Objectives Evening and night work have been associated with higher risk of injury than day work. However, previous findings may be affected by recall bias and unmeasured confounding from differences between day, evening and night workers. This study investigates whether evening and night work during the past week increases risk of injury when reducing recall bias and unmeasured confounding. Methods We linked daily working hours at the individual level of 69 200 employees (167 726 person years from 2008-2015), primarily working at hospitals to registry information on 11 834 injuries leading to emergency room visits or death. Analyses were conducted with Poisson regression models in the full population including permanent day, evening and night workers, and in two sub-populations of evening and night workers, with both day and evening or night work, respectively. Thus, the exchangeability between exposure and reference group was improved in the two sub-populations. Results Risk of injury was higher after a week with evening work [incidence rate ratio (IRR) 1.32, 95% confidence interval (CI) 1.26-1.37] and night work (IRR 1.33, 95% CI 1.25-1.41) compared with only day work. Similar, although attenuated, estimates were found for evening work among evening workers (IRR 1.18, 95% CI 1.12-1.25), and for night work among night workers (IRR 1.10, 95% CI 1.01-1.20). Conclusion There is an overall increased risk of injury after a week that has included evening or night work compared with only day work. Though attenuated, the higher risk remains after reducing unmeasured confounding.