Methods to Address Heterogeneity in Multi-database Studies in Pharmacoepidemiology: Search for Consensus and Future Directions

Multi-database studies (MDBS) are increasingly conducted in pharmacoepidemiology and other fields. Regulatory authorities often require data from multiple data sources to be used in a single study, to enhance the generalizability of results and/or obtain sufficient sample size. However, MDBS poses several challenges, especially how to manage heterogeneity that may arise from differences in data types, structure, and coding, or contexts (health policy, clinical practice, culture, and populations).

Among Asian countries with large databases, MDBBDS have been performed through AsPEN (Asian Pharmacoepidemiology Network) or in collaboration with European and North American networks/countries. However, it is extremely challenging to interpret the findings of MDBS without a clear understanding of the context and purposes of each data source, and how these differ between the data sources in the study. Despite calls for the implementation of strategies to improve replicability, increase transparency and reduce bias in MDBS, to our knowledge, there is currently no guidance for how database heterogeneity should be evaluated or even identified and reported.

DIVERSE is an ISPE-funded project to conduct a scoping review to identify the current methodological issues and solutions to address heterogeneity in MBDS and to propose a set of recommendations for future practice. In this symposium, the DIVERSE investigators will present the ongoing scoping review with examples from Asia/Pacific, Europe and North America and engage participants in proposing and discussing potential solutions and recommendations on how to identify, address, and interpret heterogeneity in MBDS.

After a brief introduction by the moderator (SS), 4 presenters will address the following (45 min):
1. Heterogeneity Issues in Asian pharmacoepidemiologic MDBS - AsPEN examples (EL)
2. A conceptual framework to represent heterogeneity in Europe (RG)
3. Review of existing methods, tools and recommendations to identify and address heterogeneity between data sources in MDBS (RP)
4. A proposal for recommendations on how to identify, address, and interpret heterogeneity in MDBS (RWP)

•Edward Lai, National Cheng Kung University, Taiwan
•Rosa Gini, ARS Toscana, Florence, Italy
•Romin Pajouheshnia, Utrecht University, Netherlands
•Robert Platt, McGill University, Canada
•Carlos Duran Salinas, University Medical Centre Utrecht, Netherlands
•Gillian Hall, Gillian Hall Epidemiology, UK
•Jenni Ilomäki, Monash University, Australia
•Ju Hwan Kim, Sungkyunkwan University, South Korea


Panel discussion (Chair, SS) (45 min): A panel discussion with DIVERSE investigators (RG, RP, GH, RWP, EL), other MDBS investigators (JS, JI), and the symposium participants will address the following questions/points after each panellist introduces briefly his/her multi-database network and specific heterogeneity issues.
1. General feedback/discussion on the proposed recommendations
2. Are there data-specific issues in identifying, addressing, and interpreting heterogeneity in MDBS? Can we identify universal recommendations based on common underlying issues? Should recommendations and guidance be region specific?
3. Should the set of recommendations on heterogeneity issues in MDBS be organized by types of research questions, designs, and approaches (common data model vs common protocol) or by data model types? We can observe differences in the reasons why the data is collected in the first place: how should this be identified and reported?
4. Any specific heterogeneity issues for medications, biologics, vaccines, medical devices? Any unique issues by indications (diseases) or demographics?