Assessment of occupational exposure patterns by frequency-domain analysis of time series data

Laboratory evidence increasingly points to exposure pattern characteristics, including the duration, frequency, and timing of the exposure during the day, as important factors influencing the biological response to extremely low-frequency magnetic fields. An exploratory analysis of exposure patterns was conducted in 113 electric utility workers employed as electricians, cable splicers, line workers, and power plant operators. The purpose of the study was to describe extremely low-frequency magnetic field exposure pattern characteristics of electric utility workers and evaluate grouping strategies for classifying occupational exposures based on their exposure pattern characteristics. Exposure patterns describe the cyclic fluctuation in exposures over time, and were evaluated by partitioning the variation of the time series into frequency components using frequency-domain analysis of the transformed and processed time series. The study samples were classified using traditional grouping strategies based on occupation and time-weighted average (TWA), and non-traditional grouping strategies based on cluster analysis of the standardized, low-frequency exposure pattern components. Rules for classifying samples into each group were developed using linear discriminant analysis, with the performance of each grouping strategy evaluated using a crossvalidation study design to estimate the rate of misclassification. Exposure patterns appeared unrelated to grouping strategies based on quartiles of the workday TWA, but were related to pattern clusters and occupation. The linear discriminant function produced very low misclassification error rates for the cluster grouping strategy (10%) compared to occupation (50%) and TWA quartile (69%) grouping strategies. Significant differences in the exposure patterns occurring between clusters and between occupational groups were observed, indicating that at least one of the spectral estimates in two of the groups were significantly different. However, patterns clusters produced the greatest contrast in exposure patterns of all grouping strategies, explaining 99 percent of the total variation compared to 58 percent of the total variation by occupation.