Inter-organizational collaboration in networks is the preferred approach to address the complex challenges the healthcare sector faces, as well as to deliver high value health services. Research on inter-organizational networks has flourished but remains conceptually and empirically siloed. Using a sequential mixed-methods design, this study answers recent calls for multi-level studies of healthcare networks. We first quantitatively model system-oriented patient sharing networks using claims data of 48 thousand dementia patients in 31 regions in the Netherlands across three years. We subsequently conducted qualitative interviews in two of these regions to understand actors’ perception of their network structure and how they purposefully collaborate within these ties. Our quantitative results reveal low density and high centralization of networks, albeit with some variation, across the 31 networks in our sample within a given year. Over time, these structural features remain very stable however. Our qualitative results furthermore indicate that actors deliberately collaborate in various purpose-oriented networks including a small subset of actors and that important actors, such as social care providers and informal caregivers, are missing from the claims-based network. Networks represent a supra-organizational structure, transcending traditional organizational boundaries. However, our results suggest that paradoxically, delving into any network fails to yield a comprehensive understanding, despite the inherent purpose of conducting network analysis to attain such insights. Consequently, most network perspectives represent inherently incomplete perspectives, highlighting the intricate nature of interconnected systems and the challenges associated with capturing their entirety.