Coupling simulation with machine learning: A hybrid approach for elderly discharge planning

Elbattah, Mahmoud; Molloy, Owen

Abstract

Healthcare systems are increasingly challenged by the phenomenal growth of population ageing. Healthcare executives are, and will be, in an inevitable need of evidence-based artifacts for decision making. The paper addresses issues in the context of discharge planning for elderly patients with application to hip fracture care in Ireland. A hybrid approach is embraced that integrates simulation modeling with machine learning in an attempt to improve the validity of the simulation model outputs. In terms of simulation modeling, a discrete event simulation model is used to model the elderly patient’s journey through the care scheme of hip fracture. In tandem with the simulation model, predictive models are used to guide the simulation model. Specifically, the predictive models are used to make predictions on the inpatient length of stay and discharge destination of simulation-generated patients. On a population basis, the simulation model provides demand predictions for healthcare resources related to discharge destinations, with a focus on long-stay care such as nursing homes. Our results suggest that there may be a need to reconsider the geographic distribution of nursing homes within particular areas in Ireland in order to keep abreast of the foreseen shift in demographics. Furthermore, the incorporation of machine learning within simulation modeling is claimed to improve the predictive power of the simulation model.

• Presentation Conference Type: Conference Paper (published)
• Conference Name: SIGSIM-PADS '16: SIGSIM Principles of Advanced Discrete Simulation
• Start Date: May 15, 2016
• Online Publication Date: May 15, 2016
• Publication Date: May 15, 2016
• Deposit Date: Nov 23, 2024
• Peer Reviewed: Peer Reviewed
• Pages: 47-56
• Book Title: SIGSIM-PADS '16: Proceedings of the 2016 ACM SIGSIM Conference on Principles of Advanced Discrete Simulation
• ISBN: 9781450337427
• DOI: https://doi.org/10.1145/2901378.2901381
• Public URL: https://uwe-repository.worktribe.com/output/13461842
• Additional Information: Published: 2016-05-15