Ensemble Learning Method for In-Hospital Cardiac Arrest Prediction

Area of Science:

• Biomedical Engineering
• Data Science
• Cardiology

Background:

• Existing cardiac arrest prediction models using multivariate time series data achieve high precision but suffer from low sensitivity and high false alarm rates.
• The need for improved predictive models with better sensitivity and specificity is critical for timely clinical intervention.

Purpose of the Study:

• To develop and evaluate an ensemble approach for cardiac arrest prediction that improves upon the precision-recall performance of existing machine learning methods.
• To enhance the accuracy and reduce false alarms in predicting rapid cardiac arrest events.

Main Methods:

• An ensemble method was proposed, combining multiple machine learning models to predict cardiac arrest from multivariate time series data.
• The performance of the ensemble approach was evaluated against other machine learning methods using precision-recall metrics.

Main Results:

• The proposed ensemble method achieved an overall area under the precision-recall curve of 46.7%.
• This represents an improvement in the precision-recall balance compared to other evaluated machine learning methods.

Conclusions:

• The developed ensemble approach shows potential for more accurate and rapid response to cardiac arrest events.
• Further research may focus on optimizing the ensemble components and validating the model in diverse clinical settings.