Koh Dong-Hee, Kim Tae-Woo, Jang Seung Hee, Ryu Hyang-Woo, Park Donguk
1. Occupational Safety and Health Research Institute, Korea Occupational Safety and Health Agency, 400, Jongga-ro, Jung-gu, Ulsan 681-230, Korea 3.Present address: Department of Occupational and Environmental Medicine, International St. Mary's Hospital, Catholic Kwandong University, 25, Simgok-ro 100 beon-gil, Seo-gu, Incheon 404-834, Korea
1. Occupational Safety and Health Research Institute, Korea Occupational Safety and Health Agency, 400, Jongga-ro, Jung-gu, Ulsan 681-230, Korea.
Ann Occup Hyg. 2015 Aug;59(7):853-61. doi: 10.1093/annhyg/mev033. Epub 2015 May 2.
The purpose of this study was to evaluate grouping schemes for exposure to total dust in cement industry workers using non-repeated measurement data.
In total, 2370 total dust measurements taken from nine Portland cement factories in 1995-2009 were analyzed. Various grouping schemes were generated based on work process, job, factory, or average exposure. To characterize variance components of each grouping scheme, we developed mixed-effects models with a B-spline time trend incorporated as fixed effects and a grouping variable incorporated as a random effect. Using the estimated variance components, elasticity was calculated. To compare the prediction performances of different grouping schemes, 10-fold cross-validation tests were conducted, and root mean squared errors and pooled correlation coefficients were calculated for each grouping scheme.
The five exposure groups created a posteriori by ranking job and factory combinations according to average dust exposure showed the best prediction performance and highest elasticity among various grouping schemes.
Our findings suggest a grouping method based on ranking of job, and factory combinations would be the optimal choice in this population. Our grouping method may aid exposure assessment efforts in similar occupational settings, minimizing the misclassification of exposures.
本研究旨在利用非重复测量数据评估水泥行业工人总粉尘暴露的分组方案。
分析了1995年至2009年从9家波特兰水泥厂采集的总共2370次总粉尘测量数据。基于工作流程、工作岗位、工厂或平均暴露量生成了各种分组方案。为了表征每个分组方案的方差成分，我们开发了混合效应模型，将B样条时间趋势作为固定效应纳入，将分组变量作为随机效应纳入。利用估计的方差成分计算弹性。为了比较不同分组方案的预测性能，进行了10折交叉验证测试，并计算了每个分组方案的均方根误差和合并相关系数。
根据平均粉尘暴露量对工作岗位和工厂组合进行排序后事后创建的五个暴露组在各种分组方案中显示出最佳的预测性能和最高的弹性。
我们的研究结果表明，基于工作岗位和工厂组合排名的分组方法是该人群的最佳选择。我们的分组方法可能有助于类似职业环境中的暴露评估工作，最大限度地减少暴露的错误分类。
