What can 35 years and over 700,000 measurements tell us about noise exposure in the mining industry? (2024)

Abstract

Keywords: Occupational noise, mining, exposure modeling, big data

Introduction

Noise is one of the most common occupational exposures in the United States (US). Tak et al. (2009) estimated that 22 million workers were exposed to hazardous noise levels based on self-reported noise exposure. The National Institute for Occupational Safety and Health (NIOSH) estimates that over four million American workers are potentially exposed to hazardous noise >85 dBA, and that the excess risk of noise-induced hearing loss (NIHL) at this exposure level ranges from 8 to 16% depending on the hearing loss metric and statistical model used (NIOSH, 1998). NIHL is among the top ten leading work-related illnesses and injuries identified by NIOSH. The overall prevalence of hearing loss in the working population is estimated to be 11.4%, while the railroad and mining industries have the highest and second-highest prevalence (34.8% and 24.3%, respectively) (). NIOSH estimated that, on average, 100–200 coal, 50–150 metal and 10–40 non-metal workers per 100,000 full-time workers experienced hearing loss each year. This accounted for around one fifth of the total reported injury cases in the mining sector (NIOSH, 2000). NIOSH also estimated that by the age of 50, 90% of miners will have developed a hearing loss >25 dB at the 1, 2, 3, and 4, kHz frequency (NIOSH, 2000).

The estimated economic cost of hearing loss varies widely. The World Health Organization (WHO) estimated the cost of hearing loss to be between 0.2 to 2% of gross domestic product (GDP) for developed countries (WHO, 1997). Emmett and Francis (2015) further found that hearing loss was independently associated with lower educational achievement and lower income than those without hearing loss. The cost of compensation for hearing loss in US military Veterans alone was over $1.2 billion in 2006 (). Recently Neitzel et al. estimated that the US could save between $52 and $152 billion each year if 20% of hearing loss from hazardous noise was prevented (Neitzel et al., In Press) Additionally, there is a growing body of evidence that noise exposure may be associated with a number of important non-auditory health effects, including cardiovascular disease (Basner et al., 2014). These effects may be particularly evident among miners, as mining has traditionally been considered one of the noisiest industries in the US (Tak et al., 2009).

The Mine Safety and Health Administration (MSHA) was established by the Federal Mine Safety and Health Act of 1977 to promulgate and enforce health and safety regulations for the mining industry (Federal Mine Safety and Health Act, 1977). The permissible exposure limit (PEL) for noise was set at 90 dBA as an eight-hour time weighted average (TWA) with a 5 dB time-intensity exchange rate for all sound levels from 90 to at least 140 dBA (Federal Mine Safety and Health Act, 1977). In addition, the Federal Mine Safety and Health Act established regulations regarding requirements for use of noise controls and hearing protectors (and, in the case of coal mines, implementation of hearing conservation programs, HCPs) that varied depending on the type of mine. In 1999 MSHA published a revised rule on occupational noise which harmonized the rules regarding HCPs and the implementation of noise controls in all mines in the United States. In addition, the new rules established an action level (AL) of 85 dBA as a TWA with a 5 dB exchange rate for sound levels between 80 and at least 130 dBA, as well as harmonized requirements for HCPs (30 CFR Part 62, 1999).

To help fulfill its mandate, MSHA conducts routine noise monitoring inspections in mines of all types, and amassed a dataset of over 700,000 noise dosimetry measurements from 1979 to 2014. Most of these measurements include information on the type of mine (facility, surface, or underground), what was being mined (coal, metal, or non-metal) whether the measurement was made using the PEL or AL criteria, and job title or task description for each measurement. In 2007, Joy and Middendorf (2007) conducted an analysis of noise measurements in coal mines from 1986 to 2004. This analysis yielded important insights into noise in US mines, but was limited by the short (four-year) time period for which data were available following implementation of MSHA’s revised noise regulation in 2000. There has not been a comprehensive analysis of noise exposure in the mining industry since the analysis by Joy and Middendorf (2007). The continued high prevalence of hearing loss among workers in the mining industry warrants another careful analysis noise exposure in the mining industry (Masterson et al., 2016).

