BRIDG: A Shared Analysis Model of the Domain of Clinical Trials Research PowerPoint PPT Presentation

presentation player overlay
1 / 69
About This Presentation
Transcript and Presenter's Notes

Title: BRIDG: A Shared Analysis Model of the Domain of Clinical Trials Research


1
BRIDG A Shared Analysis Model of the Domain of
Clinical Trials Research
  • Douglas B. Fridsma, MD, PhD
  • Center for Biomedical Informatics
  • University of Pittsburgh School of Medicine

2
What is BRIDG?
  • An open community of stakeholders interested in
    developing standards for exchanging information
    about clinical trials
  • A bridge between
  • clinical trial domain experts and technical
    experts
  • different models of clinical trials information
  • A formal model of the shared semantics of
    regulated clinical trials research
  • The semantic foundation for application and
    message development in HL7, caBIG, and CDISC

3
A Team Effort
  • Modeling Team
  • Christo Andonyadis
  • Peter Abramovitsch
  • Greg Anglin
  • Lisa Chatterjee
  • Julie Evans
  • Douglas B Fridsma
  • Smita Hastak
  • Ray Heimbuch
  • Charlie Mead
  • Joyce Niland
  • John Speakman
  • Cara Willoughby
  • Diane Wold
  • The Trial Design group in CDISC
  • Leadership and collaboration
  • Sue Dubman, NCI
  • Becky Kush, CDISC
  • Linda Quade, Randy Levin, Barbara Tardiff, HL7
  • Charlie Mead, HL7

4
Team Members (2)
  • Current modeling team
  • Christo Andonyadis
  • Greg Anglin
  • Lisa Chatterjee
  • Julie Evans
  • Douglas B Fridsma
  • Smita Hastak
  • Joel Hoffman
  • Charlie Mead
  • Joyce Niland
  • John Speakman
  • Cara Willoughby
  • Diane Wold
  • Julie Evans
  • Ed Helton
  • Pierre-Yves Lastic
  • Tony Friebel
  • Don Kacher
  • Barbara Tardiff
  • Chuck Jaffe
  • Frank Newby
  • Sally Cassels
  • Nitin Gupta
  • Landen Bain
  • Laura Altshuler
  • Steve Ruberg,
  • Elaine Job
  • Becky Kush
  • Sylvia Collins
  • Udo Siegman
  • David Hardison
  • David Iberson-Hurst

5
BRIDG Development Philosophy
  • Scope keep it clear and focused (ie, solve a
    problem that exists) and standardize to the
    extend needed
  • Refine through experience, and not endless
    discussions. This keeps the modeling effort clear
    and focused
  • Keep it generic, faithful, free of implementation
    specific formalisms, and supporting the
    requirements
  • Model in the open
  • Collaborate until it hurts
  • If the tools and models dont work with reality
    it is probably the tools and the models that need
    to change

With thanks to Dipak Kalra for discussion
6
The current stakeholders in BRIDG
  • Health Level Seven (HL7)
  • BRIDG is the domain analysis model (DAM) for HL7
    RCRIM TC
  • Clinical Data Interchange Standards Consortium
    (CDISC)
  • Consortium of pharmaceutical companies
  • Develops standards for regulatory reporting and
    drug development submissions
  • Key participant in HL7
  • caBIG CTMS workspace and NCI
  • Has real needs for applications and
    interoperability standards
  • Funding the participation of cancer centers
  • Has created a focus for integrating various
    standards organizations

7
BRIDG organization
  • BRIDG advisory board
  • Strategic direction with representation from all
    the stakeholders (CDISC, HL7, caBIG)
  • Technical Harmonization group
  • Managing the technical process of harmonizing and
    mentoring subprojects
  • Small group of individuals charged with the
    technical aspects of harmonization
  • Have to make choices, but approval and vetting
    rests with stakeholders

8
(No Transcript)
9
(No Transcript)
10
Active Subprojects
  • Trial design (Diane Wold, GSK and Peter
    Abramowitch, Fasttrack)
  • CDISC trial design subcommittee has moved their
    work into enterprise architect and UML modeling
    to speed the development efforts
  • Update the model every 2 weeks with results of
    teleconference
  • Statistical design (Greg Anglin, Lilly)
  • eDCI subproject (Don Kacher, Oracle, Christo
    Andonyadis, NCICB)
  • General purpose HL7 message to describe a data
    capture instrument, or case report form. Using
    HL7 early adopter mechanisms.
  • Actively using gForge documents, forums,
    webpages, and CVS repository

