Prediction of neurological outcome and mortality in cardiac arrest patients: an explainable machine learning study integrating HRV, EEG, and clinical features

Objective: Early and accurate prediction of neurological outcomes and mortality in comatose patients after cardiac arrest remains a critical challenge. Multimodal data integrating heart and brain electrophysiological signals may improve prognostic accuracy, yet the distinct pathophysiological mechanisms underlying neurological recovery versus survival are not well understood. Methods: We analyzed 331 patients from the I-CARE dataset.Synchronous ECG and EEG data (obtained within 72 hours post-ROSC) and clinical variables were collected. Following preprocessing, multimodal features were extracted from HRV (time/frequency/nonlinear) and EEG (time/frequency/nonlinear/network topology), alongside heartbeat-evoked potentials (HEP). All features were temporally weighted by the time elapsed post-ROSC. Four machine learning models (Logistic Regression, SVM, Random Forest, XGBoost) were developed to predict neurological outcome and mortality. Hyperparameters were optimized with Optuna. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC-ROC), accuracy, sensitivity, specificity, precision, and the F1-score. SHAP analysis was applied to interpret the optimal model and quantified the contribution of each feature. Results: XGBoost achieved the highest performance, with an AUC of 0.95 for neurological outcome and 0.89 for mortality prediction. SHAP analysis identified ShockableRhythm as the top predictor for neurological outcome and EEG kurtosis for mortality. Neurological outcome was more closely associated with EEG complexity and synchrony, whereas mortality was linked to ECG nonlinear dynamics and heart-brain coupling. Conclusion: This study confirms that an XGBoost model integrating multimodal heart-brain electrophysiological features enables accurate early prognosis. The results reveal distinct prognostic mechanisms where neurological recovery is brain-centric while survival depends on cardiac function and heart-brain interactions. These findings provide a new direction for the development of personalized prognostic assessment tools based on physiological mechanisms.