200 likes | 304 Views
Status Update: hData Record Format, Rel. 2. SOA Working Group Meeting Mark Kramer July 22, 2013. hData Background. Created in 2009 in response to then-current healthcare standards, which were viewed as complex, somewhat technologically backward, and unsuitable for mobile health
E N D
Status Update:hData Record Format, Rel. 2 SOA Working Group Meeting Mark Kramer July 22, 2013
hData Background • Created in 2009 in response to then-current healthcare standards, which were viewed as complex, somewhat technologically backward, and unsuitable for mobile health • hData core principles: • Adoption of common web technology such as REST and Atom • Disaggregation of large medical records into granular clinical resources identified by URLs • Access to clinical resources via REST interface • Use of simplified, easily validated XML representations • hData addresses two parts of the information interoperability problem: • Framework for declaring and organizing resources (hData Record Format, or HRF) • REST realization of RLUS data operations (hData RESTful Transport, or HRT)
hData Timeline ONC NwHIN team looks at hData and REST hData REST Transport Beta(Jan 2012) hData-HL7 standardizationeffort begins(Feb 2010) begun Adopts hData for next generationdevice standards hData REST Transport Normative(Jan 2013) hData Record Format DSTUapproval(Sept 2011) 2009 2013 2012 2010 2011 HL7 “Fresh Look” kick-off (May 2011) First public presentation of hData atBalisage Conference(Aug 2009) ONC S&I Framework Project uses hData ONC NwHIN recommends FHIR + RHEx Medication hData Content profile DSTU Begins(Jan 2012)
hData and FHIR Comparison • hData and FHIR are complementary and partially overlapping • hData can be used when a combination of FHIR and non-FHIR resources are exchanged
hData Task Force • Mark Kramer • David Hay • Paul Knapp • Muhammed Asim • Sam Sayer • Erik Pupo
Task Force Process • Weekly meetings since Atlanta meeting • Meetings documented on HSSP Wiki • Major activities: • Reviewed and classified DSTU comments • Obtained consensus on direction for each issue • Created technical approaches • Implemented and reviewed document updates • Aiming for normative ballot in August-September
Goals for HRF Release 2 • Terminology clarification* • Root file content profile incorporation • Content profile simplification • Path templates* • Resource prefix* • JSON support* • Atom Feed FHIR Alignment* • Atom feed metadata simplification* *FHIR Alignment
Classification of DSTU Comments (by ID #) • Terminology Issues: 256 • Root File Multiplicity (including URL templates): 251, 257 • Query, Atom Feed Issues: 252, 253, 258 • Root File Links (to profiles, xsds, etc.): 83, 264, 265 • Other Issues: • Editorial Comments: 54, 250 • XML Issues: 53, 261
Terminology Changes • Intended to make the specification more accessible and to use terminology in a more standard way; also better aligned with FHIR
hData Record Format: Root Files (Release1) • HRF contains the recipe for creating “root” files • Root file is how an hData server documents itself • Client retrieves the root file (HTTP) from known URL • Root lists URLs (sections) and resource types (extensions) Root section @path: string 1..1 @extensionID: string 1..1 @requirement: extension: string 1..1 sections 0..N Root Id: string 1..1 version: integer 1..1 created: dateTime 1..1 lastModified: dateTime 1..1 extension extensions 0..N @extensionID: string 1..1 @contentType: string 1..1 (MIME type) extension: string 1..1 (definitional URL)
hData Record Format, Release 2 NEW XML or JSON NEW
JSON Conversion • Root file and Atom feed can be provided in XML or JSON • We follow FHIR rules of thumb with some differences: • FHIR can’t handle complex elements with text values • FHIR JSON doesn’t distinguish attributes from child elements and therefore isn’t perfectly reversible • FHIR JSON doesn’t deal with XML namespaces and so naming collisions are possible
Patient Specific Endpoints (Release 1) Patient-specific baseURL Assumed prior knowledge +/patient ID Root hierarchy URL fragment Root hierarchy … No assumption about behavior here Root hierarchy Probably the same root files (but not required)
Path Templates • Templates denoted by {curly_brackets} • [baseURL]/Patients/{Patient.id}/Conditions • [baseURL]/Patients/{Patient.id}/Allergies • Templates allow "dependent resources" that are owned by an existing resource • DICOM: Studies > Series > Images Example: ../Patients/{Patient.id}/Radiology/Studies/{Study.id}/Series/{Series.id}/Images ../Patients/1234/Radiology/Studies/567/Series/2/Images • Templates can also be used for parametric access, e.g.: ../Patients/{Patient.id}/Radiology/Studies/{YYYY}/{MM}/{DD} ../Patients/{Patient.id}/Radiology/Studies/2013/07/22
Patient-Specific Hierarchy, Release 2 • Template example: [baseURL]/Patients/{patientID}/medications The only required prior knowledge Possible service to get patient IDs /Patients Base URL /patient 1 hierarchy Root Patient-specific resources /patient 2 hierarchy … /patient N hierarchy Same hierarchy
Resource Prefix Use of templates can lead to confusion between sections and resources: ../Patients/1234/Radiology/Studies/567 → Study #567 ../Patients/1234/Radiology/Studies/2012 → Study #2013 ?!? Solution: Use resource prefix, like in FHIR (“@”) ../Patients/1234/Radiology/Studies/@567 → Study 567 (resource) ../Patients/1234/Radiology/Studies/2012 → List of studies from 2012 (Atom) Resource prefix used by default; user declaration can turn off
Atom Feed • FHIR Alignment • Aligned Atom feed with FHIR element profile • Title changed from URL to description • Clarified use of link element • JSON feeds supported • Simplified and aligned additional metadata
Atom Feed: Metadata • Release 1 required metadata on Atom <entry> to represent pedigree, confidentiality, styled after CDA header • Too complicated, should not have been required (since some resources represent their own pedigree internally) • We re-modeled the metadata after the FHIR Provenance resource
hData Content Profiles (HCP), Release 2 • Re-written to be more informative, less prescriptive • Emphasize that HCPs are just domain-specific implementation guides for using hData • We provide suggested contents for HCP, but no requirements • Removed schema for HCP Definition Documents • Sufficient for authors of HCP to provide a sample root file
HRF Release 2 Document Status • Almost complete (only missing JSON in Examples section) • Waiting on feedback from some team members • Significantly upgraded document • New FHIR-style UML diagrams and pseudo-XML, JSON • Updated schemas, examples • Better descriptive writing • Don Lloyd has indicated flexibility on submission format • Provide MS-Word, he will convert to PDF