AI tool is helping doctors prescribe antidepressants more accurately
Finding the right antidepressant can be a long and frustrating journey for millions of people around the world. Now, an artificial intelligence tool developed by the University of Oxford is showing that this process can be a lot more efficient. The system, called PETRUSHKA, was tested in a large international clinical trial, and the results show that doctors can pick the most suitable medication for each patient right from the first attempts, cutting down the exhausting cycle of trial and error that has dominated psychiatry for decades.
The study was published in the Journal of the American Medical Association and is considered the first time a clinical prediction tool in mental health has demonstrated proven effectiveness in a controlled trial of this scale. The research involved 500 adults diagnosed with Major Depressive Disorder, spread across 47 centers in the United Kingdom, Brazil, and Canada, which gives the study significant global relevance.
How PETRUSHKA works in practice
PETRUSHKA is not just another generic recommendation algorithm. Its name stands for Personalising Antidepressant Treatment for Unipolar Depression Combining Individual Choices, Risks and Big Data, and it was designed with a very clear purpose: to cross-reference all available clinical data on depression with each patient’s personal information to recommend the best possible antidepressant from the very start of treatment.
In practice, the way it works is relatively straightforward from the user’s perspective. The patient answers a series of detailed questions about their health history, past experiences with medications, symptom severity, and perhaps the most distinctive part, which side effects they are or aren’t willing to tolerate on a daily basis. It could be weight gain, insomnia, sexual dysfunction, or any other effect that directly impacts quality of life. The artificial intelligence then processes this information alongside a massive database built from clinical studies and real patient records from around the world.
The result is a personalized recommendation that the doctor can use as support for their clinical decision. It’s not about replacing professional judgment but rather adding an extra layer of data-driven information that can make the choice more accurate right from the first prescription.
This last point is critical because, when it comes to treatment for depression, medication adherence is one of the biggest challenges faced by both healthcare professionals and people living with the disorder. Estimates suggest that around 80% of the millions of people who receive antidepressant prescriptions end up stopping their medication within a few weeks. Often this happens because of unexpected side effects or simply because the drug didn’t bring any noticeable improvement within the expected timeframe.
PETRUSHKA aims to tackle this bottleneck by putting the patient at the center of the decision, turning what used to be a nearly random process into one guided by data and the preferences of the person who truly matters in this equation.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
Henry Winchester’s story and his experience with the tool
Henry Winchester, 45, from Bristol, England, knows firsthand the weight of living with depression for years on end. Since his college days, he has faced recurring episodes of the disorder that have deeply affected his daily routine.
In his own words, the experience has been very challenging at many points. He describes struggling with intense anxiety and periods of very low mood, which end up limiting what he’s able to do in life. According to him, there are so many things he’d like to accomplish, but he feels held back by his own brain.
Before joining the PETRUSHKA study, Winchester spent five years trying different medications without success. The side effects were too severe and kept leading him to drop the treatment repeatedly — a cycle that anyone who’s been through something similar will recognize as incredibly draining.
When he entered the clinical trial, he answered a detailed questionnaire with numerous questions about his health and, most importantly, about which side effects he didn’t want to deal with. Winchester described the process as quite interesting and said there was an exciting moment when the tool finally revealed which antidepressant would be the best fit for his profile.
The impact was remarkable. He says he now feels much more optimistic because many of the things that used to bother him don’t affect him nearly as much anymore. Beyond that, he says he’s developed an inner confidence he had never experienced before. For someone who spent years fighting side effects and the ineffectiveness of medications prescribed the traditional way, this change represents a major transformation in quality of life.
Results of the international clinical trial
To test whether PETRUSHKA actually delivers on its promise, the researchers conducted a robust randomized clinical trial. The study split the 500 participants into two groups: half received the antidepressant recommended by the artificial intelligence tool, while the other half followed the conventional prescribing process, where the doctor selects the medication based on their own clinical experience and traditional guidelines.
Professor Andrea Cipriani from the University of Oxford was the lead investigator. He also serves as an honorary consultant psychiatrist at the Oxford Health NHS Foundation Trust and oversaw the research conducted across the 47 centers worldwide.
