As psychotherapists, we face the difficult task of understanding a person’s presenting psychological problems, conceptualizing how their thoughts, behaviors, and emotions interact to produce psychological distress, and choosing intervention techniques accordingly. Typically, clinicians use DSM criteria to assign patients one or more diagnostic labels, which theoretically should inform the treatment approach. For example, most cognitive behavioral therapy (CBT) treatment protocols have been developed for specific diagnoses. However, there is considerable heterogeneity in symptom expression within diagnostic categories. For example, anxiety and depressive disorders can be expressed very differently from person to person, and these differences can have important practical implications for functional impairment and treatment (Nelemans et al., 2014; Unick, Snowden, & Hastings, 2009). Thus, planning effective and efficient treatment may be more complicated than simply matching a person to a protocol.

Recently, a transdiagnostic approach to understanding and treating psychopathology has gained popularity, focused on developing and testing treatment protocols that target shared mechanisms underlying different diagnoses (Sauer-Zavala et al., 2017). In other words, transdiagnostic protocols can be used with individuals presenting clinically with a range of diagnoses because they target common processes that go deeper than symptoms, such as experiential avoidance and cognitive inflexibility. Despite this advantage, some psychologists have raised concerns that the transdiagnostic approach may obscure important differences between individuals with different symptom presentations (Harvey, Watkins, Mansell, & Shafran, 2004).

This begs the question: is a transdiagnostic protocol—one that can be used to treat social anxiety just as readily as it can treat obsessive-compulsive disorder, for example—simply a one-size-fits-all approach? Or can such a protocol offer a targeted, personalized approach to psychotherapy?

One example of a transdiagnostic treatment protocol is the Unified Protocol for the Transdiagnostic Treatment of Emotional Disorders (UP), developed at the Center for Anxiety and Related Disorders at Boston University (Barlow et al., 2017). This cognitive behavioral protocol focuses on correcting unhelpful emotion regulation strategies that are common across depression, anxiety, and related disorders. The UP teaches emotion regulation skills that are organized into distinct modules. A benefit of the UP is that each core module targets a different psychopathological process thought to underlie the full range of emotional disorders. Collectively, these emotion regulation skills are meant to alter unhelpful patterns of responding to emotional experiences (e.g., negative evaluation of strong emotions, suppression or avoidance of emotions). In addition to these core modules, there are modules focusing on enhancing motivation and maintaining treatment gains after treatment ends. Thus, the UP contains a range of distinct, empirically supported intervention strategies. This modular approach affords a flexibility that lends itself to personalization based on the needs of each client; theoretically, these modules can be selected and organized on a person-by-person basis.

Decisions to make personalized modifications to a course of therapy should follow a strong, individualized assessment of the patient’s presenting problems. Recently, researchers have been working to develop empirical, data-driven techniques for assessing individual differences in psychopathology that may be relevant for treatment. It is believed that these individual differences could determine what therapeutic interventions are required and how to prioritize their delivery. Some researchers are capitalizing on the UP’s modular format to study the best ways to personalize its delivery.

One approach is based on assessing patient’s preexisting strengths and weaknesses. Patients differ in the extent to which treatment skills (e.g., flexible thinking, mindfulness) may come easily to them. Research findings, mostly relating to depression, are mixed as to whether targeting relative strengths or deficits is more helpful (Cheavens, Strunk, Lazarus, & Goldstein, 2012). This approach was tested using the UP by assessing patients’ skill level for five emotion regulation strategies corresponding to five core UP modules, and then having the patients start treatment with modules that corresponded with either their strengths or weaknesses (Sauer-Zavala, Cassiello-Robbins, Ametaj, Wilner, & Pagan, 2019). Patients experienced equivalent symptom reduction regardless of whether the module order was consistent with their strengths or weaknesses, but those who started off with content focusing on their strengths saw faster improvement. Thus, one potential way to personalize treatment is to capitalize on patients’ existing strengths earlier in treatment.

