✏️ AI Helps Choose Antidepressants So Patients Stick With Treatment

There are dozens of antidepressants that target different biological mechanisms linked to depression. But predicting which medication will work best for a specific patient is extremely difficult, so doctors often rely on trial and error. As a result, many patients stop taking their medication within the first few weeks, often due to side effects.

Researchers at the University of Oxford used data from 130 clinical studies to train a system called PETRUSHKA to recommend antidepressants. Doctors enter basic patient information, including age, sex, symptom severity, and any other medical conditions. Patients can also indicate which side effects they most want to avoid. The algorithm then generates a ranked list of three possible medications.

👨 In a trial involving 520 patients, after eight weeks, 17% of those whose treatment was guided by AI stopped taking their medication, compared with 27% in the control group. Among those who quit due to side effects, the numbers were 9% versus 16%. The system also helped doctors choose more effective treatments: after six months, patients in the AI-guided group showed greater improvements in depression and anxiety symptoms.

The researchers note an important limitation: the study was open-label, meaning that patients were aware whether AI was guiding their treatment, which may have influenced the results.

Should AI help choose medications?

❤️ — Yes, if a doctor confirms the recommendation
🔥 — No, that's the doctor's job

@hiaimediaen