203. Addressing Emergency Department Overcrowding through Efficient Queue Management: a Deep Reinforcement Learning Approach
Invited abstract in session MD-3: ED and ICU, stream Sessions.
Monday, 13:30-15:00Room: NTNU, Realfagbygget R9
Authors (first author is the speaker)
| 1. | Luca Zattoni
|
| DEI, University of Bologna | |
| 2. | Andrea Eusebi
|
| DEI, University of Bologna | |
| 3. | Cristiano Fabbri
|
| IRCCS Azienda Ospedaliero Universitaria di Bologna | |
| 4. | Marco Leonessi
|
| IRCCS Azienda Ospedaliero Universitaria di Bologna | |
| 5. | Enrico Malaguti
|
| DEI, University of Bologna | |
| 6. | Paolo Tubertini
|
| IRCCS Azienda Ospedaliero Universitaria di Bologna |
Abstract
Emergency Department (ED) overcrowding is a growing concern in modern healthcare systems, as it often leads to prolonged waiting times, patient discomfort, and potential deterioration of clinical conditions, as well as increased stress on medical staff. This stress may also negatively affect decision-making capabilities, further exacerbating inefficiencies in patient management. While structural limitations and increasing demand represent major challenges, they also emphasize the need for advanced Decision Support Systems capable of adaptive, real-time responses. In this work, we propose a Deep Reinforcement Learning (DRL) approach to dynamically manage patient queues, aiming to mitigate overcrowding by optimizing the allocation of medical services. By modelling the ED as a sequential decision-making environment, our model learns to prioritize patients based on their clinical features and expected resource needs, while considering the overall system state. This enables the development of intelligent queue management strategies that adapt to evolving conditions and heterogeneous patient flows. The model is trained and evaluated in a Discrete Event Simulation environment which realistically replicates ED dynamics and variability in patient pathways. Preliminary results show a reduction in both the length of stay and the average number of patients in the system, demonstrating the potential of DRL in supporting operational decisions and improving ED performance under stress.
Keywords
- Emergency Department
- Artificial Intelligence
- Modelling and simulation
Status: accepted
Back to the list of papers