The ability to collaboratively edit data in distributed environments is essential in our information-based industry. Typically users or systems concurrently access and modify data from different locations for different purposes. Each purpose might require its own representation and subset of the shared data (i.e., a model), for the editor to be productive. Consequently, a multi-model system results, which requires multi-directional synchronization. Although some approaches exist to realize such systems, non of these supports concurrent modifications. To overcome this limitation, we extend previous work on role-oriented model synchronization with a novel conflict detection and resolution approach. Role-oriented programming has been shown to be an adequate paradigm to realize multi-model systems, as it offers separation of concerns at the level of object collaborations and allows to express context-dependent behavior. We evaluate our approach using an employee data management case study and assess the introduced performance overhead.