gms | German Medical Science

25. Jahrestagung der Gesellschaft für Arzneimittelanwendungsforschung und Arzneimittelepidemiologie (GAA)

Gesellschaft für Arzneimittelanwendungsforschung und Arzneimittelepidemiologie

22.11. - 23.11.2018, Bonn/Bad Godesberg

Data quality of pharmaceutical claims data – How to measure, assess and improve?

Meeting Abstract

Suche in Medline nach

  • corresponding author presenting/speaker Nicholas Heck - Wissenschaftliches Institut der PKV, Köln, Germany; Department of Sociology, City, University of London, London, Great Britain
  • author Christian Jacke - Wissenschaftliches Institut der PKV, Köln, Germany

Gesellschaft für Arzneimittelanwendungsforschung und Arzneimittelepidemiologie e.V. (GAA). 25. Jahrestagung der Gesellschaft für Arzneimittelanwendungsforschung und Arzneimittelepidemiologie. Bonn/Bad Godesberg, 22.-23.11.2018. Düsseldorf: German Medical Science GMS Publishing House; 2018. Doc18gaa05

doi: 10.3205/18gaa05, urn:nbn:de:0183-18gaa050

Veröffentlicht: 23. November 2018

© 2018 Heck et al.
Dieser Artikel ist ein Open-Access-Artikel und steht unter den Lizenzbedingungen der Creative Commons Attribution 4.0 License (Namensnennung). Lizenz-Angaben siehe http://creativecommons.org/licenses/by/4.0/.


Gliederung

Text

Background: High data quality is an important property of pharmaceutical claims data. General guidelines such as “Gute Praxis Sekundäranalyse (GPS)“ intend to assure methodological transparency of claims data. However, these guidelines do not report concrete steps for measurement, assessment and improvement of data quality. A general framework for the evaluation of data quality is not available. In contrast, the guidelines of the adaptive management of data quality for cohort studies and registers (GAM) from the field of medical informatics presents an interesting approach for primary data, but not for secondary data.

The present study intents to adopt the GAM to a dataset of claims data related to pharmaceuticals. A proof of concept should show how general problems of claims data fit into the general framework of data quality. Possible caveats should be identified for data holders and data generating private insurances.

Materials and methods: Private health insurances sent pharmaceutical claims data to a central clearing instance (Scientific Institute of Private Health Insurance). The data were harmonized and enriched via the pharmaceutical central number (PZN) of the ABDATA. A set of variables related to the prescription submitting population was used to define several data quality indicators between the time intervals 2011 and 2017. Person- and product-related variables defined the data quality indicators of the corresponding data levels such as integrity, organization and correctness. Raw distributions and point estimates for time series between 2011 and 2017 provided estimated parameters which served as threshold values. These historical thresholds helped to distinguish between conspicuous and non-conspicuous data. A final data quality score summarizes the results to one concise data quality score.

Results: The selected variables and data quality indicators are suitable to describe the data quality of each data quality indicator on every single data quality level. Crucial completeness issues of pharmaceutical claims data of private insurances can be addressed. A feedback for data generating instances (private health insurances) can be resumed before analysis starts.

Conclusion: The adoption of data quality screening and cleaning techniques of primary and secondary data analysis is possible. The selected variables served as examples to show high and low data quality. The immediate feedback response helps private insurances to adapt their data management and to increase their data quality on concrete variables. A constant data monitoring process based on the general GAM framework is suitable for pharmaceutical claims data.