A Markov Decision Process-based ICU Inventory Management System for a South African Hospital

• A. Steenkamp School of Industrial Engineering, North-West University, South Africa https://orcid.org/0009-0002-4426-2574
• M. Van Zyl-Cillié School of Industrial Engineering, North-West University, South Africa https://orcid.org/0000-0003-3320-706X
• C. Du Plessis School of Industrial Engineering, North-West University, South Africa https://orcid.org/0000-0002-6968-0433

DOI:

https://doi.org/10.7166/36-3-3338

Abstract

The South African healthcare system is divided into public and private healthcare. Although the private healthcare sector is funded by individuals or health insurance, private hospitals experience operational problems associated with the optimal allocation and usage of resources. The intensive care unit (ICU) of a private hospital in South Africa frequently experiences shortages of critical supplies owing to limited inventory visibility and inflexible replenishment policies. Demand variability and static inventory ordering policies are found to be the principal drivers of stock-outs in this specific ICU. This research aims to develop a probabilistic inventory management system for an ICU to enhance supply availability and to minimise risks to patient care. This framework is based on a Markov decision process (MDP), deployed as the methodological foundation for modelling the inventory state transition probabilities derived from historical usage data. A Gradio-based interface facilitates real-time interaction with the model outputs, ensuring accessibility for clinical and administrative staff. The proposed system improves visibility and supports proactive decision-making in response to stochastic demand. The findings demonstrate that an MDP, integrated with accessible data and interface tools, could reduce shortages and expired stock, thereby aligning the ICU inventory management policies more closely with clinical requirements.