Predicted adherence and ischaemic stroke risk in atrial fibrillation patients initiating oral anticoagulation: A cohort study of the medication adherence score

BACKGROUND: Medication non-adherence is a major challenge in stroke prevention in atrial fibrillation (AF), yet no validated tool exists to identify patients at risk of non-adherence at the time of oral anticoagulation (OAC) initiation. METHODS: In this retrospective cohort study at a large tertiary centre in New York City (2015-2023), adults with AF and a CHA₂DS₂-VASc score ≥ 2 initiating OAC were included. Predicted medication adherence was quantified using the Medication Adherence Score (MAS), a validated algorithm incorporating demographic, socioeconomic, and geographic attributes, dichotomised as high (MAS ≥80) or low (MAS <80). The primary outcome was ischaemic stroke within 12 months, analysed using Fine-Gray competing-risk regression. RESULTS: Of 11,233 eligible patients, 4035 (35.9%) had high and 7198 (64.1%) had low predicted adherence. Most patients received a direct oral anticoagulant (DOAC, 70.3%), 10.7% received warfarin, and 19.0% switched agents. Ischaemic stroke occurred in 6.2% of patients. High predicted adherence was associated with significantly lower stroke risk (sHR 0.67; 95% CI 0.56-0.80), with 12-month cumulative incidences of 5.08% vs. 7.97%. Continuously modelled MAS confirmed a dose-response relationship (HR 0.98 per unit increase; 95% CI 0.97-0.99). Results were consistent after excluding patients with prior stroke (sHR 0.66; 95% CI 0.53-0.83). CONCLUSIONS: In AF patients initiating OAC, low predicted medication adherence is independently associated with increased ischaemic stroke risk. The MAS may support early identification of high-risk patients, enabling targeted adherence interventions at OAC initiation.