Imma Perfetto
Cosmos science journalist
A new artificial intelligence-driven approach to more efficiently free up hospital beds will likely be rolled out across several South Australian hospitals within the year, with plans to implement it interstate in the next 18 months.
The system provides a prediction of the likelihood that the patient will be discharged within 24 hours.
Known as the “Adelaide Score,” because it gives a number between 0 and 100, the artificial intelligence program predicts the length of a patient’s stay in hospital, helping healthcare staff prioritise tasks to streamline the discharge process.
A prospective trial in April 2024 at the Lyell McEwen Hospital, a public hospital in Adelaide’s northern suburbs, found the program significantly reduced the average length of hospital stays and rate of readmissions.
Its developers hope that by increasing patient flow through hospitals it will help alleviate ambulance ramping and pressure on frontline healthcare professionals, a problem plaguing health systems throughout Australia and elsewhere.
“The Australian healthcare system is wonderful, but it’s under strain,” Dr Stephen Bacchi, a neurologist and associate professor at the University of Adelaide, told Cosmos.
“Australia can be a world lead here … I think it’s very exciting time for AI and I think it can really provide patient benefits, and we want to deliver that.”
Streamlining patient flow
The Adelaide Score is trained to take objective data from a patient’s electronic health record – such as vital signs and blood test results – and predict the likelihood a patient will be ready for discharge within 24 hours.
The prediction is not provided to inpatient surgical and medical teams and does not influence clinical decisions around medical fitness for discharge. Instead, it can be used to get the ball rolling on other logistical processes involved in readying a patient to return home.
Bacchi and Dr Joshua Kovoor, general surgery principal house officer at Gold Coast Hospital and Health Service, developed the Adelaide Score in collaboration with a multidisciplinary health team.
“If you need a wound care plan after your operation, or if you need rehabilitation booked … somebody has to do that,” Kovoor told Cosmos.
“And if that isn’t done, the moment that the consultant surgeon decides the patient is fine to go home, then the patient will be delayed in hospital.”
When a ward bed is occupied for longer than necessary, a patient in the emergency department (ED) cannot be transferred, which further prevents another patient from being admitted to the ED, increasing wait times.
“It’s been shown time and time again that adding a lot more beds usually doesn’t solve this issue, because you just sort of fill that demand with the infinite demand of healthcare services,” says Kovoor.
“What actually improves this issue is taking the resources that are already there and using them in the most strategic way possible.”
Reducing the length of patients’ stays
In practice, the Adelaide Score is used alongside a second AI – the Rosetta System – by the Supportive Weekend Interprofessional Flow Team (SWIFT).
SWIFT, comprised of a senior nurse with extensive knowledge of the local workings of the hospital and a pharmacist, helps facilitate discharges from Thursday through Sunday.
“Health is a 24/7 issue, but hospital systems are staffed by people, and inherently this means there are staffing differences overnight and on weekends,” says Bacchi.
“That means that weekend discharges can become a bottleneck, and there are fewer discharges on weekends.
“The way the algorithms work is it flags patients to the SWIFT team, who could then help liaise with the medical team and facilitate their discharges.”
They found that while the intervention significantly decreased patient length of stay in general, perhaps counterintuitively, it increased slightly for discharges on weekends.
“We think the reason for that is people who had been in hospital for a while, who would not have been discharged otherwise until Monday, were actually getting discharged on weekends,” says Bacchi.
The Adelaide Score also decreased the readmission rate at 7 days, which may be due to a more thorough discharge process before the patient leaves the hospital.
According to Kovoor, the initial problem that brings a patient to the hospital often it isn’t what causes them to be readmitted soon after being discharged.
“A lot of the times it’ll be something secondary to the thing that has been definitively treated, like their pain has gotten out of control, or their wound care plan, or they can’t take care of themselves at home,” says Kovoor.
The team plans to trial the approach again in the coming months to assess its efficacy throughout the entire week.
Plans to expand
The Adelaide Score is currently in use at another northern suburban hospital, Modbury in South Australia, but Bacchi says they have plans to expand to others very soon.
“I think it is most likely that it will be operating in the Royal Adelaide and Queen Elizabeth [Hospitals] before the end of the year, although there are a couple of steps that remain to be sorted,” he says.
They are also in discussion with multiple hospitals in Victoria, New South Wales and Queensland.
“The precise timeline for those will vary based on some local factors,” says Bacchi. “But we are optimistic that … within 12 to 18 months, it might be running interstate.”
This will require continuous evaluation, as well as engagement with local expertise, to identify how the AI system can be safely and effectively implemented in healthcare settings that may operate differently to where it was initially developed.
This includes plans to trial the Adelaide score and other AI systems in rural Australia.
“We are passionate about delivering evidence based artificial intelligence to rural Australians in an equitable manner to improve their healthcare outcomes,” says Bacchi.
