HL7’s Clinical Document Architecture Liora Alschuler HIMSS Dallas, Texas February, 2005 About me Liora Alschuler alschuler.spinosa, consultants Co-chair HL7 Structured Documents TC Co-editor, CDA Member, 2005, HL7 Board of Directors Project manager for Operation Jumpstart, (initial design of CDA) past Chair, XML SIG email@example.com Also contributing: Bob Dolin, MD, Kaiser-Permanente; John Madden, MD, Duke University Medical Center Clinical Document Architecture: CDA What is it? How does it allow you to: Do simple things simply Invest in information (doing complex things carefully) What is CDA? ANSI/HL7 CDA r1-2000 (Release one) On CD, includes prose spec, RMIM, DTDs December 2004 ballot (Release two) Structured Documents Tech Cmte, zip file on HL7.org – PASSED, will be published shortly A specification for document exchange using XML, the HL7 Reference Information Model (RIM) Version 3 methodology and vocabulary (SNOMED, ICD, local,…) CDA: A Document Exchange Specification This is a CDA and this and this and this and this and this and this CDA: A Document Exchange Specification A CDA can be a Discharge Summary Referral (CCR is one such) Progress Note H&P Public health report … any content that carries a signature CDA: XML XML is Extensible Markup Language (www.w3c.org) In XML, structure & format are conveyed by markup which is embedded into the information Sample CDA CDA = header + body CDA Header Metadata required for document discovery, management, retrieval CDA Body Clinical report Discharge Summary Referral Progress Note H&P Public health report … any content that carries a signature CDA Header The header describes: The document itself (unique ID, document type classification, version) Participants (providers, authors, patients…) Document relationships (to orders, other documents…) Metadata sufficient for document management CDA Body: two types of markup Human-readable “narrative block”, all that is required to reproduce the legal, clinical content Optional machine-readable CDA Entries, which drive automated processes CDA Body: Human-readable paragraph list table caption link required content revise (delete/insert) subscript/superscript special characters (e.g., symbols, Greek letters) in Unicode emphasis line break renderMultiMedia (non-XML graphics, video…) CDA Body: Machine Processible Clinical statement Observation Procedure Organizer Supply Encounter Substance Administration Observation Media Region Of Interest Act Optional CDA Body: Why isn’t XML + SNOMED enough? “hives”: SNOMED CT 247472004 “Dr. Dolin asserts that Henry Levin manifests hives as a previously-diagnosed allergic reaction to penicillin” First: human readable Allergy to penicillin Next: series of related statements Observation: RIM-defined Prior dx: SNOMED Allergy to penicillin: SNOMED Hives Prior dx: SNOMED Hives: SNOMED Hives is a manifestation of a reaction to penicillin Relationship: RIM-defined Then: supply context Who is the subject? Target: RIM-defined Id: local How are these concepts, relationships CDA RMIM defined? CDA Header CDA Body, Section, and Narrative Block CDA Entries Extl Refs Investing in Information CDA XML can be simple CDA XML can be complex Simple encoding relatively inexpensive Complex encoding costs more You get what you pay for: like charging a battery, the more detailed the encoding the greater the potential for reuse CDA: Return on Investment Low end: Access to documents “please send referral letter to…” “please get me the discharge summary…” “what imaging reports are available from the last episode?” High end: Reuse Send synopsis to tumor board Attach to claim for automated adjudication of payment Extract data for clinical research Low End Applications for CDA Persistant, accessible, human-readable documents Document requirements: CDA header Release One or Two body Narrative block Non-semantic markup (HTML-like) Document options: More complex markup can be inserted, to be used or ignored Low End Investment in CDA Many forms of document creation technology Voice (dictation, transcription) eForm EHR (CDA is output as “report”) What is the simplest way to create a CDA document? Enter minimal metadata Point to document body See HL7.org NLM Project: freely available application (by 3/1) What you can do with simple CDA documents: the registry hub 4. Retrieve 1. Create documents 2. Register “what imaging reports are available from the last episode?” 