11
Active Subprojects
  • Clinical Registry message (Lakshmi Grama)
  • Development of an HL7 message for registering a
    clinical trial in a clinical trial repository
  • Eudract (europe)
  • Clinicaltrials.gov, PDQ
  • Traditional HL7 project development (not early
    adopter)
  • Adverse events (Joyce Niland, COH)
  • First significant submodel to be harmonized
  • Will lead to both application development and
    message specification
  • Harmonization branch and maintenance
  • New concepts/classes/definitions added from CTOM
    (Smita Hastak, ScenPro)
  • Cleaning up the model and definitions
  • Moving away from strict HL7 RIM classes an
    emphasis on clarity in describing the domain

12
What have we accomplished?
  • BRIDG
  • Established excellent collaboration with CDISC,
    HL7 and other caBIG modelers
  • Constructed the initial pieces of a comprehensive
    model still much to do
  • Have established this model as the HL7 Domain
    analysis model for regulated clinical research TC
    (RCRIM)
  • BRIDG will serve as an integrator for CDISC
    message harmonization
  • Development of a clinical trials reference
    implementation based on BRIDG (CTOM, caAERS)
  • Establishing both organizational and procedures
    to maintain and grow the model
  • This model will serve as the semantic foundation
    for data interchange specifications in HL7,
    CDISC, the NCI, and the clinical trials space of
    caBIG

13
Rich semantics captured in BRIDG
  • CDISC glossary of terms
  • Semantics and examples at the class and attribute
    level
  • Datatypes help to clarify the semantics around
    the classes and attributes

14
(No Transcript)
15
Current Priorities
  • Static Structures
  • Cleaning up the current model to make it more
    accessible to both domain experts and technical
    experts
  • Development of a harmonizable analysis model of
    the CDISC SDTM (submission data tabular model)
  • Harmonization of adverse events model from CoH,
    harmonization with HL7 Patient Safety SIG UML
    model(s)
  • Dynamic Structures
  • Defining the Dynamic models at present
  • Harmonize with COH dynamic models
  • Linking dynamic structures to static structures
  • Developing processes to support collaboration,
    input into BRIDG and use of BRIDG
  • Work with Training Group to develop BRIDG
    educational bootcamp for domain and technical
    experts

16
Existing Standards Harmonization
Existing DMIM
  • Process A
  • Harmonize existing HL7 DMIM for regulated
    clinical research
  • Process B
  • extract domain knowledge from existing messages
    and systems into formal UML models
  • Process C
  • Harmonize and merge UML models into BRIDG

A
BRIDG model
C
C
UML submodel
UML submodel
B\
B
Existing HL7 message
Existing object model
17
New Message and Application Development
  • Process D
  • Generate a release of BRIDG for formal
    requirements traceability with vetting within
    stakeholder organizations
  • Process E
  • HL7 message generation using HL7 mechanisms
  • Process F
  • caBIG application development with annotation,
    semantic connector, UML loader

UML loader Semantic Cntr
HL7 message generation
DMIM
Annotated UML model
E
F
BRIDG model release version X
D
BRIDG model
18
Tools to Support the People Process
  • www.BRIDGproject.org
  • Open source
  • Used for projects like firefox and mozilla
  • Bug and feature tracking
  • Discussion forums
  • Source control with CVS check-in/check-out
  • Project administration
  • Email and listservs with notification of code,
    discussion group, or document changes
  • Ability to link and synchronize CVS repositories
    across sites

19
Working with VCDE
  • The need for coded concepts and terminology
    within BRIDG
  • Can show some examples of these in the model
  • Semantic harmonization at the analysis level
  • The difficulty of mapping ODM with RIM at the
    implementation level (CDISC experience)
  • Integration at the implementation level requires
    terminology and vocabulary. However, to get
    there, you need to have shared semantics at the
    analysis level.
  • If the semantics are different at the analysis
    level, then they will map into different CDEs and
    terminologies at the implementation level.

