ABSTRACT NUMBER: ESOC2026A704 FEDERATING EUROPEAN REGISTRIES FOR STROKE (FERES): PRINCIPLES, DATA MODEL AND FIRST SHOWCASE ANALYSIS FOR A FEDERATED ANALYSIS OF MULTIPLE NATIONAL STROKE REGISTRIES

Abstract Background and aims National stroke registries provide high-quality real-world data, yet multinational analyses remain limited by data fragmentation, interoperability gaps, and regulatory constraints. The Federating European REgistries for Stroke (FERES) initiative was designed to enable large-scale European stroke research through privacy-preserving federated analytics. We report the methodological framework of FERES and its formal validation using a predefined clinical showcase. Methods FERES employs a GDPR-compliant federated architecture based on harmonization of national stroke registries to stroke-specific Common Data Elements (CDEs) and distributed analysis using a secure medical informatics platform. Validation was performed using a large harmonized stroke registry cohort, partitioned into independent nodes to emulate a multi-site federation. A predefined Minimal Analysis Subset (Figure 1) addressed a standard comparison between anterior and posterior circulation ischemic strokes. Identical analytical pipelines were executed centrally on pooled data and in federated mode. Agreement was assessed using absolute percentage error (APE) for descriptive metrics, concordance of hypothesis-test decisions, effect-size consistency, and agreement of regression coefficients in log-odds space. Results Federated execution reproduced centralized pooled-data analyses with numerical equivalence. For all descriptive metrics, APE for means, standard deviations, and proportions was effectively 0%, with exact matching counts. All predefined anterior–posterior circulation hypothesis tests showed complete decision concordance (all p<0.05). Effect sizes (Cohen’s d, rank-biserial correlation, Cramér’s V) and multivariable regression coefficients demonstrated only negligible differences across execution modes. No individual-level data were transferred. Conclusions FERES provides a validated, privacy-preserving federated framework that reproduces centralized stroke registry analyses. It enables scalable, multinational real-world evidence generation benchmarking and quality improvement in stroke care. Conflict of interest The FERES project received financial support by the European Academy of Neurology (EAN) and non-financial support from the European Stroke Organization (ESO). All authors : Nothing to disclose Figure 1 - belongs to Methods