Ensure a dynamic organisational performance of an Emergency Department by exploring the structure of agent interactions

The Emergency Department (ED) is a strategic sector in the hospital chain. While the reception provided there is the first stage in the care of certain patients in the hospital, the ED nevertheless has a number of limitations, particularly when it comes to managing the flow of arrivals, which has a major impact on the patient pathway.

As a pillar of the industry 4.0 paradigm, a Digital Twin (DT) is defined as a digital model of a product or system, capable of synchronising in real time with its physical twin, and simulating and predicting its behaviour in order to offer steering and control of the physical system as close to reality as possible [Negri et al., 2017; Kritzinger et al., 2018]. The benefits of using JNs for healthcare organisations have been clearly stated [Patrone 2019; Moyaux et al., 2023], with the aim of reducing patient waiting times or length of stay, improving resource planning, etc.

In this vein, the ANR JUNEAU project aims to propose a DT for the ED that can both provide the data needed to visualise the department's behaviour in near-real time and also predict and anticipate its behaviour, by adding a simulation engine coupled with optimisation to respond to the hazards inherent in this type of department, in order to 'slave' and control the 'time patients spend in emergency' indicator.

Given the specific features of the ED, one of the scientific objectives of the project is to propose a hybrid management system by combining centralised approaches (which generally give rise to NP-hard problems) and decentralised approaches (derived from multi-agent approaches) to tackle the problem of the patient pathway. This will allow us to approach the accumulation of the best of both: the efficiency of the centralised in the absence of disruptions and the responsiveness and resilience of the decentralised to react after each disruption [Cardin et al., 2017]. Our hypothesis is that the application of this coupling to the management of the patient pathway of an SAU will make it possible to pilot this flow efficiently while dealing with unforeseen events [Florencia et al., 2023].

In fact, the core of the DT is a Digital Model that is strongly based on an agent-based approach, where organisational performance emerges from interactions between the agents that make up the model, enabling the various stakeholders to explore and evaluate alternatives in a virtual world [Lachtar, 2012].

The challenge is to enable this digital model to evolve in such a way as to ensure a level of performance that meets the expectations of decision-makers in a highly dynamic and uncertain context. Indeed, a complex system such as a DT evolves over the course of its life, as much through the dynamics of the physical twin itself as through the evolution of the JN's modelling assumptions. The question of how to control such a system is a major one, especially if we take into account the application system concerned.

Thus, by studying the interactions between all the agents in the system and understanding the associated dynamic system, it is possible to obtain organisational modelling and control. The aim of this thesis is to qualify the exploration of organisational structures, by :
• Identifying performance indicators that can be used to steer a switch between centralised decision-making and a distributed structure,
• Capturing and characterising the gaps between the Physical Twin and the DT, which are sources of evolution in our model.
This work will be based on the main application of the JUNEAU project: Paediatric Emergency at the Saint-Etienne University Hospital.