Impact of expert versus measurement-based occupational noise exposure estimates on exposure-response relationships

Abstract

Expert-judgment has frequently been used to assess quantitative exposure for epidemiologic studies, but accuracy varies widely dependent on the type of exposure and the availability of measurements to anchor estimates. There is limited empirical evidence of the sensitivity of exposure-response relationships to expert- versus measurement-based exposure assessment strategies. We examined the sensitivity of the exposure-response relationship between occupational noise exposure and acute myocardial infarction (AMI) mortality using both expert- and measurement-based occupational noise estimates in a retrospective cohort study of sawmill workers (n = 27,499). Expert-based noise estimates were evaluated by four industry experts who rated 54 sawmill jobs on a four-point scale. Measurement-based noise estimates were derived from statistical models that accounted for job, mill, and time period differences. The model-based estimates were adjusted to account for the use of hearing protective devices (HPD). We examined the shape, goodness of fit, precision, and expected versus observed attenuation of the exposure-response relationships between cumulative noise exposure and AMI mortality (910 deaths). The correlations between the expert-based and the measurement-based unadjusted and HPD-adjusted cumulative noise estimates were 0.81 and 0.57, respectively. The HPD-adjusted model-based estimates provided the most precise exposure-response relationship; no associations were observed with the unadjusted or expert-based noise estimates. In a subgroup with minimal HPD use (n = 8,700, 520 deaths), the expert- and model-based noise estimates resulted in similar relative risks; the model-based approach was 12% more precise. The measurement-based approach was more precise, as expected, but experts were reasonably able to rank occupational noise exposures. The experts' assessment was, however, unable to account for HPD use, which made a substantial contribution to exposure misclassification in this study. The experts' noise estimates would be more useful for risk assessment if they were calibrated against units of noise exposure.