workers’ job titles and survey item responses. Survey item coefficients are generally smaller than those of job title. Standard errors associated with the survey items and with the job title ‘warehouse maintenance’ were generally large compared to the coefficients themselves, indicating that these estimates are not particularly stable.
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
The current study assessed an alternative means of assessing noise exposure using workers’ subjective perceptions of their own exposure levels. Evaluation of perceived exposures in three different noise environments suggested that perceived exposure information is of greatest use at worksites with continuous and highly variable noise and possibly of less use for worksites with intermittent noise. Within-shift analyses suggested that subjects in all three noise environments maybe able to perceive differences in exposure levels and variability within a workshift, and cross-shift analyses indicated that survey items can create exposure groups with contrast and precision similar to, or slightly better than, that provided by job title. The ratio metrics suggested by Seixas et al. (2005a) performed as intended, identifying the continuous site as having the lowest mean Leq/Lavg and Lmax/Leq, the highly variable site as having the highest mean Lmax/Leq and the intermittent site as having the highest Leq/Lavg. Overall, the results of this study suggest that perceived exposure data may provide new information about occupational noise exposures—especially when considered in combination with job title—which may be particularly useful in situations where direct Leq measurements on workers are infeasible.
Trends in Leq levels across response categories for some survey items were in the expected direction, but some items had no differences across categories. All survey item responses were collapsed to strengthen trends across categories. Only one survey item for the first shift measures, and no 2-week items, had statistically significant trends in noise levels across categories, though several items had significant trends for time spent >85 dBA. At the continuous and highly variable sites where multiple job titles were studied, the differences in mean noise levels between job titles were only statistically significant for the first shift measures at the highly variable site. Contrast between exposure groups defined by job title was exceeded by most survey items at the highly variable site. This finding provides support for the first hypothesis that subjective survey item responses would create groups with greater exposure contrast than that provided by job title. At the intermittent site, exposure could not be evaluated by job title, and survey items generally yielded groups with minimal contrast. The bias of the predictions made from job title and survey items was negligible since the predictions were compared to the original data set from which they were estimated. These bias values need to be reevaluated in a validation data set. The precision of predictions modeled from survey items was equivalent to or slightly better than that provided by job title, which supports the second study hypothesis that the precision associated with subjective item predictions would be better than that associated with job title. When survey items were added to job title in the prediction models for the continuous and highly variable sites, precision was improved beyond that provided by either job title or survey item alone, and model fit was improved, in some cases substantially.
The generally poor performance of the subjective assessment approach at the intermittent worksite may be due to subject difficulties in summarizing intermittent exposure or accounting for periods of low noise. The continuous and highly variable sites rarely had periods of quiet, while the intermittent site had a number of quiet periods in each work shift. This wide range of exposure levels of varying durations may simply have been too difficult for subjects to mentally average, or subjects may have focused on periods of intense exposure and discounted quieter periods. The modest trends in noise levels across survey item responses at all sites may also be at least partially due to the small sample size of 20 subjects per site. Finally, it is possible that Leq measures are not well matched to the way the human ear perceives sound. Although the Leq is widely considered more representative of the risk of NIHL from continuous, intermittent and impulsive noise than the Lavg (NIOSH, 1998), the larger 5 dB Lavg time–intensity exchange rate may correlate better with human perception of loudness, which doubles with an increase of 10 decibels (Kryter, 1994). To evaluate this possibility, all analyses of the relationships between noise-related survey items and first shift and 2-week average noise levels were repeated using the Lavg. Lavg measures were found to perform slightly better than the Leq for nearly all survey items. However, since the Leq is the better representation of NIHL risk and is the primary metric previously utilized for exposure assessment in the Seattle construction worker cohort and since the relationship between Leq and Lavg can vary unpredictably with noise level variability, the Leq was considered the better metric for use here.