3. Discover Ubiquitous access to distributed information By class of document, patient, provider, encounter (CDA header metadata) Documents remain under local control Document creation technology evolves under local control Registry (hub) for access control, identifier xRef A tip of the hat to… Aluetietojärjestelmä 40% of Finnish population covered including Helsinki Investing in Information Simple documents retrieval, display metadata registry Two examples of higher-level investment: HIMSS 2004, Dr. John Madden, Duke University Medical Center, created a CDA pathology note that doubles as a tumor board report Also at Duke, the Starbrite “Single Source” Proof of Concept for clinical trials A single data REPRESENTATION standard facilitates multiple document PRESENTATION standards ! Pathologist view: CAP/ACoS standards compliant, templatedriven data entry Repository view: HL7-CDA standard XML with XQuery-ready, context-linked SNOMED encodings South Hospital Clinician view: Traditional format, print/electronic delivery South Hospital Tumor registrar view: Irrelevant items filtered, stage computed automatically One CDA, many applications: pathology Display or print (referring physician’s view Source CDA Archival CDA XML (pathologist, author’s view) Tumor Board, synopsis, meets CAP reporting guidelines Investing in Information “Single Source” Create once Use many Reuse clinical data in clinical trials Duke Clinical Research Institute Proof of Concept Principals: Landen Bain, Rebecca Kush, Liora Alschuler Microsoft, primary technology partner The Challenge: Integrate Patient Care and Clinical Research Data Patient Care World Clinical Research World Electronic Medical Record The Void Single Source vs Previous Solutions eSource & electronic data capture redundant with creation of clinic note require information reside in EMR/EHR proprietary data formats CDA & CDISC in “single-source” capture trial data, merge it into clinic note (re-use) work with current technology, workflow open, non-proprietary data formats CDA in Starbrite Trial Manual creation and re-entry of CRF HIS lab, ADT, meds, source documents LIS validation display db manual entry to CRF CLINIC re-key CRF CRO Current processes (dual source) HIS lab, ADT, meds, source documents LIS display dictate chart note Redundant creation of chart note CDA in Starbrite Trial Merged workflow: electronic CRF re-used in chart note HIS lab, ADT, meds, source documents LIS validation display CDA/ ODM db eCRF Proposed processes (single source) dictate chart note CLINIC ARO One CDA, many applications: clinical trials Clinic note inserted into patient chart Source CDA (principal investigator, author’s view) Archival CDA XML Case report form submission to research database See demo here, Microsoft pod! Investing in Information cost 80/20 √ benefit Disecting the curve What is easy: Header Human-readable body Low degree of coding What is hard: Concensus on semantic content requirements Model/vocabulary interface Investing in Information Example of what is hard TermInfo 2004 conference NASA August 1-4, 2004, Houston, TX Notes posted to the summit’s web page: http://csd2.no-ip.net/Composition/ and www.terminfo.org New project within HL7 Looked at issues raised by David Markwell (and previously identified by others) Investing in information: what is hard? Issue 1: Code/value dichotomy abdominal tenderness is observed examination (code) / abdomen tender (value); abdominal examination / abdomen tender; abdominal palpitation / abdomen tender; abdominal tenderness / present, etc. Investing in information: what is hard? Strong collaborative effort established to address issues Most syntatic issues addressed, full concentration on semantic interoperability Issues will be resolved, but will take time and experience Investing in Information: phased approach Lay groundwork CDA header metadata XML R1 or R2 CDA body Build Concensus on requirements Understanding of modeling process Vocabulary glossary Understand Relationship of vocabulary to model Introduce interoperable semantic content as requirements and business drivers dictate CDA: doing simple things simply, & more complex things slowly What can be done now Basic (Level One) CDA (R1 or R2) can be created with any degree of technical sophistication Document scanner+web form Transcription Electronic health record What can be done later Increase coding sophistication as business requirements dictate (return on investment and regulation) http://www.intersystems.com/mt_washington_vision.pdf Thank you! Questions?