AI TOOL HELPS DOCTORS PREDICT WHEN PATIENTS CAN SAFELY STOP ANTIDEPRESSANTS

New machine learning models from the University of South Australia could help clinicians judge when long-term antidepressant users can safely stop their medication.

The research utilised Pharmaceutical Benefits Scheme dispensing data to identify patterns associated with successful withdrawal among 100,000 patients over a ten-year period.

With antidepressant use high in countries including Australia, Iceland, Portugal, Canada and the UK, the UniSA team says artificial intelligence can support general practitioners in deprescribing when ongoing treatment is no longer clinically recommended.

Long-term use can carry side effects such as weight gain, sexual dysfunction and cardiac issues. Yet, half of patients experience withdrawal symptoms when they stop, underscoring the need for better decision support.

“Healthcare providers are often reluctant to cease antidepressant prescriptions due to their concerns about withdrawal effects, making it difficult for doctors to know who can safely discontinue treatment,” Dr Ranwala says.

“By applying AI to the PBS database, we have identified patterns linked to successful withdrawal, forecasting which patients are most likely to succeed when taking them off antidepressants.”

In the study, successful deprescription meant no antidepressant medicines for at least one year after more than 12 months of prior use.

If the medicine strength increased within six months of a reduction attempt, it was recorded as a failure.

Two models were trained and tested.

One assessed final prescription records and achieved an accuracy rate of 81%. A second tracked patients from their first prescription, following dose reductions and outcomes, and reached 90 per cent accuracy.

The researchers say the more granular model better reflects real-world prescription attempts and could give clinicians greater confidence to initiate withdrawal where appropriate.

“These results show real promise,” says UniSA co-author Associate Professor Andre Andrade.

“The most accurate model was the one that offered a more nuanced picture of deprescription attempts, better reflecting patient experiences,” Assoc Prof Andrade says.

The team adds that administrative health data can help predict clinical outcomes and improve medical decision-making.

“This data is passively collected, underused by medical professionals and a good candidate for AI use.”

Next steps include improving accuracy and usability, testing the tool in clinical settings, and exploring similar approaches for other medicines.