Designing a secure rare disease data-sharing platform leveraging electronic health records: a system architecture and simulation study

The persistent fragmentation of rare disease data significantly hinders medical research and orphan drug development. This study aimed to design and technically validate a secure data-sharing platform capable of dismantling institutional data silos by activating underutilized electronic health records (EHRs). A multi-tier platform architecture was developed to integrate EHR data, incorporating the domain-adapted Rare-RoBERTa model and ORPHA-codes for clinical semantic standardization. To assess technical feasibility, a prototype was implemented on the Hyperledger Fabric blockchain to facilitate zero-trust, audited data exchange. Platform performance was evaluated using the Chinese Biomedical Language Understanding Evaluation (CBLUE) benchmark and synthetic clinical records, focusing on entity extraction accuracy, blockchain throughput, latency, and dynamic privacy-preserving de-identification. In simulated data governance evaluations, Rare-RoBERTa achieved an overall F1-score of 87.3% in named entity recognition (NER), outperforming general baseline models. The simulated blockchain network reached a peak write throughput of approximately 980 transactions per second (TPS) under 600 concurrent users, while the de-identification mechanism preserved research data utility with an information loss of 14.5% and a re-identification risk below 0.08%. These metrics represent simulated technical performance in a controlled environment and should not be interpreted as evidence of clinical effectiveness or nationwide scalability. The proposed architecture provides a preliminary, simulated proof-of-concept demonstrating the technical feasibility of rare disease data-sharing infrastructures. While these laboratory simulation metrics do not establish clinical effectiveness or operational scalability in live hospital environments, this study outlines a plausible architectural foundation to reactivate dormant clinical data and potentially support future health equity. Rigorous multi-center pilot trials using real-world heterogeneous EHRs remain essential prior to real-world clinical deployment, ultimately facilitating evidence-based policymaking and rare disease governance.