Nomograms for Predicting the Risk of Acute Heart Failure in ICU Patients with Atrial Fibrillation

Abstract Background This study developed two nomograms to predict the incidence of acute heart failure (AHF) in patients of atrial fibrillation (AF) and evaluated the predictive value of the nomograms. Methods 18179 patients of AF from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database were randomly divided into two groups in a seven-to-three ratio to form a training cohort (n = 12725) and a validation cohort (n = 5454). Least absolute shrinkage and selection operator (LASSO) regression analyses were used to identify associated risk variables. Two nomograms were established. The new models were assessed in terms of the concordance index (C-index), the area under the curve (AUC) of receiver operating characteristic (ROC) analysis, calibration curve, and decision curve analysis (DCA). Results Least absolute shrinkage and selection operator (LASSO) regression analysis identified seven potential predictors of acute heart failure in nomogram 1 and three potential predictors in nomogram 2. Multivariate logistic regression analysis was used to evaluate the effects of these predictors and create final models. The concordance index values were 0.768 and 0.696, respectively. The areas under the curves for the training and validation sets in nomogram 1 were 0.768 and 0.763, respectively. The areas under the curves for the training and validation sets in nomogram 2 were 0.696 and 0.692, respectively. Conclusion Seven variables were identified as predictors in nomogram 1 and three variables were identified in nomogram 2. Our nomograms are reliable convenient approaches for predicting acute heart failure in patients with atrial fibrillation.