| Abstract
| - Multi-tiered sampling approaches are common inenvironmental and occupational exposure assessment,where exposures for a given individual are often modeledbased on simultaneous measurements taken at multipleindoor and outdoor sites. The monitoring data from suchstudies is hierarchical by design, imposing a complexcovariance structure that must be accounted for in orderto obtain unbiased estimates of exposure. Statisticalmethods such as structural equation modeling (SEM)represent a useful alternative to simple linear regressionin these cases, providing simultaneous and unbiasedpredictions of each level of exposure based on a set ofcovariates specific to the exposure setting. We test the SEMapproach using data from a large exposure assessmentof diesel and combustion particles in the U.S. trucking industry.The exposure assessment includes data from 36 differenttrucking terminals across the United States sampledbetween 2001 and 2005, measuring PM2.5 and its elementalcarbon (EC), organic carbon (OC) components, by personalmonitoring, and sampling at two indoor work locationsand an outdoor “background” location. Using the SEMmethod, we predict the following: (1) personal exposuresas a function of work-related exposure and smokingstatus; (2) work-related exposure as a function of terminalcharacteristics, indoor ventilation, job location, andbackground exposure conditions; and (3) backgroundexposure conditions as a function of weather, nearby sourcepollution, and other regional differences across terminalsites. The primary advantage of SEMs in this setting is theability to simultaneously predict exposures at each ofthe sampling locations, while accounting for the complexcovariance structure among the measurements anddescriptive variables. The statistically significant resultsand high R 2 values observed from the trucking industryapplication supports the broader use of this approach inexposure assessment modeling.
|
