MACHINE LEARNING MODELS SIGNIFICANT IMPROVE OUTCOME PREDICTION AFTER CARDIAC ARREST

Conclusion In patients admitted to ICU following cardiac arrest, novel machine learning approaches significantly enhance predictive discrimination compared to classical logistic regression mortality prediction techniques, with a stacked ensemble approach proving most accurate. The impact of pre-hospital data may increase the accuracy of mortality prediction, and further efforts need to be made to enhance the explainability of such models to encourage translation to clinical application.