Employment is an important marker of functional recovery from injury. There are few population-based studies of long-term employment outcomes, and limited data on the patterns of return to work post injury. This study sought to characterise patterns of engagement in work over the four-year period following major traumatic injury, and to identify factors associated with those patterns. We conducted a population-based, prospective cohort study using the Victorian State Trauma Registry. A total of 1086 working age individuals, in paid employment or full-time education before injury, were followed-up through telephone interview at 6, 12, 24, 36, and 48 months post-injury. Responses to return to work (RTW) questions were used to define four discrete patterns: early and sustained; delayed; failed; no RTW. Predictors of RTW patterns were assessed using multivariate multinomial logistic regression. Slightly more than half of respondents (51.6%) recorded early sustained RTW. A further 15.5% had delayed and 13.3% failed RTW. One in five (19.7%) did not RTW. Compared with early sustained RTW, predictors of delayed and no RTW included being in a manual occupation and injury in a motor vehicle accident. Older age and receiving compensation predicted both failed and no RTW patterns. Pre-injury disability was an additional predictor of failed RTW. Presence of co-morbidity was an additional predictor of no RTW. A range of personal, occupational, injury, health and compensation system factors influence RTW patterns after serious injury. Early identification of people at risk for delayed, failed or no RTW is needed so that targeted interventions can be delivered.