The BRIDG Project: A Model to Support Interoperability in Clinical Research

1
The BRIDG Project A Model to Support
Interoperability in Clinical Research
You cant build a skyscraper by nailing together
10,000 dog houses . Grady Booch

• Douglas B Fridsma, MD PhD
• Department of Biomedical Informatics
• University of Pittsburgh

2
Talk Organization

• Why do we need models like BRIDG?
• What is the BRIDG project?
• What have we learned?
• Does it work?
• Current status

3
Managing the Processes of Clinical Trials
Modern clinical research requires complex,
interconnected systems

• Clinical and translational research is a complex
  systems of organizations, people and information
  systems that require detailed descriptions of
• definitions of data (what)
• the process of care (how)
• so that differences in outcomes can be analyzed
  and compared

4
Modern clinical research requires complex,
interconnected systems

• But clinical and translational data must be able
  to interoperate with the larger research
  enterprise
• Other applications
• Other systems
• Other departments
• Standardization and the ability to hand off
  data and research results is critical to
  innovation

5
Modern clinical research requires complex,
interconnected systems
Ultimately, successful translational research
requires complex connections to exchange
information within and between organizations
6
What are we trying to do?

• To define implementation-independent domain
  semantics (define the what and the how of
  clinical research)
• To uncover the myriad of semantic ambiguities
  present in the complex domain of clinical
  research
• To build a foundation for achieving computable
  semantic interoperability
• The key to data integration

7
Interoperability

• The ability of multiple systems to exchange
  information and to be able to use the
  information that has been exchanged.

Datatypes Common data elements
8
Talk Organization

• Why do we need models like BRIDG?
• What is the BRIDG project?
• What have we learned?
• Does it work?
• Current status

9
What is BRIDG?Biomedical Research Integrated
Domain Group

• A model of the shared semantics of regulated
  clinical research
• A communication bridge between
• clinical research domain experts and technical
  experts
• different models of clinical research information
  (EPOCH-BRIDG)
• An open community of stakeholders interested in
  developing standards for exchanging information
  about clinical research
• HL7, NCI, CDISC, FDA, ITN
• The semantic foundation for application and
  message development in HL7, caBIG, and CDISC
• A foundation for research in knowledge
  representation and semantic interoperability

10
What is BRIDG?Biomedical Research Integrated
Domain Group

• A model of the shared semantics of regulated
  clinical research

11
So how did BRIDG get started?

• Two important streams of development that have
  been brought together into a collaborative
  framework
• CDISC 2003, started constructing an analysis
  model to map ODM standards to HL7
• NCI 2004, started caBIG initiative to construct
  a structured protocol representation and
  interoperability among clinical research in
  cancer

12
Desiderata

• Did not want to construct Yet another standard
• The good thing about standards is there are so
  many to choose from
• Open-source
• Modeled processes and organization after other
  successful open-source projects
• Mozilla, Firefox, Linux, etc
• Model is small groups (Technical Harmonization
  Committee), vet in large groups (Advisory Board)
• Provide a mechanism to scale the development work
• Parallelize the development
• Prevent collaborators from colliding with each
  other
• Allow us to modeling in the open
• In the BRIDG project, we have tried to
  operationalize the collaborative models required
  for roadmap initiatives

13
Current Classes in Core Elements
14
What is BRIDG?Biomedical Research Integrated
Domain Group

• A model of the shared semantics of regulated
  clinical research
• A communication bridge between
• clinical research domain experts and technical
  experts
• different models of clinical research information
  (EPOCH-BRIDG)

15
What Problem(s) Does BRIDG Solve?

• The Communications conundrum
• Experts know about the how to do clinical
  research but dont understand how to build
  software
• Technologists understand how to build software
  but dont understand the intricacies of the
  clinical environments and clinical research
• Any implementation (i.e. solution) is a
  compromise of the original problem statement
• Compromises must be chosen wisely
• should be based on a deep understanding of the
  problem and a dialogue between Problem Space and
  Solution Space Experts

