Objectives Individual response style, mood, expectations, and health status may affect reporting of the psychosocial work environment, and bias associations with outcomes. Reporting bias may be avoided by aggregating individual responses, ideally preserving exposure contrast. In this study, we examined the degree of exposure contrast yielded by different grouping strategies.

Method In 2007, we enrolled 4489 public employees from Aarhus, Denmark in the PRISME-cohort, with follow-up in 2009. From pay-roll registers we grouped workers at 2 organisational levels: department (n = 22) and work unit (n = 751), and 3 occupational levels: sector (n = 7), profession (n = 46), and job title (n = 77). Exposures, calculated as means of items scored on 5-point Likert scales, included psychological demands, decision latitude, social support, effort, reward, and procedural and relational justice. To assess variance components, we fitted linear mixed effect models with exposures as dependent variables, and id and grouping variables as random effects. Results are reported as the contrast in mean exposure levels e.g. between-group variance/ (between-group variance +within-group variance).

Results Within each hierarchy contrasts rose with increasing group-level detail. Grouping by either work unit (wu) or by job title (jt) contrasts were: psychological demands: 0.28(wu); 0.26(jt), decision latitude: 0.24(wu); 0.32(jt), social support: 0.24(wu); 0.06(jt), effort: 0.23(wu); 0.16(jt), reward: 0.19(wu); 0.12(jt), procedural justice: 0.24(wu); 0.14(jt), and relational justice: 0.29(wu); 0.04(jt).

Conclusions Grouping by work unit gave the most consistent contrasts (0.19–0.29), while grouping by job title varied considerably (0.04–0.32). These preliminary findings suggest that grouping by work unit provided better exposure contrasts than grouping by job title for all exposures, but decision latitude.