320 likes | 471 Vues
BRIDG Update. January HL7 Working Group Meeting 15 January 2008. BRIDG Technical Harmonization Committee (THC) Members. Charlie Mead Smita Hastak Bron Kisler Steve Sandberg Becky Angeles Wendy Ver Hoef Lewis Frey. BRIDG Update. Overall status BRIDG Vocabulary SDTM BRIDG Review Team
E N D
BRIDG Update January HL7 Working Group Meeting 15 January 2008
BRIDG Technical Harmonization Committee (THC) Members • Charlie Mead • Smita Hastak • Bron Kisler • Steve Sandberg • Becky Angeles • Wendy Ver Hoef • Lewis Frey
BRIDG Update • Overall status • BRIDG Vocabulary • SDTM BRIDG Review Team • Education
BRIDG Scope Protocol-driven research and its associated regulatory artifacts, i.e. the data, organization, resources, rules, and processes involved in the formal assessment of the utility, impact, or other pharmacological, physiological, or psychological effects of a drug, procedure, process, or device on a human, animal, or other biologic subject or substance plus all associated regulatory artifacts required for or derived from this effort.
BRIDG: Recent Significant Progress • NCI is standardizing on the BRIDG as basis for semantic interoperability in the CTMS Workspace • CDISC’s Trial Design and SDTM standard are now represented in BRIDG
Release 2.0 Plan – 1 • Release 2.0 scheduled for April 2008 • R2.0 new content: AE, C3PR (NCI’s Patient Registry) • Full binding of all static attributes to HL7 V3 data types • Candidate terminology lists/value sets (drawn from existing standards if possible) for all attributes bound to ‘coded concept descriptor’ data types • Consolidation of ‘business process pillars’ in the model with guideline of <<not>> creating the RIM problem of ‘duplicitous attributes.’
Release 2.0 Plan - 2 • Introduction of formal business rules as class invariants using both free text and equivalent parsable OCL statements • Representation of ‘player/scoper’ semantics • Complete mapping of BRIDG Model static attributes to the HL7 RIM • Test evaluation of strategy for incorporating BRIDG Model ‘sub-domains’ (e.g. cancer-specific semantics, TB semantics) using NCI’s CTOM • Creation of more exemplary instance diagrams (e.g. AE content, SDTM IG, etc.)
Recent BRIDG Infrastructure Changes • Current in-progress model available on Gforge site (access via www.bridgmodel.org) • Look on Gforge site for bug and enhancement trackers • Look for BRIDG news on Gforge site • Project mapping spreadsheets will be owned and maintained by project teams rather then the BRIDG THC. So, these spreadsheets will no longer be published with each release. BRIDG THC will publish the first mapping from BRIDG to RIM.
BRIDG 1.0 Vocabulary • Oct-Nov / NCI EVS Team conducted thorough analysis of BRIDG 1.0 “Parent Classes” and “Attributes” and aligned with NCI Thesaurus concepts / definitions (256 items assessed) • A few discrepancies were noted that needed to be resolved by BRIDG THC (e.g., Sex vs. Gender, Participant vs. Study Subject) • BRIDG 1.0 has been loaded into caDSR
BRIDG 2.0 Vocabulary - 1 • BRIDG 2.0 “Parent Classes” and “Attributes” need to be aligned with NCI Thesaurus concepts • BRIDG 2.0 will include “vocabulary binding” for attributes with CD data types (approx. 400 value sets total) • Categorize as…(1) CDISC developed; (2) NCI developed; (3) other known Controlled Vocabulary; or (4) needs to be developed
BRIDG 2.0 Vocabulary - 2 • CDISC and NCI value sets identified to be aligned with BRIDG = 84; value sets still requiring work = 308 (Note: some value sets are shared across BRIDG attributes, such as targetSite) • BRIDG 2.0, including vocabulary binding, to be released and implemented in caDSR in April • NCI internal timelines require value sets to be defined at a faster pace than the current CDISC / RCRIM vocabulary process
SDTM BRIDG MappingReport to RCRIM Diane Wold January 15, 2008
SDTM Harmonization with BRIDG • Some UML modeling of SDTM in 2005 & 2006, but model quite different from BRIDG Release 1.0 • During harmonization discussions summer 2007, we decided it would be more useful to map from the domain tables in the SDTMIG • SDTMIG contains content beyond SDTM itself • More accessible to SDTM users • Did not include areas to be covered by other groups (e.g., AE, Trial Design) • Mapping from SDTM to BRIDG is included in BRIDG Release 1.1
SDS Team Review of Mapping • Goals • Check accuracy with a wider group of domain experts • Increase BRIDG knowledge in SDS Team • Method • For each SDTMIG domain, build a UML class diagram including relevant classes • Use notes to show mappings • Record issues, questions, and comments