16
The Communication Pyramid
Standardized Models (UML)

Non-standard Graphics
ad hoc Drawings
Problem Space
Solution Space
Implementation-Independent
Abstraction
Implementation-Specific
Structured Documents
Free-text Documents
Discussions
Communication
17
What is BRIDG?Biomedical Research Integrated
Domain Group

• A model of the shared semantics of regulated
  clinical research
• A communication bridge between
• clinical research domain experts and technical
  experts
• different models of clinical research information
  (EPOCH-BRIDG)
• An open community of stakeholders interested in
  developing standards for exchanging information
  about clinical research
• HL7, NCI, CDISC, FDA, ITN

18
Current Collaborators

• HL7
• Official domain analysis model for the HL7 RCRIM
  technical committee
• All HL7 messages must be able to demonstrate
  bi-directional semantic traceability before
  they can be balloted
• CDISC
• The integrative model for all current CDISC
  standards

19
Current Collaborators

• FDA
• Developing an HL7 message based on CDISC SDTM in
  with the RFA requires capturing the semantics in
  BRIDG
• Regulated Product Submission (RPS) is the next
  HL7/FDA message to use BRIDG
• Immune Tolerance Network (ITN)
• Using BRIDG to integrate open-source caBIG tools
  with existing MDA applications

• caBIG
• Utilized as framework for NCIs caBIG (standard
  within CTMS WS)
• Additional non-CTMS semantics (e.g. FireBird,
  CDUS, etc.) being incorporated

20
What is BRIDG?Biomedical Research Integrated
Domain Group

• A model of the shared semantics of regulated
  clinical research
• A communication bridge between
• clinical research domain experts and technical
  experts
• different models of clinical research information
  (EPOCH-BRIDG)
• An open community of stakeholders interested in
  developing standards for exchanging information
  about clinical research
• HL7, NCI, CDISC, FDA, ITN
• The semantic foundation for application and
  message development in HL7, caBIG, and CDISC

21
Achieving interoperability from a common semantic
foundation
IMPLEMENTATION SOLUTIONS
STAKEHOLDERS
22
What is BRIDG?Biomedical Research Integrated
Domain Group

• A model of the shared semantics of regulated
  clinical research
• A communication bridge between
• clinical research domain experts and technical
  experts
• different models of clinical research information
  (EPOCH-BRIDG)
• An open community of stakeholders interested in
  developing standards for exchanging information
  about clinical research
• HL7, NCI, CDISC, FDA, ITN
• The semantic foundation for application and
  message development in HL7, caBIG, and CDISC
• A foundation for research in knowledge
  representation and semantic interoperability

23
Pilot Study mapping BRIDG to EPOCH

• BRIDG
• HL7, caBIG, CDISC stakeholders
• Developed collaboratively with stakeholders
• shared domain model for protocol-driven clinical
  research
• Comprehensive
• Consensus-based
• Abstract and context neutral
• EPOCH
• Immune Tolerance Network (ITN)
• International collaborative research effort that
  sponsors clinical trials and mechanistic assays
  on immune tolerance
• EPOCH clinical trial model
• Developed at Stanford Medical Informatics
• Designed to provide semantic foundation for
  management of clinical trials

24
Approach taken to mapping between EPOCH and BRIDG

• Semantic alignment
• Overcoming representation language mismatch
• Overcoming representation choice mismatches
• Principal team members
• Douglas Fridsma (Pitt)
• Webster Kelsey (Pitt)
• Samson Tu (Stanford)
• Ravi Shankar (Stanford)
• Dave Parrish (ITN)

25
Approach taken

• Semantic alignment
• Use Excel spreadsheet to systematically review
  and document possible mappings
• Define necessary preconditions for mapping
• Overcoming representation language mismatch
• Overcoming representation choice mismatches

