Development and external validation of a 2-year mortality risk model for atrial fibrillation and heart failure with an online calculator

Heart failure (HF) and atrial fibrillation (AF) frequently coexist and are associated with increased mortality, yet practical risk stratification tools for this population are limited. This study aimed to develop and validate a two-year mortality risk prediction model for AF-HF patients to facilitate clinical risk stratification. AF-HF patients were retrospectively recruited from two hospitals and divided into a derivation cohort and an external validation cohort. Missing data were handled using multiple imputation. The Boruta algorithm and least absolute shrinkage and selection operator (LASSO) regression were used for feature selection. A multivariable logistic regression model was constructed, with coefficients pooled using Rubin’s rules and predicted probabilities averaged across the imputed datasets. Bootstrap internal validation was performed. Model performance was assessed using the area under the ROC curve (AUC), calibration intercept and slope, the Hosmer–Lemeshow test, the Brier score, and decision curve analysis (DCA). Risk groups were defined by tertiles of predicted probability, and survival differences were assessed using Kaplan–Meier and log-rank tests. An online interactive risk calculator was developed. A total of 987 patients were enrolled (555 in the derivation cohort and 432 in the external validation cohort). Nine predictors were identified: age, diabetes, renal dysfunction, NT-proBNP group (< 5000 vs. ≥5000), Hb, ApoA1, neutrophil-to-lymphocyte ratio (NLR), rhythm control strategy, and poor medication adherence. The model showed good discrimination, with AUCs of 0.848 (95% CI: 0.810–0.886) and 0.862 (95% CI: 0.820–0.903) in the derivation and external validation cohorts, respectively. The optimism-corrected AUC was 0.832. Calibration was adequate in the derivation cohort, while the external validation cohort showed mild calibration drift (intercept 0.338; E/O ratio 0.832). DCA demonstrated net clinical benefit, and survival differed significantly across risk groups in both cohorts (log-rank P < 0.0001). A multidimensional prediction model was developed and externally validated for 2-year mortality risk stratification in AF-HF patients, with an online calculator to facilitate individualized risk assessment. Not applicable.