Creating a Clinical Data Element Dictionary A Proposal

1
Creating a Clinical Data Element DictionaryA
Proposal

• CDISC Board of Directors Meeting
• 10-11 Dec 2007

2
Preamble

• CDISC has made progress on many fronts
• There is a CDISC brand
• CDISC has worked on strategies/plans over the
  years
• Currently a strategy in place
• Operational plans/objectives for 2008 in place
• 2008 budget is in place
• Fundamentally change some of CDISCs approach

3
Preamble

• This is a discussion first and foremost about
  WHAT we will do.
• What one thing, if done well and consistently,
  would have the most impact on your business?
• Ken Blanchard, Mission Possible
• What is the pearl of great price.
• If we agree on WHAT, then we can discuss HOW.

4
Motivation for the WHAT

• What standards work is done today Lilly example
• Lilly Data Element Standards
• \\Rodan\rodan.grp\GCDS_EB_PUBLIC
• 3560 pages in our Dictionary
• 25,000 variables
• Its all pdf (yuk !!!!)

5
Motivation for WHAT

• Link to Analysis Dataset Standards
• http//corpweb.d51.lilly.com/statmath/CoE/ADS/ADS_
  std.html
• Thousands of pages of documentation in our total
  ADS specifications
• For each study, CROs get hundreds of pages of
  requirements that describe the data elements that
  we want, variable names, valid values, formats,
  etc.

6
Motivation for WHAT

• Dozens of people at Lilly and CROs communicate
  using these voluminous documents
• CROs have dozens of people mapping data to
  company-specific formats, naming conventions,
  etc.
• The CDISC Business Case largely predicated on
  eliminating these activities
• Reduce mapping data from one form to another to
  transfer or to integrate it.

7
Motivation for WHAT

• http//www.wikihit.org/wiki/index.php/Main_Page
• The Clinical Data Definitions created in WikiHIT
  are not completely useful for clinical research
  studies.
• caDSR has some useful elements, but is a bit
  outdated and not entirely functional for what is
  needed for clinical research studies.
• Too complex in all its detail
• NCI EVS has some useful elements, but does not
  have all the information and functionality that
  is needed by companies involved in clinical
  research.
• Data definitions do not have all information
  (e.g. valid values)

8
The Language of Clinical Trials

• It is more important to share a common vocabulary
  than it is to have agreement on common
  grammatical rules.
• ? Content is more important than structure.
• Es ist wichtiger, einen allgemeinen Wortschatz,
  als zu teilen es Vereinbarung über allgemeine
  grammatische Richtlinien haben soll.
• A common vocabulary more important for sharing
  than understanding of typical rules of grammar.

9
CDISC Adoption by Pharma

• Make SDTM more useful, implementable
• Need more specificity
• Need more definitions on variables data
  elements
• Standard data elements this is what FDA wants,
  what pharma wants, what CROs want
• FDA under pressure to do something quickly?
• CDISC dealing with healthcare, other standards
  organizations, etc.
• Dont let the perfect hold up the good
• Need more focus from CDISC

10
CDISC Adoption by Pharma

• Its about saving dollars for pharma, CROs and
  labs by simplifying interchange of data.
• Its about helping companies and FDA integrate
  data from regulated clinical research studies.
• CDISC Business case has little to do with
  healthcare at this point.

11
Motivation for WHAT

• Summary
• There is an enormous unmet need for more content.
• The CDISC Business Case is largely dependent on
  well defined data element standards being broadly
  available.
• Others are playing in this space, but do not meet
  the needs of pharma clinical research and
  regulatory submissions.
• CDISC Terminology Program has primarily focused
  on controlled terminology supporting SDTM, but
  not the data elements themselves.

12
What Is a Data Element?

• All the pieces of information (i.e. metadata)
  needed to unambiguously describe a concept
• English dictionary analogy
• Word desk
• Phonetic spelling desk
• Part of speech noun
• Definition a piece of furniture with a flat top
  for writing
• could also be thought of as the concept
• Source Latin, discus
• etc.

13
Data Element

• A Data Element is a unit of data for which
  definition, identification, representation, and
  permissible values are specified by means of a
  set of attributes the smallest unit of data.
• The purpose of a data element definition is to
  define a data element with words or phrases that
  describe, explain, or make definite and clear its
  meaning.

14
Data Elements Vertical v. Horizontal
Vertical Data Set Structure

• Valid Values for Variable are
• HR, SBP, DBP.
• A controlled terminology

• For each term, provide the metadata to describe
  it
• Definition, units, valid values, etc.

15
Data Elements Vertical v. Horizontal
Horizontal Data Set Structure
Each variable has a name (terminology) and a
corresponding set of metadata to describe it
(definitions, units, valid values, etc.)
16
Clinical Data Element for Pharma

• Variable name (draft)
• Label / concept
• Valid values of the variable itself
• Data type (num, char, date, )
• Units
• Key words (e.g. biomarker, osteoporosis, )
  facilitate searches
• Source / reference (as needed)
• SDTM data domain
• Regulatory requirement
• A team needs to define what are the essential
  metadata pieces of information that are
  parsimonious enough to eliminate ambiguity, but
  few enough to be useful, consumable,
  understandable, burdenless.