20
Two places that we need VCDE
  • Input into the model
  • make it easier to be consistent with existing
    standards (ie, HL7 datatypes?)
  • Output from the model
  • (into semantic connector, UML loader, HL7
    message specifications, etc)
  • How do we construct good models that capture the
    domain expertise and can be used to drive
    application and message interchange development?

21
Discussion questions
  • How do we work with VCDE ?
  • Identification of best of breed coded concepts
    (SNOMED, etc)
  • Development of new enumerated lists and code sets
  • Setting Priorities for development?
  • Best practice for model and terminology
    development
  • How do we work with groups outside of caBIG?
  • HL7 defined value sets? caBIG defined value sets?
    Others?
  • Input into the model
  • How do we develop harmonizable models?
  • Output from the model
  • How do analysis models flow into caBIG tools?
  • Constrained models

22
Douglas B. Fridsma fridsma_at_cbmi.pitt.edu www.BRIDG
project.org
23
(No Transcript)
24
The goal Constructing Harmonizable Models
  • Semantically robust static and dynamic
    representations of the problem to be solved
    described in an implementation-independent way.

25
Static Portions of the Model
  • UML class diagram
  • Documentation of each class and all attributes
  • Bound to data types
  • Candidate terminologies identified
  • Other attribute value sets
  • OCL business rules (could be structured text with
    OCL equivalence) For example, person class and
    credit card class the person class might have a
    gender (MF) and the credit card class would have
    Mr/Mrs, there needs to be a rule that helps to
    disambiguate how the credit card class would
    manage this.

26
Dynamic Portions of the Model
  • Process flow
  • story boards
  • Scenarios
  • Use cases
  • Text UML activity diagrams
  • Links to static structures
  • Interaction diagrams (?)
  • Sequence diagrams
  • Collaboration diagrams (UML 2.0)

27
Examples
  • Model A with entities and roles combined
  • Model B with separation across moods
  • Study as higher-level concept
  • Protocol as higher-level concept

28
(No Transcript)
29
Interchange Standards Long-term Desired
Outcomes
  • A holistic approach to standards, facilitating
    data interchange from sites through regulatory
    submission, utilizing XML
  • Standards for data acquisition supporting the
    population of a cross-trial warehouse within FDA
  • HL7-CDISC models harmonized to yield value for
    both clinical research and healthcare sharing
    of information between EMR and clinical trials
  • Global adoption of CDISC data standards

CDISC Meeting with FDA Commissioner, April 2003
30
Biomedical Research Integrated Domain Group
(BRIDG) Model
  • A domain analysis model that reflects the domain
    space of clinical/biomedical research using
    language domain experts comprehend
  • Follows the HL7 Development Framework (HDF) led
    by HL7 RIM expert
  • Initiated in January 2004 by CDISC Board
  • Developed through numerous modeling and vetting
    sessions with domain experts from CDISC, FDA, and
    NCI/NIH
  • Is being adopted by the HL7 Regulated Clinical
    Research Information Management (RCRIM) Technical
    Committee and implemented at NCI
  • Now posted on an open website
    www.bridgproject.org
  • A means of bridging together the clinical
    research standards within CDISC and between CDISC
    and HL7 (towards interoperability)
  • Has also bridged organizations

31
CDISC and HL7Why BRIDG?
Domain Anaysis Model or Problem Space Model (a la
HDF)
Level of Abstraction
RIM
ODM
32
The BRIDG Model
  • What is it? A collaborative effort between CDISC,
    HL7, caBIGTM, NCI, FDA, and Industry to produce
    one common, shared data exchange standard
  • Vision Create a domain analysis model for the
    clinical research domain to harmonize clinical
    research standards among each other and to
    harmonize standards between clinical research and
    healthcare
  • Key Goal Define a structured computable protocol
    representation that supports the entire
    life-cycle of clinical trials protocol to achieve
    syntactic and semantic interoperability

33
Why BRIDG?
  • Reduce costs eliminate duplication of effort
  • Facilitate data sharing
  • Help to speed delivery of innovative solutions to
    the patient based on research
  • Improve the efficiency and timeliness of data
    reporting
  • Enhance patient safety during clinical trials
  • Improve the shared care of patients

