Help with shorter waiting times for emergency departments

Wednesday 09 Jan 19


PhD at DTU

Learn more about a PhD at DTU and about our Industrial PhD programme at DTU
A new DTU research project seeks to reduce bottlenecks and optimize human resources in hospital accident and emergency departments.

Very long hospital waiting times in Danish A&E departments are a familiar problem, causing stress and annoyance to patients and hospital staff alike. But what can be done about waiting times and bottlenecks in A&E, and how can hospitals plan their way out of being understaffed or overstaffed in the face of changes, or when A&E comes under extra pressure?

Industrial PhD John Bruntse Larsen of DTU Compute is investigating this in cooperation with the company PDC A/S in the ‘Hospital Staff Planning with Multi-Agent Goals’ research project. In the project, John Bruntse Larsen is developing prediction algorithms or mathematical models to optimize planning, identify patients at risk, and for inclusion in decision-making support tools in the form of an app or other IT tools for staff in a hospital A&E Department.

Algorithms and mathematical models predicting flow in A&E
“In the project, I work with very large amounts of historical data from previous patient care and ongoing, current data on patients and hospital staff from A&E and other departments at a hospital in the Capital Region of Denmark. With this in mind, I build algorithms and mathematical models to predict how the flow of patients will look hour by hour—as well as day by day. The aim is to give A&E a means of predicting how the Department will be in future. This could be a general idea of how long it will take to register patients, depending on whether the doctors decide A or B, or when is the best time to get new beds ready, and so on,” John Bruntse Larsen explains.

The challenge of imitating reality
“The mathematical models in the project mimic the interaction of staff, patients and IT systems, both in A&E and in other hospital departments, in what’s known as a multi-agent system. Artificial intelligence defines an agent as an autonomous unit capable of understanding things happening in its environment and having the ability to act in and thereby affect its environment. An agent may be a robot with sensors and motors, but an agent can also be an independent program in a computer system. In my research project, an agent is a player such as a doctor, nurse or patient in A&E or some other hospital department, and each agent is translated into an independent computer program. The different programs or players interact with each other, imitating real life,” John Bruntse Larsen explains.

“The challenge of developing mathematical models or programs is to assess what data is relevant. It requires insight into the area being studied—in this case, a hospital—and an evaluation of whether the imitation and predictions actually hold true.

“I’m working on the solutions to both challenges. I’ve visited a hospital to gain insight into their work processes, and I’ve experimented with modelling this. My final model could potentially be developed for application in other hospitals departments,” John Bruntse Larsen says.

The industrial PhD project is expected to be completed in 2019.