Development and validation of a nomogram for in-hospital mortality prediction in acute myocardialinfarction patients with cardiac arrest undergoing percutaneous coronary intervention supported by veno-arterial extracorporeal membrane oxygenation.

BACKGROUND

The combination of percutaneous coronary intervention (PCI) and veno-arterial extracorporeal membrane oxygenation (VA-ECMO) has become a widely used approach for resuscitating patients with acute myocardial infarction (AMI) complicated by cardiac arrest (CA). Nonetheless, limited research has focused on predicting in-hospital mortality in affected patients. This study aims to identify factors associated with in-hospital mortality and develop a clinical prediction model for these patients.

METHODS

Clinical presentations of AMI patients with CA undergoing PCI supported by VA-ECMO at two hospitals in Zhengzhou were evaluated. Patients were stratified based on their survival status at discharge. A comprehensive analysis, which included univariate logistic regression, LASSO regression, and multivariate logistic regression, was conducted to identify predictors and develop a nomogram for in-hospital mortality. The nomogram's predictive performance was subsequently compared to that of existing models.

RESULTS

The study included 139 patients, of whom 84 died during hospitalization. Using factors such as age, current smoking, left main culprit vessel, lactic acid levels, and serum creatinine, a nomogram model was developed. The model demonstrated good predictive performance, with an area under the curve of 0.826 (95% CI 0.757-0.894) in the training dataset and 0.783 (95% CI 0.706-0.859) in the internal validation dataset, indicating high accuracy and stability. Clinical decision curve analysis confirmed the model's utility, particularly for risk thresholds above 20%, outperforming existing models.

CONCLUSIONS

This study identified independent predictors of in-hospital mortality in AMI patients with CA undergoing PCI supported by VA-ECMO and developed a clinically applicable prediction model.