A literature search identified only two previous studies that compared self-reported noise exposure to direct noise measurements. Ising et al. (1997) had 80 randomly selected workers rate their exposures as being as loud as one of five response categories (refrigerator, typewriter, electric lawn mower, electric drill or pneumatic drill). Ratings were compared to a 1-min Leq level measured in each subjects’ workplace. Median noise levels for the five categories increased in the expected direction, and the correlation between measured and self-reported levels (r = 0.84) was much higher than that seen in the current study. The variance of 1-min levels ranged from 6 to 24 dBA across the five categories, and, as with the timeline log perceived intensity item in the current study, there was significant overlap between categories. The authors did not provide information about subjects’ occupations or what types of workplaces were evaluated; however, 1-min Leq levels would not be representative of long-term average exposures in workplaces with variable or intermittent noise. The timing of the 1-min Leq measurement and survey response was also not described; if subjects answered the survey item while the measurement was being made, the elements of recall and mental averaging of exposure levels inherent to the current study would not have been present in the study of Ising et al.
In the second study of perceived noise, Ahmed et al. (2004) collected information on noise exposure from 259 noise-exposed workers and measured each subject's full-shift noise level. Subjects indicated whether the noise level where they worked was ‘high’ and whether they had to shout to be heard over noise. The question about high noise had high sensitivity (0.93) and lower specificity (0.40) in identifying workers exposed >85 dBA, while the question on shouting had lower sensitivity (0.68) and higher specificity (0.75). The current study used two items similar to questions used by Ahmed et al. (noise frequency was relatively similar to the ‘high noise’ question, and raise voice was similar to the ‘shout’ question) and found one item to be highly sensitive and the other highly specific, as did Ahmed et al. However, the item results were reversed: in the current study, the item raise voice was found to have high sensitivity (0.94) but low specificity (0.20), and the item noise frequency had low sensitivity (0.18) but high specificity (1.0). The reason for this reversal of results is unclear. The current study confirms that screening survey items can assist in identification of workers with exposure exceeding allowable limits. However, only crude exposure predictions can be developed using this approach, and it is therefore of limited use for epidemiological NIHL studies.
Analyses of variables with the potential to influence the perceived intensity of noise (e.g. use of hearing protection, perceived hearing sensitivity, noise annoyance, sensitivity to noise and feelings toward noise exposure) failed to demonstrate any relationships between these variables and the survey items, suggesting that these variables do not need to be controlled for in analyses of perceived exposure. This is consistent with previous research (Belojevic and Jakovljevic, 2001; Miedema and Vos, 2003) suggesting that noise sensitivity has at most a very weak relationship with perceived intensity and that individuals with hearing loss perceive noise levels above their hearing threshold as a normal hearing individual would through the phenomenon of recruitment (Kryter, 1994).
The performance of subjective measures compared to measured levels was generally best among workers in highly variable and continuous noise and worst among workers with intermittent exposure. The results of this study must be considered in light of the small sample size and limited number of job titles evaluated. The job titles assessed were completely nested within site, so it was not possible to evaluate the performance of the perceived exposure items for a single job title across sites. The relationship between survey item and job title-based exposure predictions and measured Leq levels must be further validated in a separate population prior to use of this strategy in external exposure assessments, as the current study could only compare predicted levels with the mean exposure levels used to create the predictions. The performance of the subjective measures should also be evaluated over a longer period of exposure, e.g. months instead of weeks.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Assessment of exposure in epidemiological studies is problematic in environments with inconstant exposures, and even when an appropriate exposure metric has been identified, measurement error associated with highly variable exposures contributes to uncertainties regarding the risk of chronic diseases such as NIHL. The results of this study suggest that subjective assessment has the potential to be used as a sensitive and specific screening tool to identify overexposed workers for compliance purposes. Use of this approach may also improve the contrast and precision of epidemiological exposure estimates for workers with highly variable exposure profiles, especially when direct, repeated exposure measurements on such workers are infeasible, as is the case in the ongoing prospective study of noise and NIHL among Seattle construction workers. Enhanced precision and contrast resulting from use of subjective measures as alternative or complementary exposure assessment techniques may contribute to increased understanding of the often subtle changes in health which are measured in most epidemiological studies.
