Establishment and Validation of A Model for New-onset Atrial Fibrillation in Patients with STEMI; A Study Based on CMR

Background Identifying patients at risk for new-onset atrial fibrillation (AF) during hospitalization for ST-segment elevation myocardial infarction (STEMI) is clinically challenging. We aimed to develop and internally validate a prediction model integrating cardiac magnetic resonance (CMR) and clinical data. Methods In this single-center retrospective study, 503 consecutive STEMI patients who underwent CMR were analyzed. Baseline clinical, laboratory, and imaging variables were entered into a ridge-penalized logistic regression model (70% training, 30% validation). Model discrimination was assessed with the area under the receiver operating characteristic curve (AUC), calibration with decile-based plots and bootstrapping, and clinical utility with decision curve analysis (DCA). Odds ratios (ORs) were derived from a conventional multivariable model, and a nomogram-style visualization was generated. Results New-onset AF occurred in 45 patients (8.9%). Left atrial (LA) reservoir strain was the only independent predictor (OR 0.913, 95% CI 0.863–0.964, p = 0.0009). Impaired LA strain reflects atrial remodeling and wall stress, which predispose to electrical instability and cause new-onset AF. Left ventricular global longitudinal strain (LVGLS) showed a similar trend but was not significant (OR 0.931, 95% CI 0.837–1.024, p = 0.147). The ridge model achieved moderate discrimination (AUC 0.765 training, 0.695 validation) with acceptable calibration (training Eavg ≈ 0.030, Brier ≈ 0.076; validation Eavg ≈ 0.023, Brier ≈ 0.079). DCA demonstrated limited incremental benefit compared with treat-all or treat-none strategies. Conclusion CMR-derived LA reservoir strain is central to predicting in-hospital AF after STEMI.