A Clinical and Biomarker-Based Model for Predicting in-Hospital Mortality in Patients Undergoing Extracorporeal...

A nomogram was constructed for predicting in-hospital mortality in patients receiving extracorporeal membrane oxygenation (ECMO) using clinical indicators, to provide a reference for clinical risk stratification.

A retrospective cohort of 122 ECMO patients (81 survivors, 41 non-survivors) from Taizhou Hospital of Zhejiang Province (January 2022-December 2024) was used for model development, with 22 patients from the MIMIC-IV database serving as the external validation cohort. A two-step Boruta-LASSO feature selection strategy was applied to identify key predictors, followed by construction of five machine learning models; the best-performing model was used to develop a nomogram, validated internally and externally via ROC analysis, calibration curves, decision curve analysis, and SHAP interpretability analysis.

Four key predictors were identified: malignant tumor, Lactic Acid, monocyte-to-white blood cell ratio (MWR), and activated partial thromboplastin time-to-albumin ratio (AAR). The logistic regression model exhibited the best and most stable predictive performance, with a cross-validated AUC of 0.765 (95% CI: 0.667-0.863) in the primary cohort. External validation in the MIMIC-IV cohort (n=22) yielded an AUC of 0.838 (95% CI: 0.655-1.000), suggesting preliminary external validation with acceptable generalizability.

Malignant tumor, elevated Lactic Acid, decreased MWR, and elevated AAR are key predictive factors for in-hospital mortality in ECMO patients. The nomogram model demonstrated satisfactory discriminatory ability, acceptable calibration, and potential clinical utility, serving as a valuable exploratory tool for bedside risk stratification in ECMO patients. Further validation in larger, multicenter, prospective cohorts is warranted before clinical implementation.