Enhancing Multisystemic Therapy (MST) Effectiveness: A Novel Approach Using MY-SCOPE and Machine Learning for Within Session Communication Analysis - Project Summary/Abstract Adolescents exhibiting serious externalizing symptomatology, represent large and underserved populations that are at high risk of presenting significant deleterious outcomes and long-term costs for themselves, their families, communities, and society. Evidence for the effectiveness of family-based treatments for adolescents presenting such serious clinical problems (i.e., substance abuse, serious and chronic delinquency) has led to efforts to disseminate these and other evidence-based treatments (EBTs) into real-world practice settings. While model-specific training, supervision, and fidelity monitoring has been a major feature of current strategies to transport and implement a variety of EBTs for children and adolescents into community practice settings, where most children are treated, even the most extensively studied and effective EBTs are not universally effective with every family. One of the most widely studied and disseminated EBTs targeting adolescents with serious clinical problems, Multisystemic Therapy (MST), has shown large effect sizes in efficacy studies conducted in controlled settings, but like other EBTs, its efficacy in real-world practice settings remains far from optimal, especially for underserved, multi-problem and multi-stressed families. The modest success of EBTs in real-world settings can, in part, be attributed to the emphasis placed on therapists achieving treatment fidelity during implementation rather than providing guidance on how to manage the complex, nuanced interpersonal communication processes that inevitably occur during the delivery of EBTs. A largely ignored area of research in implementation science has been the examination of therapists’ skills and competencies in managing interpersonal communication processes at the heart and soul of behavior change. Recent advancements in multivariate time-series analyses and machine learning allow scientists to make significant advancements in the study of psychotherapy processes by examining therapist-client communication sequences that are differentially associated with class membership in treatment response trajectories. Using audio session recordings, taken from a secondary data source, a completed NIMH-funded longitudinal effectiveness study of MST, we will use supervised and unsupervised machine learning techniques, coupled with a validated observational behavioral coding scheme (Minority Youth Sequential Code for Observing Process Exchanges; MY-SCOPE) as labels to train algorithms, to recognize communication patterns associated with class membership in treatment response trajectories. Identifying communication sequences using MY-SCOPE and machine learning, the proposed R21 study seeks to provide actionable recommendations for refining MST training materials and advance understanding of therapist-caregiver-youth communication sequences associated with MST outcome.
