Heading - Digital Infuzion, Inc.
Download
Report
Transcript Heading - Digital Infuzion, Inc.
An exploration of quality gaps in SDTM
implementation activities and ideas on how to
address these gaps through appropriate resourcing
Dianne Weatherall: 2013-04-11
GOAL
• Adoption of CDISC standards has led to:
– new processes (aCRF, metadata, programming)
– new responsibilities
• Goal: to discuss the “best” SDTM team to
implement the new process
DEFINE THE PROBLEM
• Define what is wrong with the current setup
ROOT CAUSE OF QUALITY ISSUES
• Poll on the SDTM LinkedIn group:
What is the primary cause of quality issues in SDTM?
2%
2%
4%
4%
Votes
Not understanding SDTM
Not understanding clinical data
43%
18%
Non-standard data
Not enough time/resource
Inadequate review
27%
Protocol design
Company standards
ROOT CAUSE OF QUALITY ISSUES
• Lack of understanding of SDTM – WHY?
1
2
3
Lack of skills
Lack of training
Expensive
It is complicated
Too much room for personal
preference
Customer specific
implementations
Unclear process
Too many cross-functional
teams involved
Companies are too silo’ed
Lack of expert support
Lack of skills
It takes time and effort to
become an expert
ROOT CAUSE OF QUALITY ISSUES
• Lack of understanding of clinical data – WHY?
1
2
3
Lack of data / clinical skills of teams
Clinical data is complicated
Highly un-normalized
Database structures are often
developed for data entry and
clinician preference, not
CDASH/SDTM standards
The CRF changes over time
Therapeutic areas have complicated
study designs (e.g. cohort changes)
Poor planning from the study design
stage
ROOT CAUSE OF QUALITY ISSUES
• Non-standard data – WHY?
1
2
Legacy studies
SDTM standards are relatively
recent
Customer specific requirements
Customer specific
implementations
Therapeutic areas have
complicated study designs (e.g.
cohort changes)
Poor planning from the study
design stage
3
Customer legacy systems are not
CDISC-compatible
SUMMARY OF ROOT CAUSES
•
•
•
•
•
•
•
Company silo’s
Lack of data skills of Biostats teams
Lack of CDASH / SDTM skills of Data teams
Time and effort to build expertise
Customer-specific
Poor study planning
Expensive - join a user group!
CRITERIA FOR THE BEST SDTM TEAM
• Corporate structure
• Team scenarios
CORPORATE STRUCTURE
Operations
Operations
Biometrics
Biometrics
Operations
Data
Management
Data
Management
Biostatistics
*** Blur the line between DM and BIOS
Biostatistics
BEST TEAM SCENARIO
SDTM experts?
Programmers?
Biostatisticians?
Data genius?
Cheap?
Available?
ROLES AND SKILLS
Data collection:
CRF design
(CDASH / SDTM
experts)
SDTM
mapping: CRF
annotation/
specifications
(SDTM experts)
Programming:
(SAS experts)
Review:
(SDTM +
Biostatistics +
Data experts)
Data Management ----------------------------------------Biostatistics
TEAM SCENARIO 1
Study
CRF / DB Design
aCRF
Specs
Programming
1
A
B1 (domain A)
B2 (domain B)
Etc
Reviewer R2
B1
B2
B1
B2
Reviewer R2
Reviewer R2
Reviewer R1
Advantages
Disadvantages
Continuity from aCRF
Time consuming
SME on certain domains
Pressure (updates)
Small team
Inconsistent mapping across domains
Boring
Lack of continuity from CRF design
Inconsistent metadata
TEAM SCENARIO 2
Study
CRF / DB Design
aCRF
Specs
Programming
1
A
B (all domains)
Reviewer R1
Reviewer R2
C1
C2
Reviewer R2
C1
C2
Reviewer R2
Advantages
Disadvantages
Consistent mapping across
domains
Time consuming
Pressure (updates)
Lack of continuity from CRF design
Inconsistent metadata
TEAM SCENARIO 3
Study
CRF / DB Design
aCRF
Specs
Programming
1
A
Team B
(all domains)
Reviewer R2
Team B
(all domains)
Reviewer R2
C1
C2
Reviewer R2
Reviewer R1
Advantages
Disadvantages
Consistent mapping across
domains
Resourcing issue (more people
needed), particularly for a submission
Less advanced tasks can be
done by cheaper resources,
freeing advanced
programmers for critical tasks
Lack of continuity from CRF design
Expert group for support
TEAM SCENARIO 4
Study
CRF / DB Design
aCRF
Specs
Programming
1
A
Team B
(all domains)
Reviewer Team
Team B
(all domains)
Reviewer Team
C1
C2
Reviewer Team
Reviewer Team
Advantages
Disadvantages
Consistent mapping across
domains
Difficult to develop/find reviewer skills
(for design and mapping)
Less advanced tasks can be
done by cheaper resources,
freeing advanced
programmers for critical tasks
Expert group for support
Continuity of whole process
Other things to consider
• Submissions
– Continuity across studies
– Consistency across studies
– Change control
– Bottle necks (reviewer team)
• ADaM / statistical output resourcing
RESOURCE CRITERIA
FROM
• SDTM after design
• Just save costs
• Allocate availability
TO
• SDTM during design
• Invest in expert team
• Look at continuity
THE BEST SDTM TEAM
Expert Reviewer
Team
(CDASH / SDTM)
Team Leader
(Biometrics)
CRF Designer
aCRF / SDTM
Mapper Team
*** Understand the data
*** Understand the purpose
Programmer(s)
?????????????????????????????????????
QUESTIONS