358 Understanding factors related to between- and within-Subject variation in an effort to better define SEGs
通过分析兽医学生在解剖实验室中的甲醛暴露数据，研究如何利用混合效应模型识别暴露决定因素和变异来源，以改进相似暴露组的划分，减少暴露错分。
<h3>Objectives</h3> One approach for characterising exposures of an occupational cohort when measurement data are limited is to divide the population into Similar Exposure Groups (SEGs), or clusters of workers believed to have the same general exposure profile for the agent (s) under study, from which individual levels are established. However, when the assumed homogeneity of an SEG is incorrect, researchers risk misclassifying exposures. Through analyses of formaldehyde exposures of veterinary students enrolled in a gross anatomy laboratory course, this research examines methods for improved understanding of variability sources within a dataset in an effort to better define SEGs. <h3>Methods</h3> Initial analyses suggested classifying this cohort as one single SEG may be questionable and demonstrated the importance of an appropriate sampling strategy. A mixed-effects model was thus used to identify exposure determinants and assess sources of variation. Using formaldehyde exposure as the dependent variable, explanatory variables were partitioned into fixed effects (animal, animal part, lab location, sample collection date) and random effects (subject). Additional analyses were run separately for each animal type in an effort to examine variability by task. <h3>Results</h3> Examination of the data identified several potential sources of variability. The model indicated that animal and animal part may have a significant effect on exposure, with a within-subject to between-subject variance ratio of 2:1. The proportion of total variability attributable to within-subject variation differed by animal type, with 46.4, 98.6, and 70.2% associated with dog, goat, and horse dissections, respectively. <h3>Conclusions</h3> These results help identify and describe work characteristics influencing exposure levels within the cohort. Understanding factors related to between- and within-subject variability allows for refined sub-grouping of the population and identification of work conditions that influence day-to-day exposure variations. With on-going analyses, this work will attempt to create more informative SEGs as a way to reduce exposure misclassification.