When Dr Anton Pak was completing his PhD in Applied Health Economics at UQ, his conversations with clinicians and health professionals revealed a persistent problem in emergency departments at public hospitals: a lack of accurate information about how long patients must wait to see a doctor.
“You often hear about people experiencing very long waiting times,” the research fellow at UQ’s Centre for the Business and Economics of Health (CBEH) said. “I went a bit deeper and tried to understand what’s happening there.”
At a glance
- Long and vague waiting times in emergency departments lead to patient dissatisfaction.
- Researchers at CBEH have teamed up with the Princess Alexandra Hospital to develop a system that can provide accurate waiting time forecasts.
- The method relies on machine learning algorithms that give patients and staff in emergency departments near real-time estimates
- More accurate waiting time predictions could benefit both patients and hospital networks.
While waiting time prediction varies across different states and hospitals, most emergency departments rely on simple rolling averages, which give only a rough estimate of when patients can expect to be treated. Reliant on often inaccurate waiting time forecasts, this traditional approach doesn’t capture the unpredictable fluctuations emergency departments experience every day.
As a result, most emergency departments don’t report forecasted wait times to the public, which can lead to anxiety, uncertainty, and dissatisfaction among patients. According to Associate Professor Andrew Staib, Deputy Director of Princess Alexandra Hospital, around 3% of patients waiting in emergency departments leave before they’re seen by a doctor.
“That’s quite a lot,” he said. “In this environment, it’s very easy to feel like you’ve been forgotten or that you’re not important.”
Applying machine learning to forecast wait times
Dr Pak and Dr Staib are harnessing machine learning algorithms to forecast emergency department wait times in near real-time. Instead of relying on data from the past several hours, the system draws on a substantial dataset of patient movements to more accurately predict how long they’ll spend in the waiting room.
In 2021, the team tested their AI-driven algorithms on a dataset of roughly 120,000 low-acuity patients who had visited the emergency department at Princess Alexandra Hospital over two years. The dataset captured each patient’s journey from the moment they arrived to when they were discharged.
The results of their study showed the proof-of-concept system predicted wait times 30 to 40% more accurate than the rolling average estimates currently used in hospitals. “That’s quite substantial,” Dr Pak said.
In addition to improving patient experience, better wait time predictions could help non-emergency patients make more informed decisions about accessing the health care they need. According to Dr Pak, this would not only benefit all patients, but the whole system.
“All hospitals on average would be less busy,” he said. “That’s a great thing, because they can use those scarce resources for more urgent patients.”
Next steps
The researchers hope to develop their system into a device that provides patients and hospital staff with a snapshot of estimated waiting times, such as a mobile app or display screen installed in emergency departments. A big part of designing the tool will involve gathering insights from consumer groups to decide which approach works best for delivering information to patients.
Dr Staib noted working with researchers from CBEH has been a valuable part of developing solutions that could provide tangible benefits to patients. The collaboration has also helped connect the emergency department with the right expertise to tackle the challenge of accurately predicting wait times.
“The success comes down to the individuals and enabling them to do this kind of work,” Dr Staib said. “It’s a way of bringing people into the space that may not otherwise have come into it.”
Research Publications
Dr Pak’s research has been published in Statistics in Medicine and the International Journal of Medical Informatics.
Contact Dr Anton Pak to discuss potential partnership opportunities at a.pak@uq.edu.au.
This article has been updated to include the latest publications from Dr Anton Pak.
