Guide to the Assessment Set

What the decision is about

Clinical software must, among other things, do two different things with patient data. They are related, but they should not be confused.

Keeping data	Handing data on
The model in which an organisation retains clinical information over time – storage.	The standardised representation in which one system makes information available to another – exchange.

For the European data exchanges examined in this assessment, current EHDS implementation work, cross-border specifications and German requirements use HL7 FHIR. This establishes an exchange direction for them. It does not prescribe the model in which every organisation must keep its clinical record.

The persistence decision therefore remains open. The authoritative record may be held in a proprietary model, in FHIR, in openEHR, or in an architecture that combines these models. Each additional governed representation creates obligations for transformation, validation, version management and maintenance. A later change of the authoritative model is a migration, not a configuration change.

This guide sets out the state of play examined, explains the limits of the argument and shows which document to read for which purpose.

Scope

The recommendation addresses one defined situation: genuinely new European clinical-data infrastructure in which the authoritative storage model has not yet been fixed and standards-based exchange must be supported. New means that the architectural decision remains open. It does not mean that the organisation itself is new or that no systems are already operating.

The same recommendation does not automatically follow for established estates, migration programmes, secondary-use platforms or organisations with mature openEHR capability. In those settings, existing models, skills, interfaces and migration risks can outweigh the advantage of a simpler target architecture.

Interoperability is wider than exchange and storage

Exchanging data in a standard format, or storing them in one, is only part of interoperability. Interoperability also depends on preserving clinical meaning, constraining models tightly enough to test them, governing terminology, validating transformations, testing conformance, assigning organisational responsibility and integrating information into processes that can use it safely.

The OECD's 2026 cross-country assessment independently reinforces this wider framing. It distinguishes technical, semantic, legal, regulatory and organisational dimensions and identifies governance, trust, adoption, workforce capability and data quality as conditions of effective interoperability. This set examines the part of that larger field materially affected by data architecture and standards governance; Document 4, Appendix E-10 records where the OECD analysis supports that boundary and where its standard-specific claims require caution.

This set examines the conditions materially affected by clinical-data architecture and standards governance:

• which model holds the authoritative clinical content;
• which independently governed model boundaries that content must cross;
• how profiles, archetypes, templates and terminology bindings capture meaning;
• how transformations and conformance are validated;
• how artefacts, versions, responsibilities and export paths are governed over the life of the data.

It does not evaluate clinical workflow integration, usability, organisational readiness or health outcomes. These depend on applications, configuration and professional practice as well as on standards. The conclusions therefore concern the contribution of data architecture and standards governance to interoperability, not the interoperability of a health system as a whole.

The question, put precisely

Which model should a new infrastructure that must use FHIR exchange rely on to hold the authoritative clinical record, and which additional representations are justified by a demonstrated need and benefit?

This is often framed as a contest between openEHR and FHIR. That framing hides the operational choice. Both are open specifications, but they differ in modelling approach, governance, regulatory role, documented deployment and product support. Neither conformance to a specification nor semantic richness alone establishes interoperability.

The assessment therefore compares five architectural ideal types rather than ranking two standards in the abstract. It makes a bounded recommendation, identifies the evidence that supports it and states what the available evidence cannot establish. It neither recommends a supplier nor assumes that the same architecture is preferable in every setting.

The six documents

Document	Use it for	Main contents	Languages
1 — Decision Note	Taking or communicating the decision	The recommendation, its boundary and three questions for a procurement decision.	EN DE
2 — Executive Brief	Briefing executives and architects	The purposes of the standards, five architectural ideal types, their minimum governed model conversions, the evidence boundary, five procurement questions and the conditions that could change the recommendation.	EN DE
3 — Assessment	Examining the complete argument	The derivation from regulation, architecture and governance to procurement consequences, followed by the technical comparison of persistence, querying, terminology binding, templates, FHIR Questionnaires and OMOP. Includes a glossary.	EN DE
4 — Evidence Annex	Checking claims, cases and sources	Product and implementation evidence for both standards, European and German cases, comparison tables, source criticism, the graded source ledger, the primary-artefact register and the bibliography.	EN
5 — Technical Report: German FHIR Profile Governance	Examining German profiling and conformity	The profile families and governance record, an audited corpus of 2,996 StructureDefinitions, binding and derivation analyses, version management, the conformity-assessment regime and the counting rules.	EN DE
6 — Desiderata	Deciding what evidence and governance work should follow	Twenty-four testable requests to standards bodies, public authorities, procurers, providers, vendors and research funders. Each includes a condition by which later fulfilment can be assessed.	EN DE

Reading routes

• For a decision: Read Document 1. Add Document 2 when the architectural options, evidence limits or procurement questions need explanation.
• For assurance: Read Document 3 for the argument and the detailed Document 4 for the cases, source criticism and evidence grades. The Assessment cites the Annex where a proposition requires separate evidential appraisal.
• For German FHIR implementation and governance: Read Document 5. It contains the corpus, counting rules and audit record on which the German quantitative findings rest.
• For standards work, research and funding: Read Document 6. It turns identified gaps into testable requests and is not a procurement recommendation, but is meant as an encouragement to the specialist communities to begin the further work it identifies.

How the documents relate

Documents 1 and 2 summarise the recommendation. Document 3 derives it. Document 4 tests the evidential basis and its limits. Document 5 supplies the separate technical analysis behind the German profiling findings. Document 6 records what remains unresolved and what evidence or governance changes could alter a later assessment.

The documents are therefore complementary, not six independent statements of the same case. A summary should not be read as carrying more certainty than the underlying Assessment and Evidence Annex provide.

Citation and reuse

The set may be circulated as a whole or by individual document. Quotations of figures, counts or conclusions should identify the document, release and date because the evidence base and counted corpus can change.

Documents from different releases should not be combined without checking their cross-references, figures and evidence cut-off dates. A difference between two releases may reflect revised evidence, methods or scope rather than competing conclusions.

Method, and the use of a large language model

The literature search, the collation of sources, the drafting, and the consistency and readability checking across versions of this set were carried out with the assistance of a large language model. The editorial pass that set paragraph and sentence structure was made after the findings were fixed and did not change them. Not all of the cited literature was read in full: for part of it a summary extract was the working basis, and Appendix G of Document 4 records source by source which basis applies. Document 4 also states the limits of the method in full, including what a re-derivation of the evidence grades can and cannot establish. The grading rule, its application, the assessment and the conclusions are the author's, as is responsibility for any error that remains.