Discord and social context in exposure assessment methods: Self-reports and expert ratings from the United for Health Study

Often in exposure assessment for epidemiology, there are no highly accurate exposure data and different measurement methods are considered. The objective of this study was to explore agreement between sociodemographic influences on individual reports and expert ratings of workplace exposures. To do this, we administered questionnaires to a multiracial/multiethnic cohort of 1282 low-income employees at 4 industries (14 worksites), capturing information on specific job exposures to noise, dust, airborne chemicals, job strain components, and musculoskeletal stressors from awkward postures occurring within the past 12 months. Study team occupational hygienists constructed industry- and site-specific job exposure matrices (JEMs) based on worksite walkthrough exposure evaluations. These worksite assessments used a checklist with questions and rating scales similar to those on the employee questionnaires. Worker self-reported exposures were compared with the worksite specific JEMs using percent agreement, kappa scores, loglinear and multilevel statistical models. There was generally poor agreement between the self-reported and expert exposure assessments and a negative association (discord) between individual reports and expert ratings. This discord was significantly influenced by sociodemographic characteristics: workers who were female, and those who preferred the Spanish language survey, tended to report higher exposures than experts more frequently than male workers and those who took the survey in English; younger, foreign-born, and non-white workers tended to under-report high exposures. These discord analyses highlight the social patterning of workplace exposures and advance discussions that often attribute poor agreement between measures to misclassification of exposure. Social context should be an important part of any exposure assessment strategy.