34
Protocol and the Semiotic Triangle
Symbol Protocol
Source John Speakman/Charlie Mead
35
The Communication PyramidCommunicating
complexity is HARD
  • How do you link the people who know about
    clinical domain (domain experts) with the
    technology implementers who will develop tools to
    support clinical work so that we have semantic
    interoperability?
  • The Problem Space (Domain Space)
  • A structured way to describe and document the
    information requirements of a particular area of
    interest
  • What
  • Discovered through Analysis
  • The Solution Space (Implementation/Technical
    Space)
  • How
  • Discovered through Design
  • One Problem ? Many Solutions
  • The domain analysis model BRIDG is a bridge
    between domain experts and technical experts

36
The Communication Pyramid

Domain Experts
Technical Experts
Communication
Source Charlie Mead, MD, HL7
37
The BRIDG model update How do I get involved?
Douglas B. Fridsma, MD, PhD Center for Biomedical
Informatics Center for Pathology and Oncology
Informatics University of Pittsburgh
38
Update on activities since last meeting
  • New URL for the BRIDG project
  • www.BRIDGproject.org
  • Active participation from CDISC, pharmaceutical
    companies (GSK, Lilly, others), and industry
    (Fasttrack, Digital Infuzion, Oracle), caBIG
    (COH, UPMC, MSKCC, and others)
  • Committing real resources and time to the effort
  • A true collaboration with the various
    stakeholders in the effort

39
Technology
  • Still working on access control, roles, and
    supplemental tools to assist in the development
    effort
  • CVS repository open up the process to other SIG
    and project leads, while protecting the trunk
    of the CVS tree
  • CVS_ACL vs subversion vs. other
  • Parallel HTML pages to augment those in gForge
    for individual subprojects
  • Sig and project maintained
  • PHP and wiki features to integrate into gForge
    for collaborative authoring of the model
  • Novel methods of disseminating the model
  • Trackers, Web pages, and downloads
  • Supporting subproject development with mailing
    lists, document management, and CVS access (TDM,
    eDCI)

40
Processes
  • caBIG testing subproject coordination and
    harmonization processes with eDCI, and TDM. Will
    harmonize caERS when it is ready.
  • HL7 multiple proposals about how to use BRIDG
    for message specification.
  • Harmonization of existing message specifications
  • Model management project at HL7
  • Development of new messages
  • CDISC harmonization of glossary with vocabulary
    and definitions in the model

41
Ways to participate
  • HTML view of the model
  • www.bridgproject.org/model
  • Updated ever 110-14 days represents the current
    active model
  • Trackers
  • Modeled after HL7 Harmonization forms
  • Useful for smaller comments, definitions,
    suggestions
  • Download the model from the CVS repository or
    file release
  • file release
  • CVS repository
  • DBMS implementation (pending and testing)
  • Active model development in subprojects

42
How do I create an application based on BRIDG?
  • We have checked in the original version of the
    BRIDG model to the CVS trunk, and created a
    released that version of the model.
  • The trunk represents the released version of
    the model, and may be tied to requirements
    traceability.

Version 1
BRIDG trunk
43
How do I create an application based on BRIDG?
  • The Harmonization branch is similar to a bug
    fix it allows developers to break the model,
    fix things that need to be changed, and maintain
    versioning control while they do it.

Version 1
BRIDG trunk
Version 1
V 1.1
Harmonization branch
44
(No Transcript)
45
(No Transcript)
46
(No Transcript)
47
(No Transcript)
48
How do I create an application based on BRIDG?
  • When a team begins a new project to develop a
    message or application, they would check out the
    most recent version of the model that is
    currently developed. Typically this is the
    Harmonization branch.

Version 1
BRIDG trunk
Version 1
V 1.1
Harmonization branch
eDCI development branch
eDCI 1.1
49
How do I create an application based on BRIDG?
  • New projects will begin in the same way
    checking out the current version of the model
    from the harmonization branch
  • We now have three versions of the model, all
    working in parallel.
  • Although EA helps with synchronizing models,
    there is no substitute for good coordination and
    collaboration.

