(015) Experts’ opinion on data quality assessments for multi-database pharmacoepidemiologic studies: A qualitative analysis

Background: There is currently no comprehensive guidance on best practices and recommendations for data quality assessment in multi-database pharmacoepidemiologic studies (MDPES). As part of a larger project funded by the International Society for Pharmacoepidemiology (ISPE) with the goal of producing a guidance manuscript, and besides a scoping review of published literature, there is also a need to collect unpublished data on current practices for data quality assessment in MDPES.

Objectives: To assess which data quality assurance systems are in place for MDPES and what experts recommend for a guidance manuscript.

Methods: Researchers with a demonstrated expertise in conducting MDPES from various working sectors and world regions were invited for interviews. An interview guide was developed comprising 30 questions in five main sections, as own experience and challenges in conducting data quality assessment, knowledge about others’ experience, role of a guidance paper, role of machine learning, and benchmarking of results. Semi-structured online interviews were conducted on Microsoft Teams between July-August 2023. All interviews were videotaped, transcribed verbatim, and a thematic analysis was conducted using Atlas.ti. A combined deductive and inductive approach was used for coding. The Science-Geosciences Ethics Review Board of Utrecht University has approved this study.

Results: Four out of fifteen researchers were interviewed, and two written inputs were received for analyses. All respondents held senior positions with >10 years’ experience of conducting MDPES. Ten main themes were emerged during analysis, where the most highlighted were comparison to other data sources (including comparison to: a gold standard nationwide source, within data source, and cross-country), diversity among data sources (e.g., differences in: healthcare systems, coding systems, machinery and measurements, and types of data sources), and recommendations for a guidance paper (i.e., checklist, definitions, multi-disciplinary team). Other themes were types of data quality checks, missingness, recommendations for data quality assessment, and role of machine learning.

Conclusions: According to respondents, there is no global data quality assurance system for MDPES, but they highlighted the importance of various comparisons between data and the essential differences that exist in data sources. They recommended a checklist to be part of the guidance paper, but hesitated about inclusion of a dashboard. These findings, together with results of the scoping review will inform the following objectives of this ISPE manuscript.