A Comprehensive Clinical Research Database based on CDISC ODM and - - PDF document

A Comprehensive Clinical Research Database based on CDISC ODM and i2b2

• F. Meineke, S. Stäubert. M. Loebe, A. Winter

MIE 2014 Istanbul 4.9.2014

COREUNIT DATACENTRE OF THE

INTEGRATED RESEARCH AND TREATMENT CENTER ADIPOSITY DISEASES

• Support clinical trial conduct

– Project management – Biometry/Statistics – Data management

• Support care / outpatient clinic

– Enhance documentation data quality – Optimize documentation processes

• Clinical Informatics group

– Secure Infrastructure – Integrate obesity related research data – Research Database with services and tools for data retrieval

MIE$2014$/$F.$Meineke

Head%Prof.%Markus%Löffler

OBESITY / ADIPOSITY

• Definition

– too much body fat – ICD10 E66, WHO degree

• 1 BMI ≥ 30kg/m², 3: ≥ 40kg/m²

– one of the leading preventable death causes

• Treatment

– diet, excercise, medication, bariatric surgery

• Research examples

– see www.ifb-adipositas.de/en (Strength or

Endurance? Surgery methods? Economics? Depression? …)

3

Gastric(Band

IFB RESEARCH INFRASTRUCTURE

4

SAP=SAP%i.s.h.med=Hospital%Informa7on%System LIFE=Leipzig%Research%Centre%for%Civiliza7on%Diseases

MIE%2014 Frank%Meineke

METHODS

• IT-Landscape: Research IT-

Infrastructure

• Framework for Research

Database: Open Archival Information System

• Transport Data format:

Operational Data Model

• Process Data Handling: Curation

Lifecycle Model

• Supporting Tools: ETL,

MetaDataRepository, i2b2, …

• Access Model: Data Curation

Continuum Model (Andrew Treloar)

5 MIE%2014 Frank%Meineke

!

Preserva'on*Planning

Data Management Archival Storage

Access Ingest

PRODUCER CONSUMER

SIP

Descrip<ve Informa<on Descrip<ve Informa<on

AIP AIP

queries query!responses

• rders

DIP

Administra'on

METHODS OPEN ARCHIVAL STORAGE SYSTEM

ISO$14721:2012$Open$archival$informa8on$system

MIE!2014 Frank!Meineke 6

RESULTS OAIS BASED RESEARCH DATABASE

Storage Access Ingest

Tabular Data

ETL

ETL

Source2specific Proprietary

SQL

ETL

ODM

Ingest

Archival Storage

Access

MIE22014 Frank2Meineke 7

Research Data Base

ETL

Tabular Data

Where/is Metadata? n*jobs? Moving target?

• ne/response.

StaDc2/ Generic

For/each query…

RESULTS OAIS BASED RESEARCH DATABASE

Storage Access Ingest

Tabular Data Meta Data Import Control

I2b27for group71

ETL

ETL

SQL

ETL

Standard

Terminology

Sta@c7/ Generic SQL7for group72 ETL ODM7Pool

ODM

ODM ODM ODM

Ingest

Archival Storage

Access

MIE72014 Frank7Meineke 8

Source7specific Proprietary

METHODS: STORAGE LAYER OPERATIONAL DATA MODEL

• “is designed to facilitate the regulatory-compliant

acquisition, archive and interchange of the metadata and data for clinical research studies”

• “All of the information that needs to be shared among

different software systems during the study setup,

• peration, analysis, submission or for long term

retention as part of an archive is included”

• Hypothesis: structured HIS

Data (HIS, Lab, Biobank) can be mapped to CDISC ODM and technically

9

www.cdisc.org/odm HIS$e.g.$SAP ODM/Clinical$Trial Encounter StudyEvent Document/Form Form Group ItemGroup Data:column Item Coding CodeList

MIE%2014 Frank%Meineke

INGEST LAYER ETL FROM SOURCE TO ODM

• Data Preparation

– de-identification / anonymization – curation / cleaning (for non-trial data) – annotation (groups, metadata) – normalize (e.g. use of UCUM) – map structure (events/forms/…) – prepare Record Linkage (IDs)

• Tools

– Higly flexible, customizable generic importer Tabular!"#$ %&'() *+ ,&-(+)".(+/01)2* 3

! "#$ %&'&( conforming file

10

Digital'Cura+on'Center University'of'Edinburgh

MIE&2014 Frank&Meineke

ACCESS LAYER (1) RELATIONAL DATABASE / DATAMART

• SQL based
• Generic odm2sql ETL-job

Pros

• Fine grained access rights
• Well suited for bio-statisticians
• Support for score card

generation and reporting Cons

• SQL knowledge necessary
• Complex overarching joins

11

ODM Rel. DataBase Study Scheme StudyEvent (implicit) Form (implicit) ItemGroup Table Item Column CodeList Foreign Key?Table

MIE%2014 Frank%Meineke

ACCESS LAYER (2) DATA WAREHOUSE

• Star-Scheme / EAV
• Generic odm2i2b2 ETL-

job Pros

• Simple GUI (really…)
• cohort-searching,

feasibility tests

• data exploration, data

creativity

• overarching queries
• more use cases to come

Cons

12 MIE&2014 Frank&Meineke ODM I2b2 Study 1st&level StudyEvent (n.a.) Form 2nd ItemGroup 3rd Item 4th CodeList 5th Catalogs 5th&to&nE th

non#smoking#woman#from#18/30 BMI#greater#>#40 no#known#diabetes wri;en#consent#to#be#contacted Outpa@ent#visit#in#the#last#3#month # Answer:#pa@ent#count,#pa@ent#id#list

DISCUSSION

• Benefits of ODM centric approach

– Data is stored in standardized, stable, archivable, partly-self documenting, human readable files – Writing ETL jobs from targeting a stable file based format is much less complex then targeting an evolving proprietary database – Data quality is gained, as lack of metadata and curation deficits get visible during data conversion and data usage – The OAIS framework helps to build well defined RDB components

• Experience so far (after 9 months)

13 MIE&2014 Frank&Meineke

OUTLOOK

Now (Private Research Domain)

• Small research database, structural comprehensive
• HIS datasets imported
• Pilot-phase

Near Future (Shared Research Dom.)

• Integrate more obesity

related data including selected clinical trials

• General service for IFB

14 MIE&2014 Frank&Meineke

Andrew'Treloar,'Digital'Cura3on'Con3nuum

I&thank&you&for&your&a6en7on!

frank.meineke@imise.uni:leipzig.de