Identifying risk patterns in older adults with atrial fibrillation by hierarchical cluster analysis: A retrospective approach based on the risk probability for clinical events

Abstract

Background: Older adults with atrial fibrillation (AF) have highly diverse risk levels for mortality, heart failure (HF), thromboembolism (TE), and major bleeding (MB), thus an integrated risk-pattern algorithm is warranted. Methods: We analyzed 573 AF patients aged ≥ 75 years from our single-center cohort (Shinken Database 2010–2018). The 3-year risk scores (risk probability) for mortality (M-score), HF (HF-score), TE (TE-score), and MB (MB-score) were estimated for each patient by logistic regression analysis. Using the four risk scores, cluster analysis was performed with Ward’s linkage hierarchical algorithm. Results: Three clusters were identified: Clusters 1 (n = 429, 74%), 2 (n = 24, 5%), and 3 (n = 120, 21%). The clusters were characterized as standard risk (Cluster 1), high TE- and MB-risk (Cluster 2), and high M- and HF-risk (Cluster 3). Oral anticoagulants were prescribed for over 80% of the patients in each cluster. Catheter ablation for AF was performed only in Cluster 1 (8.9%). Compared with Cluster 1, Cluster 2 was more closely associated with males, asymptomatic AF, history of cerebral infarction or transient ischemic attack, history of intracranial hemorrhage, high HAS-BLED score (≥3), and low body mass index (<18.0 kg/m2). Cluster 3 was more closely associated with old age, heart failure, and low estimated creatinine clearance (<30 mL/min). Conclusion: The cluster analysis identified those at a high risk for all-cause death and HF or a high risk for TE and MB and could support decision making in older adults with AF.

Keywords

• Atrial fibrillation
• Older adults
• Cluster analysis
• Anticoagulation
• Rhythm control
• Rate control