A comparison of approaches to modeling the relationship between ergonomic exposures and upper extremity disorders

Background: For a study of upper extremity musculoskeletal disorders among automobile manufacturing workers, an ergonomic exposure index was constructed by summing ten psychophysical (interview) items. Here we explore the sensitivity of the exposure-response relationship to the formulation of that index. Methods: Five alternative exposure indices were constructed: three a priori weighting schemes and two sets of weights derived from multivariate regression coefficients. In addition, structural equation modeling was performed with LISREL. Results: The original index and a priori weighting schemes had similar associations with upper extremity disorders, adjusted for nonoccupational covariates. A reasonable model fit was achieved in LISREL after two modifications; the standardized solution showed that nonneutral postures were significantly related to upper extremity signs and symptoms. Conclusions: In this large population, with adequate range of exposures, the exposure-response relationship appeared generally robust to the mathematical formulation of the exposure index. Among the available exposure variables, postural strain had the strongest association. (C) 2000 Wiley-Liss, Inc.