Objective

To construct a nomogram prediction model for death in patients with cardiogenic shock after extracorporeal membrane oxygenation (ECMO) weaning.

Methods

Totally 210 patients with cardiogenic shock treated with ECMO between January 2023 and October 2024 were included. They were randomly divided into the modeling group of 147 cases and the validation group of 63 cases complying with a 7:3 ratio. The modeling group was divided into the death group of 65 cases and the survival group of 82 cases based on the 28-day death status. Cox regression analysis was used to determine independent risk factors for death after ECMO weaning in patients with cardiogenic shock. R software was used to construct a nomogram prediction model to predict the likelihood of death in cardiogenic shock after ECMO weaning. The ROC curve and calibration curve were used to verify the discrimination and calibration of the nomogram prediction model, respectively; and the decision curve was used to analyze and evaluate clinical utility and net benefit.

Results

Clinical data such as age, gender, and weight showed no statistically significant differences between the modeling group and the validation group (P > 0.05). The death group had significantly higher age, rates of obesity, CRRT, cardiac arrest before ECMO, preoperative lactate level, and intraoperative red blood cell volume compared to the survival group (P < 0.05). Increase in age, obesity, CRRT, cardiac arrest before ECMO, elevated preoperative lactate level, and increased intraoperative red blood cell volume were independent predictive factors for death after ECMO weaning in patients with cardiogenic shock (P < 0.05). The AUC of ROC curve for modeling group and the validation group was 0.994 and 0.966, respectively; the consistency between the predicted probability and the actual probability of the calibration curve was high, with Hosmer-Lemeshow test modeling group χ² = 4.472, P = 0.619; validation group, χ2 = 3.813, P = 0.874; the clinical decision curve revealed that the nomogram prediction model had good net returns within the high-risk threshold probability range of 0.04 ~ 0.98.

Conclusion

The nomogram model established in this study can effectively predict the likelihood of death in patients with cardiogenic shock after ECMO weaning, and provide reference for personalized treatment.