PREACT-digital: Study protocol for a longitudinal, observational multi-center study on wearable- and EMA- based predictors of non-response to CBT for internalizing disorders

IntroductionDespite CBTs status as a first-line treatment, a substantial proportion of patients does not experience sufficient symptom relief. Recent advances in wearable technology and smartphone integration enable new, ecologically valid approaches to capture dynamic processes in real time. By combining ecological momentary assessment (EMA) with passive sensing of behavioral and physiological information, this project seeks to track daily fluctuations in symptom-associated constructs like affect, emotion regulation and physical activity. Our central goal is to determine whether dynamic, multimodal markers derived from EMA and passive sensing can predict treatment non-response and illuminate key factors that drive or hinder therapeutic change. Methods and AnalysisPREACT-digital is a subproject of the Research Unit FOR 5187 (PREACT), a large multicenter observational study in four outpatient clinics. PREACT channels state of the art machine learning techniques to identify predictors of non-response to cognitive behavioral therapy (CBT) in internalizing disorders. The study is currently running and will end in May 2026. Patients seeking cognitive behavioral therapy at one of four participating outpatient clinics are invited to join PREACT-digital. They can take part in (1) a short version with a 14-day EMA and passive sensing phase prior to therapy, or (2) a long version in which the short versions assessments are extended throughout the therapy. Participants are provided with a smartwatch and a customized study app. We collect passive data on heart rate, physical activity, sleep, and location patterns. EMA assessments cover affect, emotion regulation strategies, context and therapeutic agency. Primary outcomes on (non)-response are assessed after 20 therapy sessions and therapy end. We employ predictive and exploratory analyses. Predictive analyses focus on classification of non-response, using basic algorithms (i.e. logistic regression, gradient boosting) for straightforward interpretability, and advanced methods (LSTM, DSEM) to capture complex temporal and hierarchical patterns. Exploratory analyses investigate mechanistic links, examine the interplay of variables over time, and analyze change trajectories. Study findings will inform more personalized and ecologically valid approaches to cognitive behavioral therapy for internalizing disorders. Ethics and DisseminationThe study has received ethical approval from the Institutional Ethics Committee of the Department of Psychology at Humboldt Universitat zu Berlin (Approval No. 2021-01) and the Ethics Committee of Charite-Universitatsmedizin Berlin (Approval No. EA1/186/22). Results will be disseminated through peer-reviewed journals and presentations at national and international conferences. One year after the last patient out, data will be fully anonymized to allow for open science practices and data sharing, while at the same time securing patients data protection rights. Trial registration numberDRKS00030915; OSF PREACT: http://osf.io/bcgax; OSF PREACT-digital: https://osf.io/253nb Strengths and limitations of this studyO_LILarge sample of clinical patients starting CBT: Rare in digital phenotyping studies C_LIO_LILongitudinal assessment combining passive sensing and EMA: Captures a dynamic, multidimensional view of participants behaviors, emotions, and physiological states; Enables analysis of within-person trajectories. C_LIO_LIState-of-the-art, consumer-grade scanwatches for all participants: Ensures high-quality, standardized data collection across participants; Ensures privacy of participant by not disclosing them as study participants C_LIO_LINo healthy control group: Limits the ability to distinguish clinical from non-clinical patterns in digital phenotyping data C_LIO_LIParticipant burden due to extensive assessments might decrease adherence and influence therapy outcomes C_LI