CC BY-NC-ND 4.0 · Methods Inf Med 2019; 58(02/03): 086-093
DOI: 10.1055/s-0039-1693685
Original Article
Georg Thieme Verlag KG Stuttgart · New York

A Generic Method and Implementation to Evaluate and Improve Data Quality in Distributed Research Networks

D. Juárez
1  Federated Information Systems, German Cancer Research Center (DKFZ), Heidelberg, Germany
2  German Cancer Consortium (DKTK), Heidelberg, Germany
,
E.E. Schmidt
1  Federated Information Systems, German Cancer Research Center (DKFZ), Heidelberg, Germany
2  German Cancer Consortium (DKTK), Heidelberg, Germany
,
S. Stahl-Toyota
3  Medical Informatics in Translational Oncology, German Cancer Research Center (DKFZ), Heidelberg, Germany
,
F. Ückert
3  Medical Informatics in Translational Oncology, German Cancer Research Center (DKFZ), Heidelberg, Germany
,
M. Lablans
1  Federated Information Systems, German Cancer Research Center (DKFZ), Heidelberg, Germany
2  German Cancer Consortium (DKTK), Heidelberg, Germany
› Author Affiliations
Further Information

Publication History

15 February 2019

07 June 2019

Publication Date:
12 September 2019 (online)

  

Abstract

Background With the increasing personalization of clinical therapies, translational research is evermore dependent on multisite research cooperations to obtain sufficient data and biomaterial. Distributed research networks rely on the availability of high-quality data stored in local databases operated by their member institutions. However, reusing data documented by independent health providers for the purpose of care, rather than research (“secondary use”), reveal a high variability in terms of data formats, as well as poor data quality, across network sites.

Objectives The aim of this work is the provision of a process for the assessment of data quality with regard to completeness and syntactic accuracy across independently operated data warehouses using common definitions stored in a central (network-wide) metadata repository (MDR).

Methods For assessment of data quality across multiple sites, we employ a framework of so-called bridgeheads. These are federated data warehouses, which allow the sites to participate in a research network. A central MDR is used to store the definitions of the commonly agreed data elements and their permissible values.

Results We present the design for a generator of quality reports within a bridgehead, allowing the validation of data in the local data warehouse against a research network's central MDR. A standardized quality report can be produced at each network site, providing a means to compare data quality across sites, as well as to channel feedback to the local data source systems, and local documentation personnel. A reference implementation for this concept has been successfully utilized at 10 sites across the German Cancer Consortium.

Conclusions We have shown that comparable data quality assessment across different partners of a distributed research network is feasible when a central metadata repository is combined with locally installed assessment processes. To achieve this, we designed a quality report and the process for generating such a report. The final step was the implementation in a German research network.