In the ever-evolving field of psychotherapy, the spotlight is on “personalization”, a trend that is fast reshaping our understanding of mental health treatment. Terms like “precision mental health” or “tailored treatments” became ubiquitous in the research community. Bergin and Garfield’s Handbook of Psychotherapy and Behavior Change, a book that is known as the “Bible of psychotherapy research”, includes a whole chapter on this topic in the most recent edition.
Two types of personalization
When it comes to personalized psychotherapy, there seems to be a discrepancy between research and practice. Experienced psychotherapy practitioners may become confused when asked about their views on the topic. They argue that psychotherapy is already tailored to individual patient needs when performed correctly. Even if they use treatment manuals, they still adjust their methods, mix and match techniques, or skip certain parts of the protocol that don't seem to fit. However, the level of personalization varies greatly from therapist to therapist. Some may excel at it, while others may struggle. The process can change rapidly as therapists encounter new information or become interested in other methods. Although treatment guidelines exist for certain problems based on empirical research, therapists are unlikely to always follow them. Often, guidelines recommend therapy methods that a therapist is not trained in or does not feel confident using. This makes personalization based on subjective impressions and information available to individual therapists difficult to replicate. Psychotherapists are a heterogenous bunch after all: many have their preferred methods or therapy styles, or they have just attended a particular training that sharpens their view of new cases in some aspects, but also narrows it in others.
Another approach to personalization is often referred to as “precision mental health”. Here, treatment decisions are based on data and statistical models processing them. The type of treatment that is most effective for an individual patient is selected based on their characteristics. For this, you need large amounts of data from patients who were treated with various types of psychotherapy. Using statistical modeling or machine learning techniques, researchers then try to find patient characteristics that indicate a better fit for a specific method. One of the most impressive studies was led by Brian Schwartz from the University of Trier. Using data from more than 1300 patients, the researchers created a statistical algorithm that indicates whether a patient will profit more from cognitive-behavioral (CBT) or psychodynamic therapy. When they applied it to a new data set, they found that patients who received the type of treatment that the algorithm predicted to be optimal, their treatments were more effective. However, their algorithm worked mostly by successfully matching patients who will profit more from CBT.
There is a spectrum of personalization, ranging from subjective impressions to statistical algorithms. There are many levels between these two extremes. Even if therapists have a well-validated statistical model, they may still deviate from it because they use it as a tool for decision-making rather than a final decision.
What works?
Which one of these two approaches is likely to improve psychotherapy in the long run? Some empirical research suggests that the impact of “idiosyncratic” personalization that is based on individual therapists’ subjective choices might be limited. First, studies comparing decisions based on human reasoning with decisions based on data and statistical models consistently show that the latter method is more accurate. The most recent meta-analysis included studies resulting from over 56 years of research and showed that a 13% increase in accuracy can be expected when using statistical methods. This is not a huge increase, but from this, one could still predict a slight improvement in outcomes if data-driven methods are used to make decisions in psychotherapy. Additionally, studies examining the effectiveness of manual-guided psychotherapy versus non-manual-guided psychotherapy indicate minimal or negligible differences. If we assume that psychotherapists use idiosyncratic personalization more frequently when they don’t have to adhere to a manual, we could conclude that it does not lead to more effective therapies. This is in line with the finding that outcomes of clinical trials are mostly comparable to those from routine care. Third, we can look at the studies on personalized psychotherapy that are already there. A meta-analysis that appeared in the renowned Journal of Consulting and Clinical Psychology a couple of weeks ago found a small but robust increase in effectiveness if patients are matched to their optimal form of therapy.
Pitfalls
Despite these promising results, there are many reasons to be cautious. As mentioned, the improvements in effectiveness we can expect from personalized psychotherapy are small. Thus, it is important to check whether the effort to produce them is justified. When the method by which patients can be assigned to their optimal therapy is easy and inexpensive to perform, there is reason for optimism. However, if you are dealing with complex processes that generate large additional personnel costs or require special training to use, then it might not be worth the effort to personalize. Second, skeptics object that the patient characteristics commonly used for personalization vary far too much from study to study. This suggests that some or even all of them may be statistical artifacts that make generalization of the generated algorithms for personalization problematic. To resolve this, we would need much larger samples than we normally find in clinical trials. Some suggest that this could be resolved by combining data from clinical trials with routine care outcome data. Third, it is a consistent finding across studies that develop statistical selection algorithms that for a large proportion of patients, the treatment alternatives studied will be equally effective. This means that personalization may only be significantly beneficial for relatively small subgroups of patients for whom an optimal treatment method can be identified.
Conclusion
The advancement of personalized psychotherapy holds great potential, blending traditional therapeutic approaches with the capabilities of modern data analytics. It is crucial to acknowledge that while the use of treatment manuals does not inhibit the effectiveness of psychotherapy, personalizing the treatment method may enhance the therapy outcomes for specific subgroups of patients. We can assume that a personalized selection of the optimal treatment method can increase the effects for some patients. The advantage of these methods is that they are easy to automate and can be performed before the start of psychotherapy. This means that it is not necessarily the therapists themselves who have to use the statistical models, but they can also be implemented by health insurance companies to help their customers find the best method for their problems. Thus, even before the decision for a therapist is made, it is somewhat more likely that the best possible treatment option will be found.