The Health Data Platform Landscape: A Field Guide from openEHR to FHIR to the Cloud
— by Vivax
Two earlier posts mapped the same territory from two angles — who really owns the clinical record, and what a modern FHIR platform plus a Learning Health…
Two earlier posts looked at the health-data world from two angles: who really owns the record (the case for open standards like openEHR), and what a modern FHIR-based platform plus a Learning Health System actually looks like. This post pulls the whole vendor field together in one place — a field guide to the systems hospitals actually evaluate when they decide where their clinical data will live. It helps to group them by approach rather than by brand, because the architecture matters more than the logo. Three tiers cover most of the field: openEHR-native repositories that store the record in a vendor-neutral clinical format; FHIR-native repositories that make the exchange standard the storage model too; and the hyperscaler and enterprise platforms that bundle data, imaging, and AI tooling into managed services. None of what follows is about price. What matters is how each one answers the same question: how do you turn scattered, messy clinical data into a clean, trustworthy, AI-ready foundation?
Start with the openEHR-native tier, where the clinical record is stored in an open, vendor-neutral format defined by clinician-authored archetypes rather than a proprietary schema. EHRbase is the open-source anchor here — an Apache-2.0-licensed openEHR server that any hospital, vendor, or research group can deploy, inspect, and extend, with faithful support for the openEHR specifications and their query language, AQL. Alongside it sits Better Platform (from the company formerly known as Marand), a mature commercial openEHR data platform built around a vendor-neutral clinical data repository, with authoring tooling for archetypes and templates and a low-code layer for building applications on top. The two are complementary illustrations of the same idea: EHRbase shows that an open standard can be run on fully open-source infrastructure, while Better shows that the openEHR model is robust enough to underpin large-scale national and regional deployments. In both cases the point is the same — the data outlives the application that happens to read or write it.
The FHIR-native tier makes a different bet: take HL7 FHIR, the standard most often used to exchange health data between systems, and use it as the storage model itself. The clearest example is Smile CDR, the commercial clinical data repository from Smile Digital Health. It is built on top of HAPI FHIR — the widely-used open-source reference implementation of the FHIR standard — and wraps it with the security, identity, terminology, and operational features a production system needs, feeding the broader Smile Omni suite for data quality, measurement, and compliance. The appeal is directness: when the exchange format and the storage format are the same, far less translation sits between an external interface and the database, and integrations that would otherwise be brittle one-off projects become more uniform. Where openEHR draws a sharp line between persistence and exchange, the FHIR-native approach deliberately collapses it.
The third tier is the hyperscaler and enterprise platforms, which bundle storage, standardization, and AI tooling into managed services. Oracle Health Data Intelligence — the population-health and data platform that grew out of Cerner's HealtheIntent after Oracle's acquisition — aggregates data across sources to drive analytics and population-level insight. The big three clouds each offer a managed FHIR foundation: AWS HealthLake ingests disparate data, transforms it to FHIR R4, and adds medical natural-language processing; Microsoft Azure Health Data Services pairs a managed FHIR service with DICOM imaging and a MedTech device pipeline; and Google Cloud Healthcare API bridges FHIR, HL7v2, and DICOM into the rest of Google Cloud, where the data can flow into BigQuery and Vertex AI for analytics and model building. InterSystems HealthShare takes the health-information-exchange angle, building a unified care record and interoperability layer on the company's IRIS data platform. And IBM belongs in the story for historical reasons: its Watson Health portfolio was a prominent enterprise health-data play before those assets were divested and continued under the Merative brand — a useful reminder that this market consolidates and reshapes itself constantly. What unites the tier is the managed, all-in-one operating model: convenient and powerful, but with the data centralized inside someone else's cloud.
Across all three tiers the trade is the same: you get an AI-ready data foundation, but usually by centralizing the record inside a particular vendor's platform or cloud — and that is exactly the trade the Vivax open-source data platform is built to avoid. Like an openEHR-native repository, the Vivax Standardized Data Platform first grades every record for quality and reliability, then cleans, validates, and harmonizes it into unified, structured forms — and, like the best of the open tier, publishes those as open-source data archetypes and a new device-and-software standard that minimizes the layers between hardware, records, and model. But where a cloud lake centralizes data in order to learn from it, Vivax inverts that: in-place federated learning, differential privacy, and secure aggregation let a shared clinical world model improve across many hospitals while the raw data never leaves each hospital's own walls. The clinical world model stays open-source at its core, with only the implementation layer commercial on top. The result is what every platform in this field guide promises — a clean, standardized, AI-ready foundation — without the ownership and privacy trade-offs that usually come bundled with it.