How to improve quality control in a data conversion process?

Transcription

How to improve quality control in a data conversion process?
How to improve quality control in
a data conversion process?
By extended usage of metadata!
Dimitri Kutsenko
Entimo AG - Berlin/Germany
Data Conversion Process
SDTM Example
Generic view
SOURCE
ALGORITHMS
TARGET
CHECKS
PROCESS
Define
Dataset
Structure
Core process
Annotate
CRF
Define
Mapping
© Entimo AG | Stralauer Platz 33-34 | 10243 Berlin | www.entimo.com
Generate
and Run
Mapping
Program
Perform
SDTM
Checks
19-Oct-2010
Generate
Define
2
Cost of Change Curve
Early QC
Paradigm
QC Tasks
© Entimo AG | Stralauer Platz 33-34 | 10243 Berlin | www.entimo.com
19-Oct-2010
3
Maintain Standards Library
§  Invest effort into standards definition!
§  Metadata Types
Maintain Standards
Library
§  Dataset descriptions
(industry or company domains)
§  Conversion algorithms (text and code)
§  Mappings
§  Terminology…
§  Challenges
§ 
§ 
§ 
§ 
Global, project, trial and study levels
Multiple standards / versions
Interlinked dimensions
Reusability
Define Dataset
Structure
Annotate CRF
Define Mapping
Generate and Run
Mapping Program
Perform SDTM
Checks
Generate Define.xml
Analysis of cross-level and cross-study dependencies required!
© Entimo AG | Stralauer Platz 33-34 | 10243 Berlin | www.entimo.com
19-Oct-2010
4
Standards
Metadata Profiling
§  Compare metadata domain
definitions
§  Detect and review discrepancies
as early as possible!
Data Profiling
Metadata
Metadata
AE Domain
(Global)
AE Domain
(Study)
© Entimo AG | Stralauer Platz 33-34 | 10243 Berlin | www.entimo.com
46
1
1.77 24.5
44
0
1.62 30.2
46
0
1.66 33.1
62
1
1.80 21.9
19-Oct-2010
5
Standards
Impact Analysis
§  Track deviations between
(changed) standards and related targets!
§  Analysis/update shall include impact on datasets
linked to standards
§  Versioning/audit trail required for updates
Metadata
Common
(Global)
Metadata
AE Domain
(Study)
© Entimo AG | Stralauer Platz 33-34 | 10243 Berlin | www.entimo.com
Data
AE
Dataset
19-Oct-2010
6
Define Metametadata
Define Dataset Structures
§  Define Metametadata
(description of dataset metadata):
§  Columns
§  Formats to support checks
Maintain Standards
Library
Define Dataset
Structure
§  Configuration rules for metadata
Annotate CRF
§  Rule examples:
CDISC type – character, mandatory
seq – integer, unique, starts with 1
§  Define Dataset Structures:
Define Mapping
Generate and Run
Mapping Program
Perform SDTM
Checks
§  Use metadata domain templates
§  Derive from datasets, metametadata
© Entimo AG | Stralauer Platz 33-34 | 10243 Berlin | www.entimo.com
Define
Metametadata
19-Oct-2010
Generate Define.xml
7
Define Mapping
§  Challenges:
Define
Metametadata
§  Repeated algorithms
§  Scarce experts
§  Redundancy
Maintain Standards
Library
§  Exploit standards at maximum!
§  Standard metadata (domains)
§  Standard conversion algorithms
§  Standard mappings
Define Dataset
Structure
Create aCRF
Define Mapping
Generate and Run
Mapping Program
§  Quality checks in the mapping
Perform SDTM
Checks
definition:
§  Consistency checks
§  Review can be done
Generate Define.xml
language-independent
© Entimo AG | Stralauer Platz 33-34 | 10243 Berlin | www.entimo.com
19-Oct-2010
8
Perform SDTM Checks
§  Integrate checks into conversion
workflow! (incl. tracing)
§  Flexible definition of check criteria
required:
§  Standard SDTM conformance checks
§  Customer checks
Source Data
Source
Checks
Executable
Program
SDTM Data
Maintain Standards
Library
Define Dataset
Structure
Create aCRF
Define Mapping
§  Checks from the 1st domain on!
Mapping
Program
Define
Metametadata
Generate and Run
Mapping Program
Perform SDTM
Checks
Target
Checks
Generate Define.xml
Check
Report
© Entimo AG | Stralauer Platz 33-34 | 10243 Berlin | www.entimo.com
Check
Report
19-Oct-2010
9
Generate Define
Structural Checks
§  Metadata driven,
Metadata Driven Process
template based
definition
§  Check of data vs.
metadata prior to:
TARGET
METADATA
Mapping
Specs
§  Import/production
Mapping
Programs
Target Data
§  Define creation
Define
(xml)
Structure
Check
Define
Metametadata
Maintain Standards
Library
Conservative Process
TARGET
METADATA
Mapping
Specs
Mapping
Programs
Target Data
Define Dataset
Structure
Create aCRF
Define
(xml)
Define Mapping
Generate and Run
Mapping Program
Perform SDTM
Checks
Generate Define.xml
© Entimo AG | Stralauer Platz 33-34 | 10243 Berlin | www.entimo.com
19-Oct-2010
10
Metadata Challenges
§  Model related challenges:
§  Metadata variation vs. metadata inflation
§  Interlinked dimensions
§  Automatic collection and processing
§  Organization related challenges:
§  Effort for standard development
§  Consistent metadata management
§  Model independence
§  Different business processes
§  Regulatory requirements
© Entimo AG | Stralauer Platz 33-34 | 10243 Berlin | www.entimo.com
19-Oct-2010
11
Metadata Based Data Conversion
Process
§ 
Metadata QC leveraged by smart
tools/systems help you:
§ 
§ 
§ 
§ 
§ 
Increase data quality and consistency
Reduce cost of errors
Flexibly support available and future standards
Increase reusability of process components
Set up controlled and traceable process
© Entimo AG | Stralauer Platz 33-34 | 10243 Berlin | www.entimo.com
19-Oct-2010
12
END
Many thanks for your attention!
Questions…?
VISION STARTS NOW!
Visit at entimo s booth or email to
dku[at]entimo.de
© Entimo AG | Stralauer Platz 33-34 | 10243 Berlin | www.entimo.com
19-Oct-2010
13