It is unlikely that any single quantitative or subjective exposure estimation approach will produce estimates with optimal bias and precision. A more promising approach involves combination of information from multiple sources of exposure data (e.g. job title, task-based evaluation and subjective information) to produce a hybrid estimate of exposure. Incorporation of the job title mean level in this hybrid estimate would reduce bias, while incorporation of individual-level task-based and perceived estimates could improve precision. Although methods have previously been developed to combine information from multiple quantitative exposure estimates, the authors are not aware of any studies which have attempted to combine subjective and quantitative estimates. Future research on exposure assessment for workers with highly variable exposures should explore the integration of multiple sources of exposure information to produce precise and unbiased exposure estimates.
|
FOOTNOTES |
|---|
The free full text of this article can be found in the online version of this issue.
Received June 24, 2008; in final form August 20, 2008
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
ACGIH. Threshold limit values for chemical substances and physical agents and biological exposure indices (BEIs): noise. (2006) Cincinnati, OH: American Conference of Governmental Industrial Hygienists.
Ahmed HO, Dennis JH, Ballal SG. The accuracy of self-reported high noise exposure level and hearing loss in a working population in Eastern Saudi Arabia. Int J Hyg Environ Health (2004) 207:227–34.[CrossRef][Web of Science][Medline]
ANSI. Determination of occupational noise exposure and estimation of noise-induced hearing impairment (S3.44-1996) (1996) New York, NY: American National Standards Institute. Report No.: S3.44-1996.
Belojevic G, Jakovljevic B. Factors influencing subjective noise sensitivity in an urban population. Noise Health (2001) 4:17–24.[Medline]
Checkoway H, Pearce N, Kriebel D. Research methods in occupational epidemiology (2004) 2nd. New York, NY: Oxford University Press.
Daniell WE, Swan SS, McDaniel MM, et al. Noise exposure and hearing loss prevention programmes after 20 years of regulations in the United States. Occup Environ Med (2006) 63:343–51.
Earshen J. Chapter 3: sound measurement: instrumentation and noise descriptors. In: The noise manual—Berger E, Royster L, Royster J, Driscoll D, Layne M, eds. (2000) 5th. Fairfax, VA: American Industrial Hygiene Association. 41–100.
Hessel PA. Hearing loss among construction workers in Edmonton, Alberta, Canada. J Occup Environ Med (2000) 42:57–63.[CrossRef][Web of Science][Medline]
Ising H, Babisch W, Kruppa B, et al. Subjective work noise: a major risk factor in myocardial infarction. Soz Praventivmed (1997) 42:216–22.[CrossRef][Web of Science][Medline]
Jokitulppo J, Toivonen M, Bjork E. Estimated leisure-time noise exposure, hearing thresholds, and hearing symptoms of Finnish conscripts. Mil Med (2006) 171:112–6.[Web of Science][Medline]
Koushki PA, Kartam N, Al-Mutairi N. Workers’ perceptions and awareness of noise pollution at construction sites in Kuwait. Civ Eng Environ Sys (2004) 21:127–36.[CrossRef]
Kromhout H, Heederik D. Occupational epidemiology in the rubber industry: implications of exposure variability. Am J Ind Med (1995) 27:171–85.[Web of Science][Medline]
Kryter K. The handbook of hearing and the effects of noise: physiology, psychology, and public health (1994) London: Academic Press.
