The Global Safety Knowledge Graph Linking Case Reports, Clinical Records, Scientific Literature, Regulatory Actions, and Product Changes

Medicine-safety evidence is distributed across spontaneous case reports, clinical records, registries, observational studies, scientific publications, regulatory assessments, product labels, and manufacturing or product histories. These sources differ in structure, terminology, provenance, temporal coverage, jurisdiction, and evidentiary meaning, limiting the reconstruction of how an emerging concern develops into a safety signal, causal assessment, regulatory response, or product change. This original knowledge-graph architecture article proposes a Global Safety Knowledge Graph that represents source records, normalized concepts, evidence claims, inferred relationships, assessment states, regulatory actions, and product versions as distinguishable but connected objects. The architecture combines an evidence-source layer, semantic and identity-resolution services, a temporal-provenance graph core, analytical and conflict-management functions, and governed review and feedback processes. Product versions, manufacturing changes, and exposure contexts are represented separately from active ingredients so that product-quality or traceability hypotheses are not misattributed to an entire substance or class. Conflicting evidence is preserved through support, challenge, supersession, and unresolved-status relations rather than being overwritten during updating. Temporal representation distinguishes when an event occurred, when evidence entered the system, when an assessment changed, and when a regulatory or product action became effective. Validation is proposed across syntactic, semantic, identity-resolution, provenance, temporal, analytical, transportability, workflow, and decision-impact levels. The architecture may improve evidence traceability and support testable safety hypotheses, but it does not convert association into causation, eliminate source bias, establish regulatory acceptability, or justify automated clinical or regulatory decision-making. Its principal contribution is a unified, evidence-bounded model for connecting global medicine-safety knowledge while preserving uncertainty, provenance, temporal history, and human responsibility.