Methods Article for a Study Protocol: Refining Stroke Prediction in Atrial Fibrillation Patients in an Ethnically...

Background: Atrial fibrillation (AF) increases ischaemic stroke (IS) risk, which can be mitigated using risk prediction models to guide anticoagulation decisions. The resultant widespread use of anticoagulants has reduced IS rates globally. However, commonly used risk prediction scores were validated in mainly European cohorts. Cardiology society guidelines recommend the local refinement of such risk scores to improve risk prediction. This study aims (1) to determine trends in the prevalence of AF in IS in Auckland, (2) to perform a validation study of the CHA₂DS₂-VASc risk score (congestive heart failure, hypertension, age ≥75 [doubled], diabetes, IS/TIA/thromboembolism [doubled] - vascular disease [e.g., ischaemic heart disease and aortic plaque], age 65-74, and sex [female]) and determine if additional ethnicity factors (i.e., Māori and/or Pacific peoples) improve risk prediction, (3) to identify associations with anticoagulant failure (ie, IS on anticoagulation). Methods: This study will utilise data from the Auckland Regional Community Stroke Study (ARCOS IV [2011-12] and V [2020-21], respectively), a comprehensive registry of stroke patients. The comparative controls will be Auckland residents diagnosed with AF between 1988 and 2020, sampled from the National Minimum Dataset (NMD) - a database of hospital discharge codes collated by Manatū Hauora (the New Zealand Ministry of Health). Firstly, we will investigate trends in the prevalence of AF in IS in ARCOS IV and V. Secondly, we will use a nested case-control design by combining ARCOS V and NMD to determine the model performance of CHA₂DS₂-VASc and risk score refinements stratified by ethnicity. The effect of (a) stroke aetiology, (b) antithrombotic prescribing factors, and (c) potential interactions will also be assessed in the data analysis. Based on sample size estimations, we will require a sample of 1,493 controls and 374 cases with IS/TIA. Conclusion: Utilising data from three datasets will allow us to assess the burden and management of AF at a population level, identify trends in disease, address knowledge gaps in the management of ethnically diverse populations, and explore associations with treatment failure. Our reporting will adhere to the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) guidelines.