Development and validation of a risk prediction model for in-hospital mortality among patients with acute myocardial infarction complicated by ventricular arrhythmia

Abstract Background With the widespread adoption of percutaneous coronary intervention (PCI), patients experiencing acute myocardial infarction (AMI) complicated by ventricular arrhythmia (VA) continue to encounter a significant risk of in-hospital mortality. The predictive accuracy of existing scoring systems for this specific high-risk subgroup requires enhancement, as there is a notable absence of specialized predictive tools that integrate electrophysiological characteristics (such as fragmented QRS waves and electrical storms) with clinical metabolic indicators. This study aims to identify the independent factors influencing in-hospital mortality among patients with AMI complicated by VA, develop and validate a Nomogram prediction model, and provide a reference for early clinical risk stratification. Methods In this study, a retrospective cohort design was employed, encompassing patients diagnosed with AMI complicated by VA who were admitted to the Department of Cardiology at a tertiary first-class hospital in Anhui Province between November 2020 and October 2025. A comprehensive dataset comprising 38 variables was collected, including demographic information, clinical evaluations, laboratory tests, and electrocardiogram (ECG) physiological indices. To address data dimensionality and identify key variables, the Least Absolute Shrinkage and Selection Operator (LASSO) regression was utilized. Subsequently, multivariate logistic regression analysis was conducted to ascertain the independent factors influencing in-hospital mortality. A nomogram model was developed using R software, with its performance assessed through receiver operating characteristic (ROC) curve analysis, calibration curve, and decision curve analysis (DCA). Rigorous internal validation was performed using the bootstrap method with 1,000 resamples and 10-fold cross-validation. Results Among the 236 patients studied, 62 individuals (26.3%) succumbed during hospitalization. The LASSO regression analysis identified eight significant predictor variables: heart failure, modified shock index, TIMI flow grade, abnormal blood potassium levels, abnormal blood creatinine levels, late onset of VA, electrical storm, and fragmented QRS waves. The nomogram model, developed based on these factors, demonstrated excellent discrimination, with an area under the curve (AUC) of 0.845 (95% confidence interval [CI]: 0.783–0.908), surpassing the GRACE score’s AUC of 0.740 (95% CI: 0.670–0.811) and the TIMI risk score’s AUC of 0.723 (95% CI: 0.656–0.708). Following bootstrap validation, the nomogram’s concordance index (C-index) was 0.820, and the AUC from 10-fold cross-validation was 0.849, indicating that the model is not overfitted. The calibration curve demonstrates a high degree of agreement between the predicted probabilities and the actual incidence rates (Spiegelhalter’s Z test, P = 0.627). The DCA curve further confirms that the model provides substantial clinical net benefit within a threshold probability range of 0.07 to 0.99. Conclusion In this study, a risk prediction model for in-hospital mortality among patients with acute myocardial infarction complicated by ventricular arrhythmia was developed and validated. The model incorporates electrophysiological characteristics and biochemical markers, demonstrating high predictive accuracy and generalizability. This model serves as a valuable tool for clinicians in the early identification of patients at extremely high risk, facilitating the formulation of targeted intensive intervention strategies.