
Machine Learning Based Predictive Modeling of Readmissions Following Extracorporeal Membrane Oxygenation Hospitalizations
This study developed and assessed machine learning models, particularly using XGBoost, to predict 90-day nonelective readmissions following extracorporeal membrane oxygenation (ECMO) hospitalizations. Analyzing data from the Nationwide Readmissions Database (2016-2020), the study found that the XGBoost model outperformed traditional logistic regression in prediction accuracy and calibration. Key factors influencing readmission rates included duration of hospital stay, heart/lung transplants, and type of insurance.