26
Semantic Alignment Excel spreadsheet
27
Semantic Alignment Restrictions on EPOCH

• Mapping from EPOCH to BRDG gt Place restrictions
  on EPOCH
• Only one schedule of activities
• Period has no subperiods
• Limited temporal annotations

28
Approach taken

• Semantic alignment
• Overcoming representation language mismatch
• Overcoming representation choice mismatches

29
Overcoming representation language mismatch
BRIDG-in-OWL
Scope BRIDG Study Planned View BRIDG Complex
Data Types
30
Approach taken

• Semantic alignment
• Overcoming representation language mismatch
• Overcoming representation choice mismatches

31
Overcoming representation choice mismatch Epoch
example
EPOCH
Period
Arm1Cycle1
Arm1Cycle2
PeriodTypeIntervention
PeriodTypeIntervention
Screening0
Arm2Cycle1
Arm2Cycle2
PeriodTypeScreening
PeriodTypeIntervention
PeriodTypeIntervention
BRIDG
Intervention epoch
Screening epoch
32
SWRL rule to map epochs

• EPOCHperiodTypes of periods correspond to
  BRIDGepochs
• EPOCH periodType.label corresponds to
  BRIDGepoch.code.displayName

33
Successfully used an EPOCH clinical trial to
configure BRIDG Patient Study Calendar application
SWRL rules
SWRL rules
Patient Study Calendar
Herold protocol in EPOCH
Herold protocol in BRIDG
Herold protocol in PSC XML

• Automated mappings except for one relationship
• Because of OWL/SWRLs open-world assumption,
  First epoch cannot be derived as an epoch that
  has no predecessor

34
Pilot conclusions

• Semantic interoperability requires
• Harmonization of subsets of ontologies/models
• Overcoming mismatches in representation languages
  and representation choices
• OWL restrictions and SWRL rules help to overcome
  semantic and syntactic mismatches
• Possible future work
• Continued harmonization of BRIDG/EPOCH
• Scalability and (semi-)automation of method

35
Talk Organization

• Why do we need models like BRIDG?
• What is the BRIDG project?
• What have we learned?
• Does it work?
• Current status

36
Lessons learned (so far) in the BRIDG project

• Scope keep it clear and focused (ie, solve a
  problem that exists) and standardize to the
  extend needed
• Keep the model generic, faithful, free of
  implementation-specific formalisms, and
  supporting the requirements
• Refine through experience, and not endless
  discussions

37
Scope of BRIDG

• The domain of the regulated clinical research
  information management technical committee in
  HL7
• Protocol-Driven Research with human, animal or
  device subjects, plus appropriate associated
  regulatory documentation.
• Analysis-level semantics
• Business processes
• Static structures
• Use-case driven
• in-scope and out-of-scope determined by the
  use-case
• Does not include vocabulary or terminology
  choices
• BRIDGCodedConcept Datatype links to other
  places where domain semantics can be
  represented
• Area of active research in understanding how
  ontologies and information models link to
  vocabularies and terminologies

38
Lessons learned (so far) in the BRIDG project

• Scope keep it clear and focused (ie, solve a
  problem that exists) and standardize to the
  extend needed
• Keep the model generic, faithful, free of
  implementation-specific formalisms, and
  supporting the requirements
• Refine through experience, and not endless
  discussions.
• Understand the difference between consensus,
  abstraction, and harmonization

39
The difference between Consensus, Abstraction and
Harmonization

• Consensus
• Consensus statements often achieve agreement
  through being ambiguous
• United Nations, guideline consensus statements
• Abstraction
• generalized models that are useful for a broad
  range of different domains (HL7 RIM)
• definitions are abstract, domain independent
  (although specific to an implementation) and more
  helpful for implementation than they are for
  domain experts
• Harmonization
• Creating definitions that all experts agree on,
  creating distinctions between concepts to clarify
  the semantics when they dont agree, and
  imbedding those semantics in the business
  processes

