caBIG Compatibility Evaluation

1
caBIG Compatibility Evaluation MSKCC Clinical
Research Database (CRDB)

• CTMS caBIG Compatibility SIG
• 29-30 August 2005
• John Speakman

2
CRDB

• Relational database (Oracle)
• Development began 1992 has supported 100 of
  MSKCC clinical trials since about 1995
• Fully coded shares common data elements (i.e.,
  common within MSKCC) between protocols
• Broad consensus obtained on coding conventions
  within MSKCCs research community (a huge effort
  at the time)

3
CRDB Activity 2000-2004
2000 2001 2002 2003 2004 Active
IRB Protocols 391 400 384
428 437 New IRB Protocols 131 153
133 160 147 New Patient Registrations
9458 12,904 15,298 13,604
13,198 Institutional Trials 25.1 National
Group Trials 14.1 Industrial
Trials 32.3 Other Externally Peer Reviewed
Trials 28.5

4
CRDB Patient Count
365,972 distinct patients in CRDB as of
8/17/2005 MSKCC Patients 276,525
(89.2) Non-MSKCC Patients 33,447 (10.8)
gt94,000 pt records in gt3,400 IRB
protocols gt270,000 pt records in gt240
prospective databases

5
CRDB Functionality 1

• Metadata Management
• Data Dictionary Management
• User Identity Management
• Vocabulary Management
• Access Control Management
• User Authentication
• User Authorization
• Protocol Setup
• Patient Registration

6
CRDB Functionality 2

• Research Data Collection
• Standard Clinical Research Evaluation
• Generic (EAV) Questionnaire System
• Tissue Bank / Specimen Tracking System
• Cell Marker LIMS
• AICT LIMS
• Reporting
• Ad-Hoc Query Interface
• Pre-Formatted Push-Button Reports
• Data Submission Export
• Source Data Transfer

7
Three-Tier Architecture

• Oracle Forms Java Client
• set of generic Java classes rendered as a Java
  applet
• downloaded to client the first time a user
  requests a Forms application
• cached permanently in the users browser,
• executed in the browsers JVM
• Oracle (proprietary) middle tier
• Oracle Forms and Reports Services
• Allows Oracle Forms and Reports components
  developed for client/server deployment to be run
  on the web
• Database Tier Oracle 9iRAC EE

8
Oracle Architecture for CRDB
9
CRDB Data Sharing (via custom interfaces)
CRDB
CRMIS
Protocol definition information
Tissue consent information
SMS
CRMIS
Demographic information repository
Correction requests
LCS
Laboratory information repository
Pharmacy (reference only)
Prococol-patient accrual information
Disease staging information
DMS
Industrial Sponsors
OR
Surgery information
NCI CTEP
Phar- macy via IDB
Treatment information repository
Study data
Theradex
10
Evaluation Outcomes

• CRDB is caBIG compatible at Bronze level across
  the board
• Rigorous, unambiguous data definitions, but
• Most of its interfaces and vocabularies are
  non-standard, devised prior to agreement on many
  of todays standards
• Proprietary middleware
• All its modules communicate with each other via
  the database

11
Report Card Bronze

• Programming and Messaging Interfaces
• SQL is only means of programmatic access
• There is a custom HL7 2.x interface for lab
• Vocabularies/Terminologies Ontologies
• Well-defined but not standard
• Not reviewed by VCDE Workspace
• Data Elements
• Good definitional depth
• Not reviewed by VCDE Workspace, not in caDSR
• Information models
• ER diagrams
• Enterprise Architect can reverse engineer class
  diagrams

12
Reverse Engineered class diagram (simplified)
13
Gap Analysis / Getting to Silver

• Construct a CRDB object model
• Interested in NCIs CTOM reference implementation
  of BRIDG
• Base an XML import/export API on the object model
• Enter the CRDBs data elements into the caDSR
• Review our vocabularies
• Adopt standard ones, or map using caDSR

14
(No Transcript)
15
(No Transcript)
16
Issues

• The more functionality you have, the harder it is
  to be interoperable
• Herculean task of entering our data elements into
  caDSR
• Herculean task for someone to review them!
• How do legacy systems get there from here?
• Bay Bridge or Golden Gate Bridge?

17
Thats it!

• Questions?