Miedema HM, Vos H. Noise sensitivity and reactions to noise and other environmental conditions. J Acoust Soc Am (2003) 113:1492–504.[CrossRef][Web of Science][Medline]
Neitzel R, Seixas N. The effectiveness of hearing protection among construction workers. J Occup Environ Hyg (2005) 2:227–38.[CrossRef][Web of Science][Medline]
Neitzel R, Seixas N, Olson J, et al. Nonoccupational noise: exposures associated with routine activities. J Acoust Soc Am (2004) 115:237–45.[CrossRef][Web of Science][Medline]
Neitzel R, Seixas NS, Camp J, et al. An assessment of occupational noise exposures in four construction trades. Am Ind Hyg Assoc J (1999) 60:807–17.[Web of Science][Medline]
Nelson DI, Nelson RY, Concha-Barrientos M, et al. The global burden of occupational noise-induced hearing loss. Am J Ind Med (2005) 48:446–58.[CrossRef][Web of Science][Medline]
NIOSH. Criteria for a recommended standard: occupational noise exposure, revised criteria 1998 (1998) Cincinnati, OH: US Department of Health and Human Services, Centers for Disease Control and Prevention, National Institute for Occupational Safety and Health. Report No.: DHHS (NIOSH) 98-126.
OSHA. Occupational noise exposure: hearing conservation amendment; final rule (1983) Washington, DC: US Department of Labor.
Palmer KT, Griffin MJ, Syddall HE, et al. Occupational exposure to noise and the attributable burden of hearing difficulties in Great Britain. Occup Environ Med (2002) 59:634–9.
Palmer KT, Haward B, Griffin MJ, et al. Validity of self reported occupational exposures to hand transmitted and whole body vibration. Occup Environ Med (2000) 57:237–41.
Passchier-Vermeer W. Noise-induced hearing loss from exposure to intermittent and varying noise (1973) Washington, DC: US Environmental Protection Agency. Report No.: EPA 550/9-73-008.
Petrick ME, Royster LH, Royster JD, et al. Comparison of daily noise exposures in one workplace based on noise criteria recommended by ACGIH and OSHA. Am Ind Hyg Assoc J (1996) 57:924–8.[Web of Science][Medline]
Rappaport SM, Kromhout H, Symanski E. Variation of exposure between workers in homogeneous exposure groups. Am Ind Hyg Assoc J (1993) 54:654–62.[Web of Science][Medline]
Reeb-Whitaker CK, Seixas NS, Sheppard L, et al. Accuracy of task recall for epidemiological exposure assessment to construction noise. Occup Environ Med (2004) 61:135–42.
Robinson GS, Casali JG. Chapter 14: speech communications and signal detection in noise. In: The noise manual—Berger E, Royster L, Royster J, Driscoll D, Layne M, eds. (2000) 5th. American Industrial Hygiene Association. 567–600.
Sataloff R, Sataloff J. Occupational hearing loss (1987) New York, NY: Marcel Dekker, Inc.
Seixas N, Kujawa S, Norton S, et al. Predictors of hearing threshold levels and distortion product otoacoustic emissions among noise exposed young adults. Occup Environ Med (2004) 61:899–907.
Seixas N, Neitzel R, Sheppard L, et al. Alternative metrics for noise exposure among construction workers. Ann Occup Hyg (2005a) 49:493–502.
Seixas N, Sheppard L. Maximizing accuracy and precision using individual and grouped exposure assessments. Scand J Work Environ Health (1996) 22:94–101.[Web of Science][Medline]
Seixas NS, Goldman B, Sheppard L, et al. Prospective noise induced changes to hearing among construction industry apprentices. Occup Environ Med (2005b) 62:309–17.
Seixas NS, Ren K, Neitzel R, et al. Noise exposure among construction electricians. Am Ind Hyg Assoc J (2001) 62:615–21.
Seixas NS, Sheppard L, Neitzel R. Comparison of task-based estimates with full-shift measurements of noise exposure. AIHA J (Fairfax, Va) (2003) 64:823–9.
Teschke K, Olshan AF, Daniels JL, et al. Occupational exposure assessment in case-control studies: opportunities for improvement. Occup Environ Med (2002) 59:575–93. discussion 94.
White E, Armstrong B, Saracci R. Principles of exposure measurement in epidemiology: collecting, evaluating, and improving measures of disease risk factors (2008) 2nd. New York, NY: Oxford University Press.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||