Reliability of a semi-quantitative method for dermal exposure assessment (DREAM)

Abstract

Valid and reliable semi-quantitative dermal exposure assessment methods for epidemiological research and for occupational hygiene practice, applicable for different chemical agents, are practically nonexistent. The aim of this study was to assess the reliability of a recently developed semi-quantitative dermal exposure assessment method (DREAM) by (i) studying inter-observer agreement, (ii) assessing the effect of individual observers on dermal exposure estimates for different tasks, and (iii) comparing inter-observer agreement for ranking of body parts according to their exposure level. Four studies were performed in which a total of 29 observers (mainly occupational hygienists) were asked to fill in DREAM while performing side-by-side observations for different tasks, comprising dermal exposures to liquids, solids, and vapors. Intra-class correlation coefficients ranged from 0.68 to 0.87 for total dermal exposure estimates, indicating good to excellent inter-observer agreement. The effects of individual observers on task estimates were estimated using a linear mixed effect model with logged DREAM estimates as explanatory variable; "task", "company/department", and the interaction of "task" and "company/department" as fixed effects; and "observer" as a random effect. Geometric mean (GM) dermal exposure estimates for different tasks were estimated by taking the exponent of the predicted betas for the tasks. By taking the exponent of the predicted observer's intercept (exp(omega i)), a multiplier (M(O)) was estimated for each observer. The effects of individual observers on task estimates were relatively small, as the maximum predicted mean observers' multiplier was only a factor 2, while predicted GMs of dermal exposure estimates for tasks ranged from 0 to 1226, and none of the predicted individual observers' multipliers differed significantly from 1 (t-test alpha = 0.05). Inter-observer agreement for ranking of dermal exposure of nine body parts was moderate to good, as median values of Spearman correlation coefficients for pairs of observers ranged from 0.29 to 0.93. DREAM provides reproducible results for a broad range of tasks with dermal exposures to liquids, solids, as well as vapors. DREAM appears to offer a useful advance for estimations of dermal exposure both for epidemiological research and for occupational hygiene practice.