(P 215) SHAPEHUB: A MULTICENTRE REAL-WORLD-TIME DATA PLATFORM WITH DIGITAL TWIN ENGINE FOR A LEARNING HEALTH SYSTEM IN SARCOMA

Objective: To describe the design and early deployment of SHAPEHUB, a federated real-world-time (RWT) platform supporting value-based, data-driven sarcoma care. The platform integrates digital twin representations, automated quality indicators, and causal machine learning pipelines as foundational tools of a learning health system.

Methods: SHAPEHUB connects eight institutions within a hub-and-spoke model: one central sarcoma expertise hub and multiple affiliated centers. Structured data from electronic health records, pathology, imaging, genomics, PROMs, and administrative modules are harmonised using HL7-FHIR and OMOP standards. Real-time ingestion enables synchronized multidisciplinary tumour board (MDT) dashboards, shared longitudinal views, and metadata queries. A digital twin engine mirrors each patient’s clinical journey and forms the substrate for causal counterfactual modelling. Automated algorithms for sarcoma-specific quality indicators (e.g. margin documentation, staging procedures, timely treatment) are embedded. Cost linkage pipelines are in development to enable real-time outcome-to-cost mapping.

Results: All centers are live with RWT data ingestion, dashboard functionality, and unified MDT participation. PROM capture and quality monitoring are operational; validation is ongoing. Digital twin timelines support exploratory queries and are being prepared for causal inference layers. Initial cost integration is active in one spoke center. The network operates in a fully federated architecture: institutional data custody is maintained while analytics occur across the shared environment.

Conclusion: SHAPEHUB establishes a scalable digital infrastructure for individualised, value-based sarcoma care. By combining real-time data flow, digital twins, and causal analytics within a hub-and-spoke model, it enables dynamic quality assessment and personalised decision support, positioning SHAPEHUB’s hub-and-spoke model as a blueprint for scalable learning health systems.