Phenotypes of South Asian patients with atrial fibrillation and holistic integrated care management: cluster analysis of data from KERALA-AF Registry

Figures

• 

  Fig. 1. Clustering process of this study. The cluster analysis strategy was to minimise the variance of attribute differences within clusters using Ward's approach and the algorithm of squared Euclidean distance. Initially each sample is considered as a separate cluster, the distance metric between clusters is repeatedly updated during the aggregation of clusters, and finally formed different clusters with similar characteristics. The complete hierarchical clustering process was visualised as a dendrogram, where different vertical lines indicated different clusters and the Y-axis represented the distance measure of different clusters with the further away from the end of the tree and the greater the differences between clusters. By observing the dendrogram generated by the clustering process, the median Y-axis value was around 40, and the distance between the last four mergers in the clustering process was significantly greater than the previous mergers, therefore 5 clusters were selected.

• 

  Fig. 2. Patient characteristics of different clusters by hierarchical clustering. AF, atrial fibrillation; AR, aortic regurgitation; BMI, body mass index; CVA, cerebrovascular accident; LBBB, left bundle branch block; LVH, left ventricular hypertrophy; MR, mitral regurgitation; NYHA, New York Heart Association; PAH, pulmonary hypertension; RBBB, right bundle branch block; RWMA, regional wall motion abnormalities; SE, systemic embolism; TIA, transient ischaemic attack.

• 

  Fig. 3. AF treatment managment of different clusters by hierarchical clustering. ACEI, angiotensin-converting enzyme inhibitors; AF, atrial fibrillation; ARB, angiotensin II receptor blockers; CCB, calcium channel blockers; ICD, Implantable cardioverter defibrillator; LAAO, left atrial appendage occlusion.

• 

  Fig. 4. Risk for outcomes on multivariate logistic regression analysis among different clusters by hierarchical clustering. Adjust for beta-blockers, rate-limiting calcium channel blockers, digoxin, angiotensin-converting enzyme inhibitors, angiotensin II receptor blockers, dihydropyridine calcium channel blockers, diuretics, statins, Class I AAD, Class III AAD, antiplatelet agents, anticoagulants, pacemaker implantation, surgery for atrial fibrillation, cardiac defibrillator implantation, left atrial appendage occlusion, and catheter ablation. AAD, antiarrhythmic drug; ACS, acute coronary syndrome; CI, confidence interval; CVA, cerebrovascular accident; HF, heart failure; MACE, major adverse cardiovascular events; OR, odds ratio; SE, systemic embolism; TIA, transient ischemic attack.

• 

  Fig. 5. The A criterion, B criterion, C criterion, and ABC criteria in the whole cohort and each cluster by hierarchical clustering.

• 

  Fig. 6. Impact of adherence to the ABC pathway on all-cause mortality and mace in the whole cohort and each cluster by hierarchical clustering. Adjusted for age, sex, type of atrial fibrillation, diabetes mellitus, heart failure, hypertension, dyslipidaemia, coronary artery disease, and prior cerebrovascular disease or transient ischaemic attack. CI, confidence interval; MACE, major adverse cardiovascular events; OR, odds ratio.