Deterministic Clinical Decision Support Digital Ontology
Use case: Deterministic Clinical Decision Support Domain: Healthcare – evidence-based prescribingWhat the ontology represents.
This ontology models how a clinician (or an AI agent acting on their behalf) reasons from a patient's situation to a safe, evidence-based prescription. It captures the four moving parts of evidence-based medicine: who the patient is, what the guideline says, what drug fulfils the guideline, and what would make the recommendation unsafe. The eight entities (Patient, Disease, Biomarker, Recommendation, Guideline, DrugClass, Drug, DoseInfo, Contraindication) and nine relationships form a single connected graph that can be traversed in one query to answer a real prescribing question.
Why the relationships matter. Healthcare AI fails when it hallucinates. By making the rules of reasoning explicit as graph edges – APPLIES_IF for when a recommendation fires, UNLESS for safety blocks, BELONGS_TO as the bridge from abstract drug class to concrete drug and dose, and CITES to anchor every claim to its source guideline – we get a model where an AI agent's job is reduced from generating an answer to traversing one. The same ontology that helps a junior doctor understand "why is this drug recommended and when shouldn't I give it?" lets an AI agent produce a deterministic, auditable, fully-cited answer. This is the core insight Fabric IQ is built on: shared meaning between humans and AI, expressed as a graph.
Real-world grounding
This ontology is the conceptual core of the Digital Doktor system, a deterministic clinical decision-support prototype built on Microsoft Fabric and Azure AI Search. Every entity and relationship in the model corresponds to a node label or edge type. It is a working contract between domain experts (cardiologists who validate triples), the knowledge base (ESC, FASS, Janusinfo), and the AI agent that serves clinicians.
