Saving up to $6M by predicting hospital discharge
Predicting patients’ likelihood of getting discharged from the hospital has saved significant costs for a hospital in South Australia.
Lyell McEwin Hospital, a major tertiary acute care hospital in Adelaide, applied a machine learning algorithm model called Adelaide Score in a prospective trial.
Developed by a collaborative team from the University of Adelaide, the AI analyses vital signs and laboratory test data to predict potential patient discharge from hospital within the next 12 and 24 hours. It reads data from the past 48 hours, automatically collected via linkage with an EMR system.
FINDINGS
The AI system was trialled over 28 days in April last year and was utilised to evaluate the electronic records of inpatients in 18 surgical and medical teams. It screened and ranked patients who are likely due for discharge.
During this period, the hospital noted a 5% seven-day patient readmission rate, which is lower than 7.1% in the same period in the previous year. It also recorded a shorter median stay of 2.9 days, compared to 3.1 days previously.
Findings, which were published in ANZ Journal of Surgery, showed that the resulting reduction in patient admissions saved the hospital approximately A$735,708 (around $480,000) over the course of the trial.
WHY IT MATTERS
In an interview with Healthcare IT News, Dr Joshua Kovoor, the study’s first author, said the Adelaide Score was foremostly developed to provide solutions to the ambulance ramping issue in South Australia. SA Health data revealed that ambulances have spent an average 3,000 hours per month waiting outside emergency departments since 2022.
Optimising the discharge process, then, becomes an imperative to free up congested EDs. This process, however, is also tedious and lengthy; it involves organising transport, discharge medications, wound plans, and follow-up appointments.
This is where the Adelaide Score comes in. It cuts the time used to browse through electronic records to find patients close to getting discharged.
Therefore, its application in clinical settings, according to Dr Kovoor, “results in patients having to stay less in [the] hospital and require less readmissions after discharge, creating cost savings.”
The Adelaide Score can be applied in any healthcare setting worldwide that collects vitals and laboratory parameters as part of routine clinical practice. It also has potential implementation in many clinical systems that automatically links data to an EMR system.
Following the trial at Lyell McEwin, the Adelaide Score is being considered for potential implementation across eastern states of Australia. The research team is also exploring collaborative opportunities abroad. Dr Stephen Bacchi, University of Adelaide associate professor and Adelaide Score study’s senior author, said they are also in talks with “key stakeholders regarding future expansion.”
THE LARGER TREND
Besides AI, the South Australian government implemented telehealth or virtual care models to address health system congestion. Following the recent pandemic, it invested, trialled, and implemented free telehealth services dedicated to adults, children, and seniors. A 24/7 remote health monitoring service was also launched for remote and rural communities.
