HL7 CDA R2: structured clinical documents and the Italian profiles

AIHealth

On-premise clinical platform with local LLMs, RAG on FHIR/DICOM data, diagnostic support, remote follow-up. Architecture designed for the MDR pathway.

Discover AIHealth →

Digital Health

Medical software development compliant with CE and MDR regulatory standards. Clinical decision support systems, AI integration in clinical workflows.

Discover →

From messaging to the document

The HL7 v2 standard describes events through messages: an admission, a prescription, a freshly produced report. There is however clinical information whose nature is that of a document: a hospital discharge letter, an emergency department report, a patient summary. They have an identified author, a legal signature, a version, their own lifecycle — and for this reason HL7 developed a separate and complementary standard: the Clinical Document Architecture.

Version 1 of CDA was published in 2000 (then called Patient Record Architecture), and Release 2 (CDA R2) was approved as an ANSI standard in 2005. Since then CDA R2 has been the reference document format for exchanging clinical documents between systems — and, above all, for populating national and regional Electronic Health Records in Italy and across Europe.

The model: RIM, XML, six principles

CDA R2 derives from the Reference Information Model of HL7 v3. Information objects (Act, Observation, Procedure, Organization, Patient, Encounter) and their relationships are modelled according to RIM and serialised as XML. The document schema is defined by an official XSD; presentation to the professional happens through a standard XSLT stylesheet.

The standard is built around six principles, explicitly stated by HL7:

1. Persistence — the document exists in an unchanged state for a defined period
2. Stewardship — it is maintained by a designated organisation
3. Potential for authentication — it is legally signable
4. Context — it establishes a default context for the entire content
5. Wholeness — authentication applies to the whole document
6. Human readability — narrative content is readable without specific tools

The structure: Header and Body

Every CDA R2 document is made of two parts:

• Header — machine-readable metadata: document identifier, document type (a LOINC code in code), patient (recordTarget), authors (author), custodian (custodian), legal authenticator (legalAuthenticator), related documents (relatedDocument), delivering organisation (componentOf / encompassingEncounter)
• Body — clinical content, which can be either nonXMLBody (an encapsulated PDF or other file) or structuredBody split into section

Each section contains a narrative <text> block — readable by the end user through the rendering XSLT — and, optionally, structured and coded <entry>. The obligation of human readability materialises in the relationship between narrative and entries: the narrative is not derived from the entries but is their authoritative representation; the entries are a computable annotation on the narrative.

The three levels

CDA R2 defines three levels of structuring:

• Level 1 — document with a structured header and narrative body (including encapsulated PDF). Minimal formalisation of clinical information
• Level 2 — coded sections (identified by a LOINC code or equivalent) with still mostly narrative content
• Level 3 — fully structured entries, with observations, procedures and values coded against controlled terminologies (SNOMED CT, LOINC, ICD-9-CM/ICD-10, ATC, …)

The choice of level is a design decision: a full Level 3 enables automated analysis and reasoning but requires structured production from the EHR systems; a Level 1 is always producible from a document already existing as free text.

The Italian profiles

CDA R2 is a broad standard: real use in national systems requires Implementation Guides that restrict its optionality, tighten its constraints and specify terminologies. In Italy this work is carried out by HL7 Italia, an affiliate of HL7 International, in cooperation with the Ministry of Health, AgID and the Technical Table on the Electronic Health Record.

The first-generation Italian CDA R2 profiles focus on the clinical documents most relevant to the FSE:

• Hospital Discharge Letter (Lettera di Dimissione Ospedaliera, LDO) — the main document at the end of a hospital stay, with history, diagnoses, procedures performed, significant tests, discharge therapy, instructions to the treating physician
• Laboratory Report — laboratory test results, structured as LOINC-coded observations with reference values, units, textual observations
• Emergency Department Record (Verbale di Pronto Soccorso) — closure document of an ED access with triage, history, diagnoses, procedures, outcome, discharge
• Patient Summary (Profilo Sanitario Sintetico) — summary view of the patient: demographics, active problems, allergies, current therapy, vaccinations
• Radiology Report — CDA version of the textual report, with possible integration of related DICOM studies
• Specialist and Pharmaceutical Prescription

Each profile specifies:

• The templateId of the document and sections (OIDs under the Italian root 2.16.840.1.113883.2.9)
• The mandatory and optional sections, with the LOINC code for each
• The terminologies allowed in each entry (ATC for drugs, ICD-9-CM for coded diagnoses, LOINC for observations)
• The XSLT transformation rules for presentation

Signature and preservation

CDA R2 documents destined for the FSE must be digitally signed (CAdES or XAdES on the XML file) by the authoring physician or legal authenticator, and preserved in compliance with the Digital Administration Code. Integration with remote signature systems and hospital HSMs is one of the more delicate implementation junctions, together with document lifecycle management (cancellation, replacement, addendum) which the standard handles through relatedDocument.

Tooling

Production and consumption of CDA R2 are supported by open tooling:

• MDHT (Model-Driven Health Tools) — Eclipse/OpenHealthTools project for generating Java APIs from MDMI/UML models of CDA, with automatic validation against the profile
• HAPI Structures CDA — Java library of the HAPI family, part of the University Health Network ecosystem
• Italian Schematron — HL7 Italia distributes Schematron schemas that validate the conformance of a CDA instance to the Italian profile, complementary to XSD validation

The landscape is not without friction: the complexity of CDA R2 — inherited from the RIM model — is the target of recurring criticism, and the production time of a coherent Level 3 document is not trivial. These very issues are pushing HL7 to work, in parallel with CDA, on a new model oriented to resources and REST APIs; work has started and should lead to a first public release in the coming years.

In the FSE

For the Italian FSE, CDA R2 is today — at 2013 — the intended population format. Documents produced by providers are indexed and shared through the IHE XDS frame (profiles XDS.b, XDS-SD for scanned documents), with the regional registry managing metadata and the repository storing the files. Full deployment awaits the operational technical rules — the DPCM implementing Article 12 of DL 179/2012, expected in the coming years.

References: HL7 Clinical Document Architecture, Release 2 (ANSI/HL7 CDAR2-2005). HL7 Reference Information Model (RIM). HL7 Italia CDA R2 profiles. IHE Content Profiles (XDS-SD, CCD). MDHT — Eclipse Model Driven Health Tools.