Predicting new-onset heart failure hospitalization of patients with atrial fibrillation: development and external validations of a risk score

Abstract

Abstract Background Atrial fibrillation (AF) is a well-known risk factor for heart failure (HF), and HF development as a complication of AF is associated with a decline in the quality of life and poor prognosis. However, unlike thrombotic events, incidence of HF in patients with AF has not changed for decades, and a preventive strategy has yet to be developed. Purpose We sought to develop a risk model for new-onset HF admission in patients with AF and without a history of HF. Additionally, we attempted to externally validate the developed risk model. Methods We utilized two multicenter, prospective, observational registries of AF and analyzed the patients without a history of AF. One of which is defined as a derivation cohort, which included 2,857 patients, and the other is defined as a validation cohort, which included 2,516 patients. We developed a risk model by selecting variables with regularized regression and weighing coefficients by Cox regression analysis with the derivation cohort. The external validity was tested in the validation cohort. Results During the follow-up period, 148 patients (5.2%) in the derivation cohort and 104 patients (4.1%) in the validation cohort developed HF during the median follow-up period of 1,396 (interquartile range [IQR]: 1,078–1,820) days and 1,168 (IQR: 844–1,309) days, respectively. In the derivation cohort, four predictors (age, hemoglobin, serum creatinine, and log-transformed brain natriuretic peptide) were identified as potential risk factors for HF development. The developed risk model showed good discrimination and calibration in both the derivation (area under the curve [AUC], 0.77 [95% confidence interval (CI) 0.73–0.81]; Hosmer-Lemeshow test, P=0.257) and validation cohorts (AUC: 0.76 [95% CI 0.72–0.81]; Hosmer-Lemeshow test, P=0.475). Considering death not due to HF as a competing risk, the cumulative incidence curves for HF admission stratified by the risk score were generated, which showed higher HF hospitalization rate for the higher risk score categories. Conclusion The newly developed risk model with four readily available clinical characteristics and biomarkers performed well in the prediction of new-onset HF admission of patients with AF in both derivation and validation cohort. Funding Acknowledgement Type of funding sources: None.