基于列线图的VA-ECMO患者临床死亡风险预测模型构建OA

Objective:To identify risk factors associated with in-hospital death and to develop a nomogram-based predictive model for in-hospital mortality in acute myocardial infarction(AMI)patients treated with venous-arterial extracorporeal membrane oxygenation(VA-ECMO).Methods:A total of 162 consecutive patients with AMI who received VA-ECMO support between May 2021 and June 2025 were retrospectively enrolled.The time of ECMO initiation was defined as the start of follow-up,and patients were followed until hospital discharge or death,whichever occurred first.In-hospital all-cause mortality was defined as the primary endpoint.Cox proportional hazards regression analysis was performed to evaluate the associations between candidate variables and the risk of in-hospital mortality.Variables were selected using least absolute shrinkage and selection operator(LASSO)regression,and a multivariable Cox regression model was subsequently constructed.Based on the final model,a nomogram was developed to predict in-hospital survival probability.Model discrimination was assessed using the concordance index(C-index).The 28-day time point was used as a fixed landmark for time-dependent receiver operating characteristic(ROC)curve analysis to evaluate short-term predictive performance.Model calibration was evaluated using calibration curves,and clinical utility was assessed using decision curve analysis(DCA).Results:Multivariable analysis demonstrated that cardiac troponin T,soluble suppression of tumorigenicity-2(sST2),hemoglobin concentration,prothrombin time,serum sodium level,and alanine aminotransferase were significantly associated with in-hospital mortality.White blood cell count and albumin showed borderline statistical significance in the model.The nomogram incorporating these eight variables exhibited good discriminative performance and satisfactory calibration,indicating favorable clinical applicability.Conclusion:This study identified key clinical variables associated with in-hospital mortality and successfully developed and validated a nomogram-based prediction model.The proposed model provides a simple and reliable tool for individualized risk stratification and may assist clinicians in optimizing decision-making and management strategies for this high-risk population.