Integration of Machine Learning Algorithms and Discrete-Event Simulation for the Cost of Healthcare Resources

Area of Science:

• Healthcare Management
• Operations Research
• Artificial Intelligence in Healthcare

Background:

• Effective healthcare resource allocation is crucial for managing patient throughput and minimizing wait times in healthcare institutions.
• Emergency departments (EDs) face unique challenges in balancing resource utilization with patient demand and operational costs.

Purpose of the Study:

• To estimate patient volume (p) and waiting time (w) in an ED using healthcare resource allocation models.
• To analyze the cost-effectiveness of healthcare resources based on varying cost coefficients (δi).
• To compare the performance of machine learning algorithms in predicting these key performance indicators.

Main Methods:

• Integration of discrete-event simulation (DES) with machine learning (ML) algorithms: Random Forest (RF), Gradient Boosting (GB), and AdaBoost (AB).
• Estimation of output variables (p and w) based on resource cost coefficients (δi).
• Performance evaluation of ML algorithms using accuracy metrics in training and testing stages.

Main Results:

• AdaBoost (AB) exhibited the highest accuracy in predicting patient volume (p) and waiting time (w) across various resource cost scenarios during the training phase.
• Gradient Boosting (GB) showed strong performance in the test stage, though AB outperformed it for specific p estimations at δ0.2.
• Analysis indicated that optimizing resource allocation based on the δi coefficient can increase hospital income from patient volume, but may variably impact waiting times and costs.

Conclusions:

• The integrated DES and ML approach, particularly using AdaBoost, provides a robust method for optimizing healthcare resource allocation in EDs.
• Resource allocation scenarios driven by the δi coefficient are generally favorable for EDs, as increased patient throughput (p) often outweighs resource costs.
• Careful consideration of resource cost coefficients is necessary, as higher values (δ0.2, δ0.3) can increase, rather than decrease, hospital costs related to waiting times (w).