40
Harmonize Concepts, Not Words
Symbol Protocol
Source John Speakman
41
Lessons learned (so far) in the BRIDG project

• Scope keep it clear and focused (ie, solve a
  problem that exists) and standardize to the
  extend needed
• Keep the model generic, faithful, free of
  implementation-specific formalisms, and
  supporting the requirements
• Refine through experience, and not endless
  discussions.
• Understand the difference between consensus,
  abstraction, and harmonization.
• Models are only a piece of the puzzle
• Datatypes, vocabularies and terminologies provide
  additional clarification of the semantics

42
BRIDG is only one piece of the semantic puzzle

• Common model across all domains-of-interest
• The representation of clinical trials in BRIDG
• Model grounded on robust data type specification
• Common data elements (ISO 11179) in the cancer
  Data Standards Repository (caDSR)
• Methodology for binding terms from concept-based
  terminologies
• UML loader, semantic connector, Enterprise
  Vocabulary Server, LexGRID etc
• A formally defined process for defining specific
  structures to be exchanged between machines, i.e.
  a messaging standard
• HL7 and implementation specifications
• caBIG unified process/model driven architecture

43
Lessons learned (so far) in the BRIDG project

• Scope keep it clear and focused (ie, solve a
  problem that exists) and standardize to the
  extend needed
• Keep the model generic, faithful, free of
  implementation-specific formalisms, and
  supporting the requirements
• Refine through experience, and not endless
  discussions.
• Understand the difference between consensus,
  abstraction, and harmonization.
• Models are only a piece of the puzzle
• Datatypes, vocabularies and terminologies provide
  additional clarification of the semantics
• Use a larger development framework to organize,
  iterate, and trace the semantics
• Provides a mechanism to integrate multiple
  projects, manage change

44
Translational Research is a Cycle
Support for Research Protocol
Development
Support for CT Enrollment Management
ResearchDevelopment
Pre-Trial Setup
PatientEnrollment
New IdeaGeneration
DataAnalysis
PatientManagement
Reporting Administration
Financial Billing
Support for Data Mining Analysis
Support for CT Reporting Administration
45
The Unified Process(Iterative/Incremental,
Risk-Focused, Architecture-Centric)
46
In BRIDG Process Matters

• Modeled on open-source software development
  initiatives
• Scalable processes to support coordination and
  collaboration across multiple modeling groups
• Organizational structure with stakeholder
  representation
• Focus on semantics, not representation or
  implementation allows for variation around a
  common model

47
Lessons learned (so far) in the BRIDG project

• Scope keep it clear and focused (ie, solve a
  problem that exists) and standardize to the
  extend needed
• Keep the model generic, faithful, free of
  implementation-specific formalisms, and
  supporting the requirements
• Refine through experience, and not endless
  discussions.
• Understand the difference between consensus,
  abstraction, and harmonization.
• Models are only a piece of the puzzle
• Datatypes, vocabularies and terminologies provide
  additional clarification of the semantics
• Use a larger development framework to organize,
  iterate, and trace the semantics
• Provides a mechanism to integrate multiple
  projects, manage change
• The importance of capturing dynamic semantics
  (activities) of clinical trials research
• Clarifies the data and concept definitions
• Provides the context for use of the data
  structures
• Sets the stage for a service-oriented architecture

48
Work Process Information in BRIDG

• Activity Diagrams
• Captures the workflow of clinical trials research
• Provides the context for when (and how) data is
  exchanged
• Helps to clarify the definitions (the what) by
  describing how the data is used

49
Mapping the Clinical Research Domain
50
Talk Organization

• Why do we need models like BRIDG?
• What is the BRIDG project?
• What have we learned?
• Does it work?
• Current status

