A comparison of task-based and job-based estimation of physical behavior compositions in grocery store workers

Background Job exposure matrices (JEMs) are widely used in occupational epidemiology to estimate biomechanical exposures in a trade. However, JEMs are insensitive to variability among workers within the same job, potentially leading to less accurate exposures and increased uncertainty of exposure-outcome associations. Task-based exposure assessment, in which individual exposures are derived from task-specific exposures weighted by the time spent on each task, has been suggested to improve accuracy. Objective To examine and compare the accuracy of task-based and job-based assessment of physical behaviors—sitting, standing, and moving—in grocery retail. Methods The study was conducted in 2 Swedish medium-sized grocery stores. Accelerometry was used to measure “true” physical behaviors continuously over 3 full working days for all participating employees (n = 36; 16 women, 20 men). Job-based exposure estimates were derived by averaging sitting, standing, and moving across all workers, representing a JEM. Nine discrete tasks were defined in collaboration with employees and owners in the stores. For 35 of the workers, task timelines were videotaped for 4 h and linked to accelerometry to construct a gender-neutral (unisex) task exposure matrix (TEM). All 36 participants completed diaries for the 3 d, reporting time spent on each task, and task-based estimates of “true” exposures through the 3 d were calculated for each worker as the weighted average of task exposures based on the diary-reported task distribution. Physical behaviors were analyzed using procedures from compositional data analysis (CoDA), including describing and comparing the performance of job-based and task-based approaches using Aitchison distances to the measured, “true” exposures. Results The prerequisites for successful task-based assessment were satisfied, including physical behavior contrasts between tasks and heterogeneity in task composition between workers. task-based estimates were on average closer to “true” exposures than job-based estimates, but the magnitude of improvement was modest (23%) and varied between women (35%) and men (12%). Thirty-three percent of the 36 “true” job exposures were better predicted by the job-based approach, and substantial variability in exposure persisted even within tasks. Conclusion Task-based assessment offered only modest improvement in accuracy over job-based estimates and the cost-efficiency of the task-based approach compared to the less costly job-based strategy can be questioned. Task-based assessment may therefore be most useful when task information is needed for documentation or interventions, rather than solely for improving exposure estimation.