AbstractPurposeDatabase heterogeneity can impact effect estimates. Harmonisation provided by common protocols and common data models (CDMs) can increase the validity of pharmacoepidemiologic research. In a case study measuring the changes in the safety and effectiveness of stroke prevention therapy after the introduction of direct oral anticoagulants (DOACs), we performed an international comparison.MethodsUsing data from Stockholm, Denmark, Scotland and Norway, harmonised with a common protocol and CDM, two calendar‐based cohorts were created: 2012 and 2017. Patients with a diagnosis code of atrial fibrillation 5 years preceding the 1‐year cohort window were included. DOAC, vitamin K antagonist and aspirin treatment were assessed in the 6 months prior to the start of each year while strokes and bleeds were assessed during the year. A Poisson regression generated incidence rate ratios (IRRs) to compare outcomes from 2017 to 2012 adjusted for changes in individual‐level baseline characteristics.ResultsIn 280 359 patients in the 2012 cohort and 356 779 in the 2017 cohort, treatment with OACs increased on average from 45% to 65%, while treatment with aspirin decreased from 30% to 10%. In all countries except Scotland, there were decreases in the risk of stroke and no changes in bleeding risk, after adjustment for changes in baseline characteristics. In Scotland, major bleeding (IRR 1.09, 95% confidence interval [CI] [1.00; 1.18]) and intracranial haemorrhage (IRR 1.31, 95% CI [1.13; 1.52]) increased from 2012 to 2017.ConclusionsStroke prevention therapy improved from 2012 to 2017 with a corresponding reduction in stroke risk without increasing the risk of bleeding in all countries, except Scotland. The heterogeneity that remains after methodological harmonisation can be informative of the underlying population and database.