Status of Review • Demography completed • Lab completed • Chosen because it includes most SDTM Findings variables • Vitals Signs completed (very similar to Lab) • ECG, Physical Exam and Questionnaires being reviewed as a group • Chosen to cover remaining Findings variables • Comparisons to lab to ensure consistency, speed review • Interventions not yet reviewed • Events not yet reviewed • Since AE not mapped, review will be limited
Issues Identified by Review • Corrections to mapping, many at datatype level • Example: COUNTRY mapped to “value” within AD datatype; should map to “country” • Corrections to BRIDG • Example: StudyReferenceDateRange meant to represent SDTM RFSTDTC, but definition was inaccurate. • Issues for SDS team • Example: SDTM IG allows two different uses for PEBODSYS, result of coding or pre-specified category; review group recommended eliminating the latter. • Points to be clarified with other standards (Lab, RCRIM aECG) • Example: Is method of assay part of Lab model?
Review Team Benefits • Appreciation of UML and of datatypes • Disciplined modeling approach clarified areas of “gut level” uneasiness • Overloading of variables, such as PEORRES (can contain either an observation or an assessment as NORMAL) • Handling of assessments by different mechanisms in different domains (INTP a test in ECG but NORMAL a result in PE) • Explaining issues to THC led us to better document how SDTM results variables are populated
Lab results flow Sponsor copies LBORRES into LBSTRESC Is result numeric? LBORRES populated NO YES Is there a text standardization for this test? Sponsor copies LBORRES into LBSTRESC and LBSTRESN Are there units? NO NO YES Sponsor places original units In LBORRESU YES Assumption: for a test with units, the sponsor has identified units to standardize on Sponsor performs standardization of text Original units = standard units? NO YES Sponsor places standardized text in LBSTRESC Sponsor copies LBORRES into LBSTRESC and LBSTRESN Sponsor performs unit conversion Assumption: a test cannot have both a text standardization and units. Sponsor places converted measurement in LBSTRESC and LBSTRESN Sponsor places standard units in in LBSTRESU
“Had we but known…”Starting with BRIDG might have avoided SDTM problems • During development of Microbiology domain, we realized we were using “method” variable for both method of assay and method of specimen collection • SDTM includes overloaded fields such as PEORRES and DSDECOD (COMPLETED or reason for non-completion) • We were slow to realize that TESTCD may not uniquely identify a test. E.g., a lab test may also require specimen and method. This issue now requires metadata solutions.
Conclusions • No show stoppers identified by review of SDTM mapping • Any new SDTM domains developed must be mapped to BRIDG • CDISC standards handle datatypes differently; issue needs to be addressed • Use of BRIDG seems to be fulfilling promise of harmonizing standards, also improving their quality
Unified Modeling Language • Used in the BRIDG model • The industry-standard language for specifying, visualizing, constructing, and documenting the requirements of software systems • The BRIDG model uses these UML diagrams: • Class diagrams • Activity diagrams • Instance diagrams
UML Class Diagrams • class – a concept of primary importance the domain-of-interest, depicted as a rectangle labeled with the concept’s name • attribute (including datatype specification) – a descriptive feature of a class, depicted as being contained within the class • relationship – one of several types of “lines” between classes
Class diagram example attribute class relationship multiplicity
The Pillars of InteroperabilityNecessary but not necessarily sufficient • Common model across all domains of interest • Foundation of rigorously defined data types • Methodology for interfacing with controlled vocabularies • Formal process and tools for defining interchange structures Source: Charlie Mead, MD, HL7
Foundation of rigorously defined data types • Simple vs Complex • Simple: Character, String, Text, Numeric
R1 Important Content Concepts • Planned, Scheduled, Performed Study • ObservationResult vs. Assessment • Analysis and Reporting
The BRIDG Backbone Classes • Person • Organization • Material • StudyProtocol • Documentation • Activity • ActivityRelationship • ObservationResult • ObservationResultRelationship • Assessment