Measurements from this dataset were analyzed as part of a larger job exposure matrix (JEM) for occupational noise. Our study had two goals intended to increase our understanding of past and present noise exposure in the mining industry and to help predict future exposures so that adequate controls can be implemented to protect workers’ health. The first goal of this analysis was to describe and evaluate trends in measured occupational noise levels among US miners from 1979 to 2014. The second goal was to use the measurements in the dataset to build a statistical model that could be used to estimate a worker’s occupational noise exposure based on their job title, and the type of mine.

Methods

Results

Prior to data cleaning there were a total of 619,028 PEL measurements and 283,169 AL measurements available. Table 1 summarizes the steps in the data cleaning process and the number of measurements eliminated for each exclusion criteria. The largest loss of PEL measurements was the result of missing information regarding what type of material was being mined. The largest loss of AL measurements occurred because the TWA measurements were below 60 dBA. In total, 120,159 (19.4%) PEL and 7,421 (2.7%) AL measurements were removed from the dataset, leaving 498,869 and 275,748 valid PEL and AL measurements. The mean PEL measurement prior to 2000 was 84.4 dBA with a standard deviation (SD) of 8.2 compared to a mean of 79.9 ± 8.6 dBA after 2000; this difference was highly significant (p < 0.0001). The mean AL (post-2000) was 83.7 ± 6.9 dBA.

Figure 1 shows that the average TWA for PEL measurements have been steadily decreasing over time. AL measurements, made starting in 2000, followed a similar pattern, with higher measured levels than those indicated by the PEL due to the different measurement ranges used (80–130 dBA for the AL versus. 90–140 dBA for the PEL). Table 2 summarizes the changes in average noise exposure before and after the year 2000 stratified by what material was being mined and what type of mine the measurements came from. On average, the PEL measurements decreased by 4.5 dBA for measurements made after the implementation of MSHA’s updated noise regulation in 2000. The greatest reduction in PEL exposures was seen in non-metal mines, where measurements decreased by 6.5 dBA, while coal and metal mines only by 2.7 and 2.8 dBA respectively, and underground coal mines only decreased by 0.8 dBA. The average AL measurements ranged from 81.9 in surface non-metal mines to 87.9 dBA in underground metal mines.

Figure 2 summarizes the percentage of measurements that exceeded the 85 dBA AL and 90 dBA PEL. Prior to the year 2000, 23.5% of all TWA measurements exceeded the MSHA PEL of 90 dBA, and underground metal mining had the highest noise exposures of all mine types, with 44.8% of the TWA measurements exceeding 90 dBA. Following the implementation of MSHA’s revised noise regulation in 2000, 21.7% of measurements exceeded the AL and 7.0% exceeded the PEL. After the year 2000, underground metal mining continued to have the greatest percentage of measurements (17.4%) that exceeded the PEL, while underground coal mining had the greatest percentage of measurements (48.4%) that exceeded the AL.

Table 3 shows the percentage of PEL TWA measurements pre- and post-2000 that exceeded a range of cutpoints (85, 90, 105, and 115 dBA), stratified by material mined and mine type. The percentage of measurements exceeding each of the four cutpoints dropped after the year 2000 across all mine types and materials mined, with the largest reductions in measurements exceeding these cutpoints seen in underground non-metal mines, and the smallest reductions seen in underground coal mines.

There were a total of 45 different job titles in this dataset. Appendix 1 provides the mean, standard deviation, and number of measurements for each job title before and after the year 2000. The measurements for the 45 job titles were collapsed into broad occupational groups based on their assigned SOC codes. Figure 3 provides a box plot of the distribution of measurements for each broad occupational group. The mining exposure group had the highest median exposure both before and after the implementation of MSHA’s noise standard in 2000. The miner, production, and other groups all had very similar medians but the miner group had a larger number of statistical outliers than the other group suggesting that the likelihood of exposures greater than 105 dBA is higher in this group.

