Meta-analysis of job-exposure matrix data from multiple sources

Abstract

The objectives of this study were to determine the heterogeneity of data sources used to construct a job-exposure matrix (JEM) for occupational noise, and to calculate pooled exposure estimates for different job titles using different sources. The JEM was populated with measurements from government databases, private industry, and the published literature. Data were organized by job title using the US standard occupational classification (SOC). Using data from the literature as prior information, adjusted mean exposure was calculated for both the government and industry data following a simple Bayesian approach. A meta-analysis was conducted to measure data heterogeneity across sources and to calculate a pooled exposure estimate for each SOC and SOC group. In total, 715,867 measurements across 259 SOCs were analyzed. Using the data from literature as a prior, 15 of 28 applicable SOCs in the government and industry data had adjusted mean exposures above the OSHA action level (85 dBA). The meta-analysis showed that 63% of SOCs, and 78% of SOC groups, had moderate to high heterogeneity. Fifty-one percent of SOCs and 43% of SOC groups had pooled estimated exposures >85 dBA. The pooled estimates suggested that workers in 131 of 259 SOCs (51%) were exposed beyond the threshold of 85 dBA.

Figure 1.

Relationship of SOC codes between the three sources of data. L,G, and I correspond to literature, government, and industry data respectively.

Figure 2.

Adjusted means and 95% CI for government data using the literature data as the prior.

Figure 3.

Adjusted means and 95% CI for industry data using the literature data as the prior.

Figure 4.

Distribution of noise measurements from 1968- 2014. The lower bound of the box represents the 25^th percentile, the line represents the 50^th percentile and the upper bound of the box represents the 75^th percentile.