Another approach is based on assessing how a patient’s symptoms relate to one another. The idea is that some symptoms may be directly caused by other symptoms, such as insomnia leading to loss of energy in depression, so therapists may be able to prioritize targeting the symptoms closest to the “root.” Initial efforts to measure these symptom dynamics included the Dynamic Assessment and Treatment Algorithm (DATA), which developed personalized transdiagnostic treatment plans for individuals with either major depressive disorder (MDD) or generalized anxiety disorder (GAD; Fernandez et al., 2017). DATA involves thirty days of ecological momentary assessment (EMA), or administering frequently-repeated symptom surveys to patients as they go about their lives. The results of the symptom surveys are then analyzed through a series of statistical analyses that group co-occurring symptoms into meaningful clusters and then determine which clusters of symptoms predict others at later time points (or, which symptoms may be at the “root”). Then, the algorithm creates treatment plans that prioritize treatment activities from the UP that match these “root” symptom clusters. A recent open trial offered preliminary evidence that DATA-generated treatment plans can reduce anxiety and depression symptoms for individuals diagnosed with either MDD or GAD (Fisher et al., 2019). However, it remains unclear whether these treatment plans outperform delivering the UP in standard order.

Despite this promising early work, there are a number of unanswered questions about the most effective ways to assess individuals for factors that will be relevant to treatment planning. For example, should clinicians be assessing symptoms, strengths and weaknesses, or some other target construct? How much assessment is necessary to make meaningful, informed treatment decisions? What types of statistical analyses are most useful? Clearly, there is much that we don’t yet know. Bastiaansen et al. (2019) assembled twelve research teams who specialize in idiographic assessment and provided them with EMA symptom surveys from a sample patient. Each team was asked to identify the treatment target they would prioritize during treatment. There were several major differences between the teams’ approaches, including which symptoms were included in their analyses, the methods used to group symptoms into clusters, and the analyses used to ultimately select the most important target. These differences led to the teams identifying different targets and recommending different treatment approaches. Notably, despite disparate data analytic strategies, the teams all made reasonable, justifiable decisions; these results highlight the lack of consensus on best practices for analyzing idiographic data to generate individualized treatment plans.

Moving forward, it will be crucial for researchers to direct their efforts where research is most needed. First, as mentioned above, we must determine what constructs to assess: symptoms, underlying mechanisms, or something else? As developed, DATA generates treatment plans based on assessment of GAD and MDD symptoms. For a truly transdiagnostic algorithm, symptoms of other emotional disorders (such as panic disorder or obsessive-compulsive disorder) would also need to be included in each assessment, which may make the algorithm too unwieldy by including a very large number of items. Instead of focusing on symptoms, targeting underlying processes may be more parsimonious, as these processes are shared across disorders.

Second, researchers must consider assessment timing. Dynamic person-level statistics require data to be collected across many time points; thus, most methodologies involve administering surveys multiple times per day for a period of several weeks. On one hand, a treatment waitlist may be an ideal time to collect data that can be used to plan treatment once a slot opens up. On the other hand, this type of data collection is ultimately a heavy burden for patients and may present a problem during data analysis if patients are not compliant. In addition to this practical consideration, it is also important that assessment timing make theoretical sense. In other words, do the hypothesized relationships between the measured constructs unfold on an hourly basis? A daily basis? It may be impossible to capture the full range of temporal dynamics between constructs but comparing approaches with different assessment schedules may help determine the optimal timing to use.

Assuming researchers ultimately develop an algorithm that both correctly assesses constructs relevant to an individual’s psychopathology and does so at time intervals that reveal meaningful relationships between these constructs, researchers will then still have to be able to translate these findings into treatment plans. Each target being assessed should have a corresponding treatment intervention; for example, if anxiety sensitivity, or the fear of physical symptoms associated with anxiety, is found to be a strong maintaining factor or “root” construct for a particular patient, a clinician may choose to use interoceptive exposure exercises designed to promote tolerance of uncomfortable physical sensations early in treatment. Similarly, if an algorithm highlights the importance of a patient’s sleep disturbances in predicting later interfering symptoms, a clinician may help a patient improve sleep hygiene before moving on to other interventions.

Idiographic research methods are gaining popularity due to their potential to allow for the personalization of evidence-based treatments. The UP, a modular transdiagnostic treatment, lends itself to personalization because modules can be rearranged based on patient needs. Existing research suggests that rearranging UP modules may make treatment more effective or efficient; however, future research is needed to develop best practices for making treatment decisions.

+16

-5

Andrew J. Curreri, M.A.

Andrew is a third-year graduate student in the Transdiagnostic Treatment Lab at Boston University. He is interested in understanding mechanisms of change in cognitive behavioral therapies with the goal of optimizing treatment delivery based on individual patient characteristics. He is also interested in nonspecific treatment factors such as expectancy, adherence, and therapeutic alliance. Andrew graduated from Boston University in 2014 with a B.A. in psychology and worked as a research technician in the National Center for PTSD at the VA Boston Healthcare System prior to joining the Transdiagnostic Treatment Lab.