Becoming aware at the point of care: a context-aware architecture for bridging semantic representation and interpretation of inter-organizational patient information

Background Fragmented patient records across healthcare organizations limit clinicians’ awareness of relevant patient information, while increased information availability may complicate identifying and interpreting information relevant to the current clinical situation. Although interoperability standards enable data exchange between systems, they do not necessarily create awareness of, identify, or help interpret information relevant to the current clinical situation. This highlights a gap between semantic representation and clinicians’ interpretation and use of information in practice. This study investigates how context-aware information access and contextualization can help bridge this gap. Methods Guided by Design Science Research, we developed the Context Resolver architecture as a detailed architectural specification of a key component within a previously established information-centric architecture framework. The architecture combines vector-based semantic retrieval, lightweight information tokens, provenance-preserving mechanisms, and Retrieval-Augmented Generation (RAG) to support context-aware information access. A prototype was demonstrated through a multimorbid patient scenario using synthetic clinical data. Two ex-ante and one ex-post formative evaluations involving clinical and technical experts assessed problem relevance, architecture design, and the prototype’s representation of proposed architectural mechanisms. Results The Context Resolver architecture addressed requirements related to contextual cueing, provenance, adaptive summarization, and support for information interpretation. The architecture promotes awareness of relevant cross-organizational clinical activities before deeper information access. Vector-based retrieval provided mechanisms intended to support access to contextually relevant information, while bounded GenAI synthesis generated provenance-linked summaries designed to reduce information overload through contextual filtering and summarization. The evaluations suggested that the identified problem is clinically relevant and that the proposed architecture provides a plausible response aligned with stakeholder needs, governance requirements, and clinical information practice. Conclusions The Context Resolver illustrates an architectural approach that combines machine-readable information representation with mechanisms intended to support contextual interpretation and use of information in clinical practice. The study contributes a detailed architectural specification of the Context Resolver as socio-technical design knowledge for semantic interoperability in health information systems and illustrates how awareness, context-aware retrieval, provenance-preserving access, and responsible GenAI may support information access across organizational boundaries. The findings suggest that semantic interoperability can benefit from complementing machine-readable information exchange with mechanisms supporting awareness, contextualization, and interpretation in clinical practice.