Machine Learning for Outcome Prediction in Electroencephalograph (EEG)-Monitored Children in the Intensive Care Unit

Area of Science:

• Pediatric Critical Care Medicine
• Neuroscience
• Biomedical Informatics

Background:

• Predicting in-hospital mortality in critically ill children is crucial for timely intervention.
• Continuous electroencephalography (cEEG) monitoring is common in intensive care units (ICUs) for critically ill children.
• Existing predictive models may not fully leverage complex data patterns.

Purpose of the Study:

• To evaluate the performance of various machine learning algorithms in predicting in-hospital mortality.
• To compare the predictive accuracy of machine learning models against traditional explanatory models.
• To assess the utility of machine learning in enhancing mortality prediction for critically ill children on cEEG.

Main Methods:

• Retrospective analysis of a database containing 414 critically ill children undergoing cEEG in the ICU.
• Implementation and evaluation of machine learning algorithms including stepwise selection/elimination, LASSO, support vector machine (SVM) with linear kernel, and random forest.
• Comparison of model performance using the area under the receiver operating characteristic curve (AUC).

Main Results:

• Stepwise selection/elimination models achieved the highest predictive performance (AUC = 0.82).
• LASSO and linear kernel SVM models showed strong performance (AUC = 0.79), followed by random forest (AUC = 0.71).
• Traditional explanatory models demonstrated poorer discriminative ability (AUC = 0.63 without etiology, AUC = 0.45 with etiology).

Conclusions:

• Machine learning techniques, particularly stepwise models, significantly enhance the prediction of in-hospital mortality in critically ill children on cEEG.
• These algorithms offer valuable insights beyond traditional explanatory models, even with limited variables and patient numbers.
• The findings support the integration of machine learning into clinical decision-making for pediatric ICU patients requiring cEEG monitoring.