Purpose This study investigated the association between return to work self-efficacy (RTW-SE) and sustained return to work (RTW) at two different time points, over a 12-month period. The primary objective of the study was to examine if the relationship between RTW-SE and a sustained RTW changed over the RTW timeline. Methods This study used survey responses from a longitudinal cohort of n = 410 workers’ compensation claimants with either an upper-body musculoskeletal injury or a psychological injury. A path analysis tested the associations between RTW-SE and a sustained RTW at two time-points. A Wald χ2 test compared nested models to determine if the association changed over time. Results RTW-SE measured at time- point 1 (T1) was associated with a sustained RTW at time-point two (T2) (β = 0.24, P < 0.05) but no association was found between RTW-SE at T2 and a sustained RTW at time-point three (T3) (β = 0.017, n.s.). Model comparisons revealed significant differences in the associations between RTW-SE and a sustained RTW, with the relationship being stronger in the early phase of RTW compared to the latter phase (χ2 = 5.002, p = 0.03). Conclusions The results indicate that RTW-SE at 4–6 months post-injury is important for a sustained RTW 6-months later although RTW-SE at 10–12 months post-injury had a negligible association over the same duration. Further research should investigate whether these findings generalize to other populations and what factors other than RTW-SE are associated with RTW in the later stages of the RTW process.