Mary E Davis1. 1. Department of Urban and Environmental Policy and Planning, Tufts University, 97 Talbot Avenue, Medford, MA 02155, USA. mary.davis@tufts.edu
Abstract
OBJECTIVES: Many occupational hygiene surveys are designed to collect pollutant monitoring data from multiple locations simultaneously to better reflect the reality of work-related exposure. The exposure model must account for the complexity inherent in this study design, as well as be flexible to extrapolating exposures across an occupational cohort for dose-response modelling and risk assessment. This paper explores the structural equation model (SEM) as a tool to analyse pollutant monitoring data from occupational studies with multiple concurrent sampling across exposure locations. METHODS: This study uses exposure data from a comprehensive assessment of diesel exhaust in the US trucking industry to test the strength of SEMs over more standard analytical approaches such as ordinary least squares (OLS). The exposure data consist of concurrent sampling of elemental carbon from multiple co-located monitors on individual workers, work area and background levels at 36 different trucking terminals across the USA. RESULTS: The SEM is compared with two separate OLS specifications-one that focuses only on predicting personal exposure and excludes data from the additional monitoring sites, and a second that estimates three separate OLS specifications. When compared with the OLS specifications, the SEM provided a better fit to these layered exposure data. The OLS specifications suffered from bias in the coefficients, including downward bias in the work area and background exposure levels and overstatement of the smoking effect. Additionally, many theoretically valid covariates were significant only in the SEM. CONCLUSIONS: This study provides evidence in favour of more widespread use of SEMs in occupational health. SEMs represent a more robust and realistic framework for modelling multiple exposure pathways and have the potential to reduce exposure misclassification bias and strengthen the linkages between studies of exposure and disease outcomes.
OBJECTIVES: Many occupational hygiene surveys are designed to collect pollutant monitoring data from multiple locations simultaneously to better reflect the reality of work-related exposure. The exposure model must account for the complexity inherent in this study design, as well as be flexible to extrapolating exposures across an occupational cohort for dose-response modelling and risk assessment. This paper explores the structural equation model (SEM) as a tool to analyse pollutant monitoring data from occupational studies with multiple concurrent sampling across exposure locations. METHODS: This study uses exposure data from a comprehensive assessment of diesel exhaust in the US trucking industry to test the strength of SEMs over more standard analytical approaches such as ordinary least squares (OLS). The exposure data consist of concurrent sampling of elemental carbon from multiple co-located monitors on individual workers, work area and background levels at 36 different trucking terminals across the USA. RESULTS: The SEM is compared with two separate OLS specifications-one that focuses only on predicting personal exposure and excludes data from the additional monitoring sites, and a second that estimates three separate OLS specifications. When compared with the OLS specifications, the SEM provided a better fit to these layered exposure data. The OLS specifications suffered from bias in the coefficients, including downward bias in the work area and background exposure levels and overstatement of the smoking effect. Additionally, many theoretically valid covariates were significant only in the SEM. CONCLUSIONS: This study provides evidence in favour of more widespread use of SEMs in occupational health. SEMs represent a more robust and realistic framework for modelling multiple exposure pathways and have the potential to reduce exposure misclassification bias and strengthen the linkages between studies of exposure and disease outcomes.
Authors: Audra L Gollenberg; O Yaw Addo; Zhiwei Zhang; Mary L Hediger; John H Himes; Peter A Lee Journal: Horm Res Paediatr Date: 2015-01-24 Impact factor: 2.852
Authors: Robert B Gunier; Asa Bradman; Michael Jerrett; Donald R Smith; Kim G Harley; Christine Austin; Michelle Vedar; Manish Arora; Brenda Eskenazi Journal: Environ Sci Technol Date: 2013-09-23 Impact factor: 9.028
Authors: Scott Carver; Julia A Beatty; Ryan M Troyer; Rachel L Harris; Kathryn Stutzman-Rodriguez; Vanessa R Barrs; Cathy C Chan; Séverine Tasker; Michael R Lappin; Sue VandeWoude Journal: Parasit Vectors Date: 2015-12-23 Impact factor: 3.876