The future of CBT and evidence‐based psychotherapies is promising

Stein et al ¹ point out that, while evidence‐based psychotherapies and particularly cognitive behavioral therapy (CBT) represent a “remarkable step forward”, their implementation in mental health systems globally is “arguably best conceptualized as representing incremental progress”.

Modest implementation is tied to several factors, including incompatibility with other psychotherapeutic models, frequent departure from evidence‐based guidelines in routine care, and lack of trained clinicians. Further, even with embedded training in evidence‐based therapies, as exemplified by the UK Improving Access to Psychological Therapies (IAPT) program, the authors report that rates of clinically significant improvement are estimated at only 26% when assuming poor treatment response among dropouts ¹ .

In line with 2004 modeling to suggest that universal provision of evidence‐based practices will reduce the global disease burden by only 40%, Stein et al ¹ raise the specter that the burden of mental disorders will never be significantly reduced. In further support of this bleak outlook, they refer to the treatment‐prevalence paradox of increased treatment uptake without corresponding reductions in population prevalence rates (as documented for depression).

Herein, I argue that a more promising future of CBT and other evidence‐based psychotherapies is achievable through: a) more mechanistically targeted interventions, that b) are personalized or matched to individuals and c) are scaled with fidelity by harnessing technology.

The majority of randomized controlled trials (RCTs) to date evaluate CBT packages of multiple elements (e.g., cognitive restructuring, relaxation, exposure), designed for individuals classified according to diagnostic nosologies. Yet, within a set of therapeutic elements, some are likely to be more effective than others for a given individual, increasing the risk of iatrogenic effects, inefficiency, and treatment dropout. Moreover, diagnostic categorization for treatment selection ignores the substantial heterogeneity within diagnoses (e.g., within post‐traumatic stress disorder, some people experience numbing and dis­sociation whereas others suffer from heightened emotional arousal). Transdiagnostic symptom dimension models, such as hierarchical latent structural models and symptom network approaches, promise greater precision in personalization of mental health care. Shifts towards treatment elements rather than packages, and symptom dimensions rather than diagnoses, will enable more targeted interventions that are more effectively matched to individuals. Evidence in support of prescriptive matching to specific treatment elements is beginning to emerge ² .

A treatment elements approach also a­ligns with targeting specific dysregulations in physiology, cognition, behavior or emotion that correlate with or contribute to psychopathology. Exemplars include advances in neuroscience and behavioral science of fear extinction, that have led to refinements of exposure therapy for fear and anxiety symptoms ³ . Corresponding advances in the area of reward processing have led to treatments that target reward hyposensitivity for anhedonia symptoms across anxiety and depressive disorders ⁴ . Feedback from evaluation of target engagement can then inform iterative intervention refinement.

With moderated mediation approaches, we may further learn that mediators (as measures of purported mechanisms) have differential relevance across persons. As an illustration, prediction error generalization may be a stronger driver of exposure therapy effects for some people, whereas re‐appraisal of feared outcomes may be more relevant for others, such that different versions of exposure therapy may be tailored for each individual. Consequently, theoretically relevant features of responding could be matched to targeted interventions more precisely and thereby more effectively, as a step beyond moderation based on standard features of clinical presentation (e.g., symptoms and functioning).

Advances in the mechanisms contributing to psychopathology, continuing development of intervention elements that specifically target mechanistic features, along with prescriptive algorithms for selecting the right intervention for a given person, represent an enormous research agenda, but one that is nonetheless underway, with the US National Institute of Mental Health's emphasis upon experimental therapeutics for clinical trials and the recent Wellcome Trust initiative of “Finding the next generation of mental health treatments and approaches”.

Alongside the development of more tar­geted and personalized intervention elements, technologies can facilitate screening and triaging to the type of care predicted to be most effective, with rapid adaptation of care as needed, for more scalability and more effective outcomes.

Online screening and tracking of mental health status and related variables is suitable for large scale deployment, particularly adaptive testing which increases measurement precision and minimizes participant burden relative to traditional fixed length instruments ⁵ . Automated feed­back from scoring algorithms can then guide treatment selection. Prescriptive treatment selection algorithms generated from machine learning or other modeling of an array of relevant data may improve over­all outcomes relative to standard clinical ­decision making, as has been demonstrated when selecting between low‐intensity versus high‐intensity care within IAPT using a limited range of predictive variables (i.e., symptom severity, impairment, personality traits, employment status, race/ethnicity) ⁶ . As mentioned, theoretically relevant variables (e.g., emotion regulation, response inhibition, and threat expectancy) may enhance accuracy of treatment response prediction for specific treatment elements (versus levels of care).

Rather than adapt level of care after a patient shows non‐response or prematurely discontinues treatment (as is typical in stepped care models), ongoing predictive modeling can facilitate adaptation to higher levels of care or to different therapeutic elements before failure occurs. This just‐in‐time treatment approach has the potential to improve effectiveness and re­duce attrition, as patients may be more engaged in treatment when they are receiving what they need most at the time they most need it. Adaptive interventions can also increase the efficiency of service delivery and reduce downstream service costs. Furthermore, adaptation extends to maintenance goals, so that care can be rapidly reinitiated upon signs of symptom worsening to prevent full relapse.

Task‐sharing through non‐specialized providers is a cost‐effective strategy for scalable mental health care ⁷ , but is challenged by scalability of training and su­pervision and by fidelity assurance (adherence and competency). Digital tools can address these issues, such as training courses with interactive feedback for skill development and ongoing competency evaluations, as well as computerized session guides to maintain fidelity ⁸ .

Digital CBT and other evidence‐based psychotherapies via phone, computers and other electronic devices increase access to care, and overcome barriers of stigma, financial difficulties, time constraints, and location of services. The available ev­idence clearly supports their efficacy, although more research is needed in low‐ to middle‐income countries. Digital therapies are particularly suited to the research agenda of prescriptive algorithms for selecting specific intervention elements most likely to benefit an individual. Yet, user uptake, engagement and dropout are problematic, especially in routine clinical care settings. Since human support mitigates these concerns ⁹ , models that combine non‐specialist providers with digital interventions have unique potential to expand reach, engagement and effectiveness.

Mechanistically targeted and personalized intervention elements that are match­ed to individual needs and adapted as needs change over time, delivered digitally or by clinicians, that can be scaled up through online tools and artificial intelligence technologies, offer a future in which delivery of evidence‐based care will reduce the global disease burden of mental health by more than 40%. Challenges include the enormous research agenda for developing mechanistically targeted interventions and their prescriptive matching to individ­uals.

Implementation will continue to be chal­lenged by transportability of digital technologies into under‐resourced areas, lack of resources for the most severely ill, and cultural adaptations to avoid simple exportation of Western constructs. Whether systems will choose to endorse evidence‐based psychotherapies, in spite of the view that they are overly reductionistic or do not address complex refractory or comorbid cases, will most likely depend upon the success of that implementation.