The regression coefficients for PEL and AL models created using the training dataset are presented in Appendix 2. The PEL model contained measurements from 1979 through 2014 while the AL model contained measurements from 2000 through 2014. In both models the year variable was centered to the first year that measurements were collected. The adjusted-R² for the models were 0.1540 for the PEL and 0.1339 for the AL model. When controlling for job title, material being mined, and the type of mine, both models predicted that noise exposure has decreased over time (−0.331 and −0.243 dBA per year for the PEL and AL model). This indicates that on average noise levels in the mining industry have decreased by about 0.3 and 0.2 dBA each year for PEL and AL measurements. Coal mines were predicted to be noisier than metal and non-metal mines in the PEL model but metal mines were found to be noisier than coal and non-metal mines in the AL model. Underground mines were found to be noisier than facility and surface mines in both the PEL and AL models. Roof bolters were estimated to have the highest exposure in the PEL model while landscaping and grounds keeping workers were estimated to have the highest exposure in the AL model.

The overall mean for the predicted values from the training dataset PEL model was 82.0 ± 3.4 dBA compared to a mean of 82.0 ±8.7 dBA for the validation dataset. The interquartile range (IQR) for the predicted values was 5.0 dBA compared to 12.2 dBA for the validation data. For each job title, predictions from the PEL model were on average 0.9 dBA different than the actual measurements in the validation dataset. The results of the t-tests found that six job titles had significantly different (p<0.05) predicted and actual mean exposures. Only three job titles had a mean difference greater than 2 dBA and these job titles had a smaller number of measurements compared to other job titles in the dataset. The overall mean for the predicted values from the AL model was 83.7 ± 2.5 dBA while the validation data had a mean of 83.7 ± 6.9 dBA. The IQR was 3.4 and 8.7 dBA for the predicted and validation datasets, respectively. For each job title, predictions from the AL model were on average 0.7 dBA different than the validation values. Two job titles were found to have predicted values significantly different from the validation values, the difference between the predicted and validation values for both job titles were less than 2 dBA.

Discussion

The results from this analysis indicate that mean noise exposure in the mining industry has been decreasing every year. This concurs with the results from Joy and Middendorf (2007), who found that the overall annual median noise dose declined 67% for surface coal mining and 24% for underground coal mining from 1986 to 2004. The reductions in exposure noted in our analyses are likely due, at least in part, to the implementation of MSHA’s revised noise regulation in 2000 and to improvements in mining technology and noise control (Kovalchik et al., 2007; Smith et al., 2011). However, this reduction in noise exposure does not appear to be evenly distributed among different types of mines, nor has it been completely monotonic. Workers in underground coal mines in particular had a smaller decrease in noise exposure than workers in other mine types mining other materials when comparing measurements before and after the implementation of MSHA’s noise standard. There was also a small increase in the percentage of facility coal miners exposed to noise >105 dBA and underground coal miner exposed to noise > 115 dBA. Our analysis does not allow us to know why this increase has occurred or if it is statistically significant. However, it is possible that as mining technology becomes more automated and requires less workers a small sub-group of miners may be exposed to very high levels of noise as they operate machinery.

Regardless of what types of materials were being mined, workers in underground mines were found to have the highest noise exposure. This is not surprising, as underground mine work involves use of noisy heavy equipment in tightly-enclosed, reverberant spaces (Peterson et al., 2006). This suggests that additional resources should be directed to design and implement new noise control technologies that can be used in underground mines. Prior to 2000, coal, metal, and non-metal mines had a similar mean exposure level; however, after the year 2000, non-metal mines experienced a much larger decrease in noise exposure than coal and metal mines. A portion of this difference can be attributed to the smaller percentage (5.9%) of underground mine measurements from the non-metal mining sector that were taken post-2000 compared to underground coal (52.5%) and metal (44.2%) mines. The mean exposure in underground mines in the non-metal mining sector was still 2–3 dBA lower than coal or metal mining. This difference could be caused by the differences in tools and techniques for extracting coal and metals compared to non-metals (Peterson et al., 2006), or perhaps by differing production demands.

