Development and validation of a prediction model for risk of atrial fibrillation in elderly patients with nonalcoholic fatty liver disease

Objectives To develop a prediction model for risk of atrial fibrillation(AF) in elderly patients with nonalcoholic fatty liver disease (NAFLD). Methods This study collected data from 2035 elderly patients over 65 years old diagnosed with NAFLD at Northern Jiangsu People's Hospital. Using a 7:3 ratio, participants were separated into two groups through random assignment: a model development cohort (n = 1424) and an internal verification cohort (n = 611). 511 elderly NAFLD patients from The First Affiliated Hospital of Soochow University were collected as an external validation set to further assess the predictive capability of the model. Using multivariate logistic regression analysis, we developed a predictive model for AF and visualized it through a nomogram. The discrimination, calibration, and clinical application value of the model were comprehensively evaluated using receiver operating characteristic (ROC) curves, calibration curves, decision curve analysis (DCA), and clinical impact curves (CIC). Results In this study, 2035 elderly patients with NAFLD, 116 cases (5.7%) were found to have AF. The baseline features between validation and training groups exhibited no significant statistical differences( P  > 0.05). Direct bilirubin (OR = 1.405, 95% CI: 1.281–1.549), body mass index (BMI) (OR = 1.168, 95% CI: 1.087–1.255), low-density lipoprotein cholesterol (LDL-c) (OR = 0.494, 95% CI: 0.377–0.643), albumin (OR = 0.804, 95% CI: 0.735–0.878), and age (OR = 1.075, 95% CI: 1.039–1.112) were all independent risk factors for AF in elderly NAFLD patients ( P  < 0.001). The combined prediction model composed of these five indicators had an AUC value of 0.829 (95% CI: 0.774–0.884) in the training group, 0.855 (95% CI: 0.794–0.916) in the internal validation group, and 0.826 (95% CI: 0.742–0.910) in the external validation set. The consistency of the prediction model was effectively confirmed through its calibration curve. The DCA and CIC showed that the risk threshold probabilities for AF in the training and validation groups were 5%-79.5% and 5%-90%, respectively. Conclusions The nomogram model constructed in this study has good predictive efficacy and clinical application value for the risk of AF in elderly NAFLD patients.