Version 1
BRIDG trunk
Version 1
V 1.1
V 1.2
Harmonization branch
CAERs development branch
caERs 1.1
eDCI development branch
eDCI 1.1
eDCI 1.2
50
(No Transcript)
51
Support for subprojects
  • Developing php-based webpages that can be
    maintained and customized by subproject leads
  • Limitation in gForge with project and subproject
    descriptions
  • Developing less error-prone CVS access controls
  • check-in to CVS grants root access to the entire
    CVS tree
  • Subproject leads need flexibility in organizing
    the CVS tree to fit their needs
  • CVS root needs to be protected from inadvertent
    check-in
  • Parallel constructs with documents, webpages, cvs
    repository, etc
  • TROVE map is too high-level, but need an
    ontology to help organize and find information
    on the gForge site.
  • Improved search capabilities?

52
How do I create an application based on BRIDG?
  • Development can continue on both of these
    branches, using the master-replica features of
    Enterprise Architect

Version 1
BRIDG trunk
Version 1
V 1.1
V 1.2
Harmonization branch
CAERs development branch
eDCI development branch
eDCI 1.1
eDCI 1.2
53
(No Transcript)
54
How do I create an application based on BRIDG?
  • New projects will repeat the process of checkout,
    modeling, etc.
  • As the development continues, there will be
    periodic new releases of the model.
  • The arrow from version 1 to version two
    represents some vetting of the model from
    stakeholders in the project and incorporation of
    harmonization branch changes.

Version 1
Version 2
BRIDG trunk
Version 1
V 1.1
V 1.2
V 1.3
Harmonization branch
CAERs development branch
caERs 1.1
caERs 1.2
eDCI development branch
eDCI 1.1
eDCI 1.2
eDCI 1.3
55
(No Transcript)
56
How do I create an application based on BRIDG?
  • As a development branch matures, it can be merged
    with the harmonization branch (using
    master-replicas)
  • The model then can be vetted with the
    stakeholders, and a release generated

Version 2
Version 3
BRIDG trunk
V 1.3
V 1.3
V 1.3
Harmonization branch
CAERs development branch
caERs 1.2
caERs 1.3
caERs 1.4
eDCI development branch
eDCI 1.3
eDCI 1.2
eDCI 1.3
57
HL7 Model management processes for vetting
58
How do I create an application based on BRIDG?
Version 3
BRIDG trunk
59
How do I create an application based on BRIDG?
  • A portion of the BRIDG model will need to be
    annotated and modified for the constraints of the
    semantic connector and UML loader
  • Future releases may accept a broader range of
    relationships

Version 3
BRIDG trunk
Annotated model
60
How do I create an application based on BRIDG?
  • The output of the processes related to the
    semantic connector and UML loader will be used
    for application development

Version 3
BRIDG trunk
Annotated model
Semantic Connector
UML Loader
caCORE tools
Application
Application Module
61
Other issues
  • GREATLY simplified, but a lot of people are
    working on the processes
  • NCICB with the processes around the UML loader
    and semantic connector
  • HL7 with regard to using the model for message
    development
  • BRIDG group working on processes that touch on
    BRIDG
  • Input/outputs
  • Versioning
  • Harmonization
  • Collaboration and coordination
  • Adopters/developers trying to figure out what it
    means for them

62
Technology and Processes
  • Technology and process go hand-in-hand
  • Features in EA for synchronization
  • Features of CVS repository and gForge
  • Vertical vs Horizontal project integration
  • CVS vs subversion and branch-level access
    controls
  • The best way to move forward is to start seeing
    what works, (and what doesnt) and have the
    flexibility to adapt to new requirements to the
    process and technology

63
(No Transcript)
64
(No Transcript)
65
(No Transcript)
66
(No Transcript)
67
(No Transcript)
68
Current ongoing project
  • eDCI message
  • Participants Oracle, caBIG, CDISC and others
  • developing an HL7 message that describes
    electronic data collection instruments (case
    report forms)
  • Clinical Trial Registration message project
  • Participants caBIG, CDISC, HL7, Eudract,
    clinicaltrials.gov, PDQ
  • Developing a clinical trial registration message
  • Adverse events application development
  • Participants caBIG, COH
  • Developing an adverse events reporting system
  • Trial Design modeling
  • Participants caBIG, CDISC, FastTrack, others
  • Extending the trial design portions of the model
    in BRIDG
  • Statistical modeling extensions
  • Participants CDISC, Eli Lilly, other pharma,
    FastTrack
  • Extending the statistical modeling portions of
    the model

69
Clinical Research Back Office The motivation for
integration
6
1
5
2
3
4
Write a Comment
User Comments (0)
About PowerShow.com