ABSTRACT NUMBER: ESOC2026A1110 A DATA-DRIVEN IDENTIFICATION OF DRUG COMMUNITIES ASSOCIATED WITH IMPROVED SURVIVAL AFTER STROKE USING POPULATION-BASED CLINICAL RECORDS

Abstract Background and aims Although acute stroke care has substantially improved, effective neuroprotective strategies have not been translated into routine clinical practice. Real-world clinical data provide an opportunity to explore if commonly prescribed drugs, alone or in combination, could be associated with better outcomes after stroke. The ETNIAS-project applies data-driven methods to population-based health records to identify pharmacological patterns linked to post-stroke survival. Methods We analysed 133,605 patients admitted with stroke between 2015 and 2022 using the Andalusian Population Health Database, which contains more than 49 million clinical records. Sociodemographic characteristics, vascular comorbidities, pre-stroke medication exposure (ATC -Anatomical, Therapeutic, Chemical-classification), and 3-month survival were included. Associations with survival were assessed using bootstrap penalised logistic regression models, and variable importance was explored using tree-based methods. Also, clustering techniques were used to group medications into drugs communities. Results Among thousands of possible drug combinations, 4,882 were significantly associated with 3-month survival. Application of a predefined selection criteria (odds ratio ≥1.5 and p ≤0.02) reduced the analysis to 19 ATC communities showing consistent associations with survival. These communities included drug classes related to thyroid therapies (ATC H03CA), specific antirheumatic agents such as glucosamine and chondroitin sulfate (ATC M01CX), retinoids and vitamin A derivatives (ATC A11CA), and combinations involving cardiovascular and antithrombotic treatments. Conclusions By analysing population-based clinical data, pharmacological communities associated with post-stroke survival were identified. While causal relationships cannot be inferred, the findings provide a rational basis for prioritising candidate drugs and combinations for experimental testing in animal models of stroke as a new neuroprotection strategy. Conflict of interest Nothing to disclose