17
Creating a Clinical Data Element Dictionary (CDED)

• Task Force Members
• Steve Ruberg
• Bron Kisler
• Scott Getzin
• Doug Fridsma
• Chris Chute
• Sue Dubman
• Dave Iberson-Hurst
• Cara Willoughby

18
Proposal WHAT - Unmet Need

• Comprehensive, electronically accessible,
  organized dictionary of unambiguous data element
  standards for our industry
• One of the most fundamental problems we all face
  within our own pharma companies, but even more
  acutely across the pharma industry/enterprise.
• Consistent with Strategy Theme 2, 5, 6
• THE place where people go for clinical data
  element standards.
• THE thing for which CDISC is known ?!?!

19
Alignment and Focus

• If additional funding can be secured, standards
  specific to therapeutic areas will become part of
  the extended CDASH scope.
• CDISC Press Release 33
• 15 May 2007
• KEY QUESTION
• Given the importance of this area and the need to
  move quickly, should we re-prioritize and divert
  resources (people and ) to this effort?

20
Alignment and Focus

• FOCUS
• Where do we focus?
• ISO, AHIC, AHRQ, NLMEc, industry architecture,
• Initial focus on meeting pharma industry needs
• If others want to piggyback on that effort, that
  is fine.
• Initial focus on clinical data and clinical trial
  metadata
• Initial focus on raw/observed data
• There is a lot of territory to conquer within
  this focus area. Other opportunities
  (pre-clinical data elements, derived data
  elements) can be explored in the future.

21
Impact on Other CDISC Teams

• Clinical Data Element Dictionary (CDED)
• Terminology, SDTM, CDASH, LAB and SEND all
  converge into a common approach focused on the
  data elements and their exquisite definition
• Reduces need to harmonize CDISC models if they
  all utilize the same data element definitions
• Harmonization happens on the front end rather
  than after the fact
• The transport standard for carrying standardized
  content (ODM, HL7, SAS, other???) can be whatever
• BRIDG work continues as is
• Tightly coordinate standard data elements with
  BRIDG efforts

22
Creating a Clinical Data Element Dictionary (CDED)
Content Standards
Transport Standards
Initial Inputs
CDASH
SDTM
SAS
CDED
80
LAB
20
ODM
Protocol
TB?
HL7
CV?
Other Existing?
23
Proposal

• HOW - Business model
• An open, electronic, peer production environment
  with appropriate governance
• Like MedDRA, but open and free
• Like Wikipedia, but more governance
• Like LINUX, but more granular and dynamic
• CDISC must adopt a more flexible and rapid
  development process

24
Clinical Data Element Standards
Governance
Final
Review
Template
Submission
Anyone

• Downloadable (define.xml)
• Searchable text, key words
• search shows status (submit, review, final)

25
Governance for the CDED
Governing Board
2 Full-Time CDISC Employees
Lead 1
Lead 2
Lead 3 ... Lead k
Team 1
Team 2
Team 3
Team k
Lead 1
Lead 2
Lead 3
Lead k
. . .
6-8 SMEs
6-8 SMEs
6-8 SMEs
6-8 SMEs
26
Proposal

• WHO - CDISC
• CDISC has the opportunity to assert an even
  greater leadership role in this arena.
• Leverage CDISCs strengths Strategy Theme 1
• Independence
• Consensus building
• Strong pharmaceutical / clinical research
  expertise
• Global recognition
• Place substantial priority and focus on this
  effort
• The pearl of great price

27
Proposal

• WHEN - ASAP
• The time is right to charge ahead aggressively
• There is a large, unmet business need
• FDA and others are looking for a content
  leader
• CDISC has ongoing terminology efforts
• Technology is in place (i.e. wikis)
• Mindset is in place (i.e. people can work
  virtually)
• Others are advancing on this front and we may be
  left out

28
Budget

• Transition personnel to this effort
• Continue/finalize ongoing CDASH efforts
• Redirect some Terminology Team efforts
• Need part-time Governance team members
• Contracted for 25 of their time
• SMEs for TA or data domains
• Leverage CROs, software members of CDISC

29
Summary

• There remains a clear need to have unambiguous
  clinical data element standards (CDES)
• Considerable efforts still spent on exchanging
  data
• Considerable efforts still spent on integrating
  data
• Needed across the drug development industry
• Broad set of data domains (safety, efficacy,
  outcomes, PK, etc.)
• Independent of strategies related to messaging or
  transport technologies
• Lets act decisively and move quickly.

30
Benefits of Using Documented CDEs

• Facilitates common data collection by defining
  content and scope.
• Supports semantic data relationships.
• Defines valid values for enumerated data.
• Improves understanding of data.
• Simplifies and documents data analysis.
• Provides historical context for data collections.
• Encourages reuse of existing data structures.
• Facilitates sharing of data across organizational
  entities.
• Facilitates integration of data across studies.