Mixed-effects models for the evaluation of long-term trends in exposure levels with an example from the nickel industry.

Longitudinal studies play an important role in evaluating the temporal behavior of occupational exposures. The purpose of this paper is to examine certain features of longitudinal data and to present a general conceptual framework by which these features may be taken into account so that statistically valid inferences can be made. Statistical methods that rely on the application of mixed-effects models are proposed for evaluating long-term trends in exposures to workplace contaminants. The mixed-effects model presented herein has fixed effects for trend components and random effects for workers, job groups, buildings and plants. These models differ from conventional techniques in that they accommodate hierarchically structured data and account for the correlation that may arise due to the clustering of measurements based on when and where the data were collected. While primary interest is focused on determining the magnitude of trends in exposure levels over time, the model also provides information about the magnitude of the sources of variation associated with different groupings of workers. Application of the mixed-effects model is illustrated with a large database of shift-long personal exposure measurements collected on workers exposed to nickel aerosols in the nickel-producing industry.