Integrating Machine Learning with Discrete Event Simulation for Improving Health Referral Processing in a Care Management Setting

TL;DR

This paper presents a machine-learning-guided discrete event simulation framework to optimize health referral processing in post-discharge care, reducing delays and improving system efficiency.

Contribution

It introduces an integrated simulation and machine learning approach to enhance health referral workflows in care management settings.

Findings

Reduced referral creation delay time

Improved process efficiency in referral handling

Demonstrated the effectiveness of predictive models in healthcare systems

Abstract

Post-discharge care management coordinates patients' referrals to improve their health after being discharged from hospitals, especially elderly and chronically ill patients. In a care management setting, health referrals are processed by a specialized unit in the managed care organization (MCO), which interacts with many other entities including inpatient hospitals, insurance companies, and post-discharge care providers. In this paper, a machine-learning-guided discrete event simulation framework to improve health referrals processing is proposed. Random-forest-based prediction models are developed to predict the LOS and referral type. Two simulation models are constructed to represent the as-is configuration of the referral processing system and the intelligent system after incorporating the prediction functionality, respectively. By incorporating a prediction module for the referral processing system to plan and prioritize referrals, the overall performance was enhanced in terms of reducing the average referral creation delay time. This research will emphasize the role of post-discharge care management in improving health quality and reducing associated costs. Also, the paper demonstrates how to use integrated systems engineering methods for process improvement of complex healthcare systems.