The system is currently being rolled out across South Australia and may be incorporated across the country in the coming years.
A new AI-powered penicillin allergy de-labelling system is being rolled out across South Australia, with plans to roll it out across the country in coming years.
Although 18-25% of people report that they are allergic to penicillin, evidence shows that around 90% should be de-labelled.
This over-reporting is a major public health concern, contributing to antimicrobial resistance, inappropriate prescribing and even mortality.
Currently in Australia, most drug allergy testing is performed in specialist outpatient clinics or in hospitals.
The new de-labelling system aims to save patient and health system time and resources by using AI to cross reference patients’ electronic medical records against evidence-based data.
This allows the system to determine mislabeled allergies, said lead project researcher and neurologist Associate Professor Stephen Bacchi, at the University of Adelaide.
“The artificial intelligence algorithm has been trained to read descriptions of allergies and intolerances and classify individual interactions based on these previous examples,” he said.
“It can then determine if patients may be incorrectly labelled as allergic to penicillin.
“We flag this with the patient’s hospital doctors. Following discussion with the patient, in many cases, the label is then able to be removed.”
The system was trialed at the Lyell McEwin Hospital and is being led by the Royal Australasian College of Physicians.
It is expected that the SA rollout will be complete by early 2025 and will be expanded across Australia in the coming years.
Professor Bacchi said the current de-labelling process was arduous.
“The current allergy labelling process relies heavily on self-reporting and assessing patients’ immediate response to the antibiotic,” he said.
“These measures aren’t always accurate. Many patients labelled as allergic to penicillin are intolerant to the antibiotic.
“While allergies and intolerances may both present unpleasant symptoms, they are very different.
“If a patient is intolerant to penicillin, they can typically still receive the antibiotic. This is not true for allergies.”
Professor Bacchi said the AI system would help to ease strain on the healthcare system.
“Adding the penicillin allergy label to a patient’s medical records can lead to several knock-on effects,” he said.
“Since penicillin may be contraindicated in some patients with allergy labels, healthcare workers must spend time sourcing alternative antibiotics.
“This can lead to patients having to spend longer at the doctor or hospital, and increased costs for both the health system and the patients.
“By de-labelling incorrect penicillin allergy labels, we can help save time and resources for patients and for health systems.
“Many healthcare systems across the country are already strained, and processes like these can go a long way in ensuring that some of that burden is eased.”
When asked about how GPs would be kept in the loop on changes to allergy labels, Professor Bacchi said the best way to communicate regarding penicillin allergies was an area of ongoing research.
“For work, such as delabelling, that occurs in an inpatient setting, the patient’s discharge summary would be the best place to obtain information,” he said.
“However, asking the patient themselves is also always vital.
“In some instances, referrals will be made to outpatient immunology clinics, in which case further correspondence from these clinics will also provide useful information.”
