Abstract
BACKGROUND: To evaluate the association between estimated glomerular filtration rate (eGFR) and atrial fibrillation (AF) in patients hospitalized for chronic heart failure (CHF) using a retrospective analysis and establish a risk model of AF development. METHODS: In total, 1301 inpatients with CHF at Fujian Medical University Union Hospital were divided into two groups based on the eGFR: normal eGFR (eGFR ≥ 90 ml/min/1.73 m(2)) and decreased eGFR (eGFR < 90 ml/min/1.73 m(2)). All patients with CHF were randomly assigned to two groups in a 7:3 ratio: the training (910 patients) and validation (391 patients) groups. In the training group, univariate and multivariate logistic regression methods were employed to identify eGFR as a risk factor for the occurrence of AF in patients with CHF. Calibration curves, receiver operating characteristic curves, and decision curve analysis (DCA) were used to establish and validate two models. RESULTS: Of the 1,301 patients, 398 had concurrent AF. Multivariate regression analysis in the training group revealed that eGFR, heart rate, hemoglobin level, LDL-C/HDL-C ratio, left atrial diameter, left ventricular mass index, coronary heart disease were independent risk factors for AF in patients with CHF. Based on the multivariate analysis results, a nomogram model was constructed incorporating these nine independent risk factors to predict AF risk in patients with CHF. The area under the curve for the training and validation groups was 0.80 (95% confidence interval [CI]: 0.77-0.84) and 0.76 (95% CI: 0.70-0.81), respectively. The calibration curve showed good agreement between the predicted and actual values. DCA indicated that when the threshold for the training group was between 8% and 83% and that for the validation group was between 7% and 82%, using the nomogram to predict AF risk was beneficial. CONCLUSIONS: Decreased eGFR is an independent risk factor for the occurrence of AF in patients with CHF. The calibration curve of the constructed nomogram model demonstrated an excellent fit, showing good consistency between the predicted and actual AF risks, thus indicating that the model has strong predictive capability.