Background.Job-based exposure estimation using the occupational mean (JBM) is associated with substantial error. Many studies have therefore estimated job exposures from workers’ tasks, i.e. task-based modeling (TBM), typically by combining individual workers’ task proportions (TP) in the job with a general task exposure matrix. Studies of postures and muscle activity have, however, shown that TBM may be ineffective; one possible reason being that TPs are not correct. The present simulation study investigated the influence of random and systematic TP error on TBM performance.

Methods.We constructed two virtual two-task jobs with task exposure contrasts of 0.2 and 0.8. In both, TPs and task exposures mimicked likely occupational scenarios. We then simulated four cases of TP error: no error, random error, bias, and bias and random error. For each case, we varied the TP error size, and compared the absolute errors of TBM- and JBM-based job exposures for 10,000 virtual workers.

Results.For the low-contrast job, TBM with error-free TPs was, on average, only 6% more efficient than JBM, and the probability of TBM leading to a more correct job exposure than JBM was 56%. TP errors had negligible effects on effectiveness. With error-free TPs in the high-contrast job, TPM was 75% more efficient than JBM, and led to more correct job exposures for 71% of all workers. TP errors decreased TBM performance, down to being 34% better than JBM when both random and systematic errors were “large”; 62% of all individuals being more correctly assessed by TBM.

Discussion.For jobs with limited task exposure contrast, TBM was essentially equivalent to JBM, while TP errors had marginal impact. In high-contrast jobs, TBM was more effec-tive, but was also more sensitive to both random and systematic TP errors. This may feed further discussion of the cost-efficiency of TBM in occupational settings.