부산대학교 도서관

Introduction: Despite growing evidence that atrial fibrillation (AF) burden impacts clinical outcomes, few clinical risk scores incorporate individual-level arrhythmia characteristics. We analyzed predictors of hospitalization for heart failure (HF) in patients using both clinical history and ambulatory cardiac monitoring data.Methods: We evaluated patients who underwent Zio XT ambulatory monitoring in the US (2014-2020). The monitoring data were linked with clinical data from Centers for Medicare and Medicaid Services. Patients on anticoagulation or anti-arrhythmic therapy and those with evidence of sustained VT, known ventricular fibrillation/flutter, prior ablation, prior left atrial appendage occlusion, or cardiac arrest in the prior year were excluded. We randomly divided our dataset into a 70% training dataset and 30% testing dataset. Predictive models were generated using LASSO Cox regression for variable selection in the training dataset and evaluated on testing dataset among ambulatory ECG variables and the components of CHA2DS2-VASC.Results: Among 224,682 patients in the training dataset, 4,021 (1.8%) were hospitalized for HF within 1 year. A model that included components of the CHA2DS2-VASC and all ECG variables (excluding multicollinear variables) had greater discrimination and calibration for HF hospitalization (C-statistic 0.85 [0.84-0.86]), compared with the CHA2DS2-VASC score alone (C-statistic 0.73 [0.72-0.74], Figure). ECG findings with the greatest predictive value were the presence of premature ventricular couplets (HR 1.54 [1.43-1.65]) and presence of AF (HR 1.53 [1.35-1.72]).Conclusion: A model that incorporates both ECG variables from ambulatory monitoring data and the components of the CHA2DS2-VASC provides better discrimination for the risk of HF hospitalization compared with the CHA2DS2-VASC risk score alone. Premature ventricular couplets and AF are highly associated with HF hospitalization.