Following the implementation of MSHA’s revised noise regulation, noise exposures dropped for all broad occupational groups. Pre-2000, between 15 and 45% of all PEL measurements exceeded the 90 dBA PEL, while post-2000, 3–17% percent exceeded 90 dBA. Following the introduction of the revised regulation, the median AL exposures in both miner and production groups exceeded the 85 dBA AL, suggesting that workers in these groups should be the focus of further efforts to reduce noise exposure in the mining industry.

The mean training dataset predictions from the linear models were very close to the mean measurements in the validation dataset despite having a relatively low adjusted-R² (0.1540 and 0.1339 for the PEL and AL models respectively). This occurred because the large number of samples present in both models and the validation training sets results in a very stable and unbiased mean exposure estimate for each job title (). This is the primary advantage of working with large datasets and makes the predictions generated by these models useful for both establishing a past exposures and helping predict future exposures for groups of workers. However, it is very important to recognize that there is an inherent variability in an individual worker’s exposure from day to day due to a number of factors, including the implementation of controls, workload, and personal work habits (Kromhout et al., 1993). As a result, the predictions from the model should not be used in place of noise monitoring. The best use of the model would be to predict mean yearly exposures to noise for groups of workers in each of the mine types and materials mined in order to help predict the risk of developing hearing loss in the future.

There are some limitations that need to be considered when using this model. The first is the possibility of error in exposures estimates due to misclassification of some job titles. We attempted to reduce this risk by using the SOC database to standardize job titles, and by removing measurements where a SOC code could not be assigned. It is also important to consider that grouping workers by job titles does not guarantee that all those workers have similar exposures (Rappaport et al., 1993). Another limitation is that 123,031 measurements could not be included in this analysis because they met the exclusion criteria. The majority (91,231) of these measurements were excluded because they did not provide any information on what material was being mined. We could not identify an efficient method to find the missing information for these measurements and chose to exclude them because of the overall size of the dataset would prevent the exclusion of these measurements from introducing significant error to the analysis. If information on material being mined was missing in a non-random fashion, this could have introduced bias into the estimates presented here. We also removed 25,339 measurements for being below 60 dBA. Removing these measurements likely resulted in slightly higher mean exposures in our analysis, but we believe this is justified because it is very unlikely that an eight-hour TWA at a mine site would be < 60 dBA (the noise level of an average conversation). We believe that the effects of these excluded measurements on our analyses are likely small due to the size of the dataset.

Despite these limitations, the analysis herein signifies a substantial expansion of the previous work by Joy and Middendorf (2007), and provides an up-to-date examination of noise exposure in the mining industry. The main strength of this analysis is the size and scope of this dataset makes it possible to calculate very accurate group exposure estimates. Another strength of this analysis is that the use of the SOC system provides a standardized method for future studies to classify job titles in a harmonized manner so that exposure information can be more easily compared between studies. Additionally, by stratifying exposure groups by the type of mine and what mineral is being extracted, it is possible to discern exposure differences between different mining sectors and mine types so that sector- and mine-specific controls can be implemented to reduce noise exposure. Finally, the models presented here can be used to predict a worker’s mean yearly noise exposure based on their job title, type of mine they were employed at, and the year of their employment. This information could help identify workers at increased risk of developing NIHL and help prioritize resources to implement engineering controls and ensure that the worker is enrolled in a HCP.

Acknowledgments

List of acronyms and abbreviations

Appendix 1. Comparison of permissible exposure limit (PEL) measurements before and after 2000 by job title

Appendix 2. Regression coefficients for the permissible exposure limit (PEL) and action level (AL) models

Footnotes

References