I wrote recently about a fascinating project from the University of Alberta to use artificial intelligence to automate the diagnosis and treatment of mental health disorders.  It’s part of a wider body of work that sees data and AI used to better understand and treat mental health problems.

The latest of these was documented in a recently published paper from a team of Harvard researchers who utilize artificial intelligence to determine which patients are best suited to antidepressant medications.

The researchers collected data from a recent clinical trial of antidepressant medications.  The trial, called Establishing Moderators and Biosignatures of Antidepressant Response in Clinical Care (EMBARC), was conducted across four sites (Columbia University, Massachusetts General Hospital, the University of Michigan, and UT Southwestern Medical Center), and the data was used to develop an algorithm to predict the number of individuals who could benefit from the medication.

The right medicine

The study found that most people saw little difference in outcome when randomly assigned to the medication vs placebo groups, but there were clear differences for one-third of participants.  The results suggest that this group were better suited to the antidepressants, and therefore had a much better outcome as a result.  This group were characterized as having more severe depression with negative emotionality.  They were also older and more likely to be employed.

“These results bring us closer to identifying groups of patients very likely to benefit preferentially from an SSRI and could realize the goal of personalizing antidepressant treatment selection,” the researchers say.

The next step is to try and adapt the algorithm so that it’s usable in real world scenarios.  They will also be partnering with the University of Pennsylvania to conduct a trial in a psychiatric clinic, with two or more viable treatments tested to see just how nuanced the algorithm can get.

“Our mission is to use these data-driven algorithms to provide clinicians and patients with useful information about which treatment is expected to yield the best outcome for this specific individual,” the authors conclude.

There has been a lot of discussion and attention given to the development of personalized medicine in healthcare, but this is one of the first attempts to do likewise for mental health.  It’s a welcome step in the right direction.