PSO-XnB: a proposed model for predicting hospital stay of CAD patients

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Computational Medicine

Background:

• Coronary artery disease (CAD) management requires accurate prediction of patient length of stay (LOS).
• Traditional methods for LOS prediction often lack the precision needed for effective healthcare resource allocation.
• Developing advanced predictive models is crucial for improving patient care and operational efficiency in hospitals.

Purpose of the Study:

• To introduce a novel predictive model, Particle Swarm Optimized-Enhanced NeuroBoost (PSO-ENB), for categorizing hospital LOS in CAD patients.
• To enhance prediction accuracy by integrating deep autoencoders, eXtreme gradient boosting, and particle swarm optimization.
• To validate the model's performance against existing approaches in healthcare applications.

Main Methods:

• Utilized deep neural autoencoders for dimensionality reduction of patient data.
• Employed an eXtreme gradient boosting (XGBoost) model fed with autoencoder-reconstructed data.
• Optimized XGBoost hyperparameters using particle swarm optimization (PSO) for enhanced predictive power.
• Implemented a fuzzy rule-based system to categorize LOS into four distinct classes.

Main Results:

• The proposed PSO-ENB model achieved a high overall accuracy of 98.8%.
• Demonstrated superior performance compared to traditional ensemble models and previous research.
• Achieved the highest scores in precision, recall, and F1-scores across all LOS categories.
• Validated the model's effectiveness for medical healthcare applications.

Conclusions:

• The PSO-ENB model offers a significant advancement in predicting hospital LOS for CAD patients.
• The integration of deep learning, ensemble methods, and optimization techniques yields superior predictive accuracy.
• This model shows strong potential for improving clinical decision-making and healthcare management.