Objective To test whether return to work as a binary (yes/no) outcome that includes all persons who returned to work regardless of mode of return reflects a composite or a homogeneous outcome in a cohort of workers who have sustained acute orthopaedic trauma resulting in hospitalisation.

Methods Prospective cohort study. One hundred and sixty-eight participants were recruited and followed for 6 months. The study achieved 89% follow-up. Baseline data were obtained at study recruitment and participants were further surveyed by phone at three timepoints during the study. Polytomous logistic regression was used to simultaneously examine the association between potential predictors and different modes of first return to work (RTW). A test of the equality of the ORs associated with the independent predictor variables was also undertaken.

Results Of the 152 participants with full follow-up, 46 (30%) returned first to full duties, 58 (38%) returned first to modified work and 48 (32%) did not return to work during the study period. Significant determinants of the two modes of return to work were different. A test of the equality of ORs indicated that the relative ORs for the difference in the slope coefficients for five of the 10 independent factors in the two polytomous logistic regression sub-models corresponding to each mode of return to work were statistically significant. This raises the likelihood that first RTW reflects a composite rather than a homogeneous outcome.

Conclusion The study provides evidence that RTW may reflect a composite outcome when it includes different modes of first RTW. The identified predictive factors appear to exert different mechanisms of action depending on the mode of RTW. The findings suggest that the different modes of RTW may need to be considered independently. The results of the study have potentially important implications for research and insurance practice.