Data platforms
Healthcare Data Platform Architecture
The same two words, data platform, describe an integration engine and a cloud lakehouse, and a procurement conversation in which the two sides mean different ones is the normal case. This paper names the parts. It places the platform in the five-stage stack, follows one laboratory result through every stage with the payload at each, and takes the six components in turn: the canonical model, data contracts, lineage and audit, access control, quality monitoring, and event and bulk delivery. It treats patient identity and terminology as the shared services they are, works a Fellegi–Sunter match by hand, and sets out the release calendar a terminology service has to absorb. It closes on what governs each part of the stage, from normative standards to nothing at all. Written from the specifications, the national data and the code systems themselves, with every version and status current as at October 2026 and every Canadian particular called out.
What’s inside
- The two systems that share the name data platform, compared across ten dimensions, and what each does when it is bought as the other
- The five-stage stack, and ONC’s national figures on the gap between receiving a record and integrating it
- One potassium result followed from an HL7 v2 segment to a FHIR Observation, an OMOP row, a feature and a population query, with every code checked
- The canonical model: vendor dimensional, OMOP CDM v5.5 and SQL on FHIR compared, with a working ViewDefinition
- Data contracts: the five things a contract covers, and the general-purpose standard that exists for them
- Lineage, provenance and audit, and PHIPA’s enacted but unproclaimed electronic audit log
- Access control as six layers, with the HL7 purpose and sensitivity codes and Ontario’s lockbox
- Data quality in Kahn’s terms, eighteen example checks, and what OHDSI’s Data Quality Dashboard counts
- Event and bulk delivery, with the Bulk Data export exchange step by step
- Patient identity by Fellegi–Sunter, worked through, and why Canadian matching is a different problem
- Terminology as a service: thirty scheduled releases a year across five code systems, and the one Canadian classification standing still
- Four questions that decide the shape of a platform, and residency answered at the strength the law supports
- What governs each part of the stage, from normative standards to nothing at all
- A glossary, and a source list separating specifications from regulators from data
Request this resource
Related insights
- Article · Data platforms
What a healthcare data platform is made of
The same two words describe an integration engine and a cloud lakehouse. Naming the parts, six components and two shared services, is what makes two very different platforms comparable.
· 14 min read - Article · Data platforms
The data platform is the evidence package
Of 691 AI devices the FDA cleared through July 2023, 95.5% never said who was in the training data. That is not a paperwork failure. It is a platform that could not produce the answer.
· 12 min read