According to Cipriani, the tool represents a revolutionary approach because, until now, prescribing antidepressants for people with depression essentially relied on trial and error based on the doctor’s experience. He emphasizes that this trial-and-error process is not only time-consuming but also harmful to patients, who remain exposed to unsuitable medications for extended periods.
The numbers speak for themselves. Cipriani stated that the tool’s results are remarkable because, when used, PETRUSHKA increases the likelihood by 40% that the patient will continue taking the prescribed medication. And staying on the medication is a direct indicator that the drug is both effective and tolerable at the same time.
Beyond improved adherence, participants in the AI-assisted group also showed a more significant reduction in depressive and anxiety symptoms over the weeks of follow-up. This finding is particularly important because it shows the tool isn’t just about getting patients to keep taking their meds — it’s about making sure the medication is actually working and improving the person’s life.
The data published in the Journal of the American Medical Association marks the first time a clinical prediction tool in mental health has demonstrated proven effectiveness in this way. This is a milestone that could open doors for applying similar models to other areas of psychiatry.
The importance of geographic diversity in the study
Another aspect worth highlighting is the geographic diversity of the participants. By including centers in Brazil, Canada, and the United Kingdom, the researchers were able to test PETRUSHKA’s effectiveness across populations with quite different genetic, cultural, and socioeconomic profiles. This strengthens the validity of the results beyond a European context and suggests the tool can work well across different public health realities.
This point is especially relevant when looking at the Brazilian landscape. Brazil has one of the highest rates of depression in the world, according to World Health Organization data. At the same time, access to specialized psychiatrists is still very limited across much of the country, particularly in regions far from major urban centers. A tool that helps optimize antidepressant selection could have a significant impact in settings where every medical appointment counts and where patients often don’t have the option of quickly returning to a specialist to adjust their medication.
What this means for the future of depression treatment
The arrival of PETRUSHKA on the clinical scene represents a paradigm shift in how medicine approaches depression treatment. For decades, choosing an antidepressant has basically worked on a trial-and-error model. The doctor prescribed a medication, the patient took it for a few weeks, and if there was no improvement or the side effects were too much, they moved on to another option. This cycle could repeat for months or even years, generating frustration, distrust in the treatment, and in many cases, complete abandonment of drug therapy.
With an artificial intelligence tool capable of personalizing the recommendation from the very first moment, this scenario could start to change significantly. The process becomes more efficient and also more humanized, since the patient’s preferences are taken into account in a structured way rather than just as an informal comment during the appointment.
It’s important to stress that PETRUSHKA doesn’t intend to and shouldn’t replace the role of the psychiatrist or the doctor who follows the patient. The idea is for it to function as a clinical decision-support tool, offering the professional a data-driven analysis that complements the knowledge and sensitivity that only a human being can bring during a consultation. This balance between technology and human care is precisely what makes the approach so promising. Instead of creating a scenario where the machine decides everything, the system strengthens the doctor-patient relationship by including the person’s preferences and concerns in the decision-making process.
Next steps and expansion to other disorders
Looking ahead, the Oxford researchers have already signaled their plans to expand PETRUSHKA’s application. The team led by Cipriani aims to make the tool available to general practitioners across the United Kingdom, which would enormously broaden the system’s reach beyond specialized psychiatry centers. If that happens, patients who currently rely on general practitioners for their first antidepressant prescription could immediately benefit from a more precise and personalized recommendation.
There’s also the prospect of similar artificial intelligence models being applied to other mental health disorders, such as generalized anxiety and bipolar disorder, where medication selection also involves a significant degree of uncertainty. The logic is the same: the more relevant information is available at the time of the decision, the greater the chance of getting it right on the first try.
If this trend holds, we’re looking at a moment when technology may finally help solve one of the most persistent problems in modern psychiatry: making sure each person receives the right treatment, at the right dose, from the very beginning. Currently, the process can take years, as Henry Winchester’s own story shows. Five years trying different medications before finding something that worked is a long time in anyone’s life, especially for someone dealing with a disorder that already steals energy and hope every single day.
For the millions of Americans and people around the world living with depression every day, the possibility of having tools like PETRUSHKA isn’t just promising. It’s something that can concretely transform the way treatment is carried out, making it faster, more humane, and above all, more effective 💙
