BACKGROUND: Distributed data networks allow us to address otherwiseinaccessible questions, such as the effects of uncommon treatments,risks of rare adverse drug reactions and heterogeneity in the effectsof medicines across populations. Challenges in multi‐database studiesinclude both practical barriers, such as restrictions on data access, andmethodological issues, such as variation in the collection and coding ofinformation across centers.
OBJECTIVES: This symposium will address the challenges of using dis-tributed data for pharmacoepidemiologic analyses. Specialists willpresent current solutions, weigh the advantages and limitations andhighlight areas for further research. The practical recommendationsprovided will benefit any researcher interested in conducting multi‐center studies.
DESCRIPTION: The symposium will consist of a series of talks, summa-rized by a moderator‐led discussion. 1.“Dealing with differences inhealthcare systems in multi‐database studies”. Examples of the impact of differences, such as the capturing and registration of exposure andoutcomes, across healthcare systems will be presented. Differentapproaches to deal with this, including common data models and stan-dardized protocols, will be discussed(20 min).2.“Methods for meta‐analytic pooling of results from distributed networks: random effects,fixed effects and Bayesian methods”. Distributed networks usuallywork with site‐specific analyses pooled via meta‐analysis. Thestrengths, weaknesses, and assumptions underlying different methodsas well as a proposal of best practices for the analysis of distributednetworks will be presented(20 min).3.“Distributed analytic anddata‐sharing methods that enable robust statistical analysis withoutthe need to share individual‐level data”. Methods that ensure consis-tent data processing and statistical analysis across data sources, andmethods that allow database‐specific analysis (e.g., covariate selection)will be presented. Different analytic options based on different typesof outcomes and different adjustment methods will also be discussed(20 min).4.“Making the most of it: how to handle systematically miss-ing information in multi‐database studies”. Several approaches exist tohandle missing data across a distributed network. A hierarchy ofapproaches to leverage information across a network, along with ahierarchy of their requirements will be presented for discussion(20 min). 5. Panel discussion(10 min)
