Background Direct/ measurement of work activities iscostly, so researchers need to distribute resourcesefficiently to elucidate the relationships betweenexposures and back injury.

Methods This study used data from full-shiftelectromyography (EMG; N¼133) to develop threeexposure metrics: mean, 90th percentile and cumulativeEMG. For each metric, the components of variance werecalculated between- and within-subject, and betweengroupfor four different grouping schemes: grouping byindustry (construction, forestry, transportation,warehousing and wood products), by company, by job andby quintiles based on exposures ranked by jobs withinindustries. Attenuation and precision of simulatedexposureeresponse relationships were calculated for eachgrouping scheme to determine efficient sampling strategies.

Results As expected, grouping based on exposurequintiles had the highest between-group variances andlowest attenuation, demonstrating the lowest possibleattenuation with this data.

Conclusion There is potential for grouping schemes toreduce attenuation, but precision losses should beconsidered and whenever possible empirical data shouldbe employed to select potential exposure groupingschemes.