Project title: Improving Estimates of Occupational Noise Exposure
Completed in: 2009
Exposure assessment in occupational epidemiological studies is difficult, and particularly challenging in dynamic industries like construction. To inform the exposure assessment strategy of a longitudinal study of noise-induced hearing loss (NIHL) among construction workers, this study evaluated the accuracy and performance of various assessment techniques. These techniques included trade-mean (TM) and task-based (TB) approaches, a worker-based subjective rating (SR) technique, and novel hybrid combinations of these techniques.
The performance of the techniques was evaluated in several settings. Short-term (single shift and two-week) TM and SR estimates were created for subjects in three different noise environments: continuous, intermittent, and highly variable (n=20 subjects per environment). Exposure estimates were then compared to subjects’ measured average exposures. The TM and SR techniques were validated in a separate cohort of construction workers (n=68 workers) over a longer four-month period. In addition to TM and SR estimates, TB estimates were also created for this cohort, and estimates from each of these three techniques were compared to subjects’ measured average exposures. Finally, hybrid techniques were developed for the cohort, which incorporated TM, SR, and TB estimates via arithmetic mean combination, linear regression combination, and modification of TM and TB estimates using SR information. Hybrid estimates were compared to subjects’ measured exposures and to estimates from single techniques.
TM and TB estimates exhibited a large amount of measurement error compared to measured exposures, with the TM estimates being the less accurate of the two. SR estimates generally demonstrated less error, and provided greater exposure contrast, than did TM estimates, but included more error than TB estimates. SR items also correctly evaluated different noise environments and exposure variability. Hybrid estimates generally had less measurement error than estimates from single techniques. The best-performing hybrid techniques combined TB and SR estimates, and resulted in substantial improvements in estimated exposures compared to single techniques. Hybrid estimates were not improved by the inclusion of TM information. The strongest and most precise exposure-response estimates in the longitudinal cohort of construction workers being evaluated for NIHL are expected to result from use of a hybrid regression technique combining TB and SR estimates.