Clinical Knowledge Representation (CKR): A Formal Framework for Machine-Readable Clinical Knowledge

Contemporary digital health infrastructure excels at representing data (DICOM), clinical resources (FHIR), concepts (SNOMED CT), and biomedical relations (knowledge graphs). Yet a structural gap persists at the foundation of clinical artificial intelligence: the knowledge that clinicians derive from these artifacts — structured, evidenced, provenance-bearing, and indexed to individual patients — is never explicitly represented in a computationally tractable form. AI systems are consequently forced to reconstruct, at inference time, the knowledge that was never encoded. This reconstruction is the root cause of clinical AI brittleness, opacity, poor generalizability, and regulatory non-compliance. This article introduces Clinical Knowledge Representation (CKR) as a formal architectural layer in computational medicine. CKR is defined as a semantic computational infrastructure that explicitly represents clinical knowledge through entities, attributes, relationships, evidence, provenance, temporality, context, and confidence, enabling explainable reasoning by both humans and artificial intelligence systems. We formalize CKR as an 8-tuple (E, A, R, V, P, T, C, F) with mathematical properties, specify a five-level provenance model, and demonstrate through systematic analysis that no existing standard or framework — including computable-knowledge and provenance models — satisfies all eight required properties of a clinical knowledge claim simultaneously. We further show that this gap constitutes a structural bottleneck for five major clinical AI paradigms: foundation models, retrieval-augmented generation, agentic AI, neuro-symbolic AI, and explainability. CKR complements — and does not replace — DICOM, FHIR, SNOMED CT, and knowledge graphs. It is the layer that makes them composable into explicit, auditable clinical knowledge. Note: This is an author-formatted preprint, deposited to establish priority and enable open community feedback prior to formal peer review. Dr. Rafael Jacobus, DDS – ([email protected])