0165 A job exposure matrix from self-reported physical exposures: a method to control bias from symptoms

Background Self-reported physical exposures are prone to bias from symptoms. Job Exposure Matrices (JEM) can give unbiased estimates but do not capture variation of exposures within the same job. This study created JEMs from self-reported exposures, both including and excluding responses from symptomatic workers, and compared these three exposure measures and their associations with hand/wrist pain. Methods 35 000 employed French workers from the CONSTANCE cohort reported work-related physical exposures and current symptoms in the hand and other body areas. We grouped self-reported exposures at the level of the job title to create two JEMs - a full JEM including all workers, and an asymptomatic JEM excluding workers with pain symptoms. We examined differences in exposure estimates between the full and asymptomatic JEMs and the change of within-group and between-group variance. Log-binomial models tested strength of association between individual, full and asymptomatic JEMs to hand pain. Results Most job-level exposure estimates were lower in the asymptomatic JEM. There was smaller within-group variance and larger between group variance in the asymptomatic JEM, demonstrating more homogenous exposure groups. Preliminary regression models showed similar estimates of association between each exposure method (individual, full JEM and asymptomatic JEM) and hand pain (Prevalence ratios with asymptomatic JEM=1.15–1.34; all p Conclusions A JEM using responses only from asymptomatic workers created more homogenous exposure groups, but initial analyses showed no other significant evidence of biassed exposure estimates due to symptoms. JEMs are a useful method of exposure assignment for some epidemiological studies of musculoskeletal disorders.