Efficient planning of vehicles and personnel is essential for cost-effective railway operations. Traditionally, circulation planning for the traction units and shift planning for the traction unit drivers are performed sequentially, which limits flexibility in the latter. In this paper, we propose a semi-integrated approach where the circulation plan is computed first, but the shift planning stage retains the freedom to choose the exact positioning of empty runs from a set of predefined options. This added flexibility enables more efficient crew schedules without affecting the quality of the circulation plan. Circulation planning is modeled as a classical network flow problem, while shift planning follows a two-stage process: feasible shifts are generated using a constraint programming model and selected using a set partitioning model. A case study based on data from the Austrian Federal Railways demonstrates the practical applicability and its potential to reduce crew scheduling costs. On average across all instances, the efficiency gains amount to 3.7 %. These results show that semi-integrated methods offer a promising alternative to sequential planning while avoiding the complexity of fully integrated models.