51
Recent Experience With BRIDGThe CTMS WS
Interoperability Project (1)

• Goal In 9 weeks (including Holidays), build
  interoperable support for three Use Cases across
  five applications in the Clinical Trial
  Management System (CTMS) Work Space
• Use Cases to be defined by group of SMEs as pain
  points
• Applications include
• Patient Study Calendar
• Static model has been harmonized with BRIDG
• Lab Hub
• Static model has been harmonized with BRIDG
• Adverse Event Reporting System
• Application development began concurrent with
  Interoperability Project
• Early analysis work already done and compliant
  with BRIDG
• Patient/Clinical Trial Registry
• Repository of links between patients and trials
• COT CTDMS (Oracle Clinical)
• Trial repository export capability only

Source Charlie Mead
52
Recent Experience With BRIDGThe CTMS WS
Interoperability Project (2)

• Process Iterative/Incremental SEP utilizing
  BRIDG as DAM
• Two one-month iterations
• Limited ability to change existing code base
• Process began with Business Modeling
• Activity Diagrams for each Use Case mapped to
  BRIDG
• BRIDG Extract generated based on AD ? BRIDG
  mapping
• BRIDG Extract ? XMI ? analysis-level XSD ?
  implementation-level XSM (common wire format)
• short time-line ? no formal messaging
  structures (V3 messages will be developed at a
  future date)
• The Conclusion The project could not have
  succeed without the use of a DAM
• The previous harmonization/common application of
  BRIDG by 3/5 of the applications (and the ability
  of the other 2 applications to map appropriate
  static structures to BRIDG) enabled the project
  to succeed

Source Charlie Mead
53
Talk Organization

• Why do we need models like BRIDG?
• What is the BRIDG project?
• What have we learned?
• Does it work?
• Current status

54
What have we accomplished?

• BRIDG is no longer a modeling exercise, but a
  robust model used by
• the National Cancer Institute for all application
  development
• Commercial clinical trials software developers
• CDISC to unify their existing models
• HL7 for all HL7 messages related to clinical
  trials research
• FDA for electronic data submission standards
• Other collaborators (immune tolerance network) to
  integrate applications and clinical trials work
  processes
• BRIDG release V1 June, 2007
• In the process we have
• Developed generalizable scaleable processes to
  support collaboration across organizations,
  models, and domains
• Begun to understand how to overcoming semantic
  and representational differences between
  different ontologies and terminologies

55
Value of analysis modeling for semantic
interoperability

• If you understand
• The processes (activities) that you do within
  your organization to support clinical and
  translational research
• The information (data) that you use for these
  activities
• You can
• Reduce redundancy in your organization
• Redesign organizational processes
• Know what to share (and what not to)
• Integrate custom and commercial applications (you
  know where they fit in your organizational
  activities and data)
• Create a shared understanding of the work of
  clinical trials research (facilitates culture
  change)
• This becomes critical for any program in which
  you want to
• Exchange data between different disciplines
  (translational research)
• Clinical trials and translational programs (CTSA)

56
Acknowledgements leadership

• Leadership and collaboration
• CDISC board members
• Becky Kush, CDISC
• Ken Buetow, Peter Covitz, Sue Dubman, Christo
  Andonyadis, John Speakman, NCI
• Charlie Mead, HL7, BAH
• Barbara Tardiff, Linda Quade, Ed Trip, Ed Helton,
  Randy Levin (FDA), RCRIM technical committee of
  HL7

57
AcknowledgementsOrganizations

• Technology companies
• ScenPro
• IBM
• SAS
• Fast track
• SAIC
• BAH
• Oracle

• Pharmaceutical companies
• AstraZeneca
• Boehringer-Ingelheim
• Eli Lilly
• GlaxoSmithKline
• Merck
• Novartis
• Pfizer
• Sanofi-Aventis

58
(No Transcript)
59
(No Transcript)
60
www.BRIDGproject.org

• Fridsma_at_cbmi.pitt.edu
• The Technical Harmonization Committee
• Smita Hastak, ScenPro
• Julie Evans, CDISC
• Charlie Mead, BAH
• Douglas Fridsma, University of Pittsburgh