Clinical informaticians and architects asking how the record is represented
The clinical knowledge graph
Before any model runs, the record has to mean something: people, episodes, observations, orders and organisations, linked and typed.
What is represented
Entities
Person, guardianship, episode, encounter, observation, order, result, plan, organisation and practitioner role.
Relations
Which observation justified which decision, and which plan step produced which appointment.
Terminologies
Codes are attached at capture, so a query does not depend on how a clinician phrased a sentence.
Why it matters for the AI layer
A model is only given the sub-graph a task needs. Scoping retrieval by graph shape is what makes minimisation enforceable instead of aspirational.
The six questions
- What is it?
- A typed graph of clinical entities and their relations, bound to recognised terminologies rather than to free text.
- Who is it for?
- Clinical informaticians, data leads and architects reviewing how the record is modelled.
- Which problem does it solve?
- A record stored as documents cannot answer a question that crosses two episodes or two organisations.
- How does it work?
- Every write creates or updates typed nodes and edges; codes are resolved at capture, not at reporting time.
- How is it different?
- The graph is the record, not an analytics copy of it, so an answer and a chart never diverge.
- What is the proof?
- The capability catalogue is generated from the same graph structure the product writes to.
This page is part of a platform of 2,156 clinical and operational capabilities across 8 business domains.
Explore the catalogueContinue
Before any model runs, the record has to mean something: people, episodes, observations, orders and organisations, linked and typed.
Medical review: 2026-08-24 · Legal review: 2026-08-24