A Network Approach to Feedback: Identifying Intersession Processes Linked to Symptom Outcome in Psychotherapy

Abstract Background There is robust evidence that routine outcome monitoring (ROM) systems exert small but consistently positive effects on psychotherapy outcomes (e.g., de Jong et al. in Clin Psychol Rev 85:102002, 2021; Lambert et al. in Psychotherapy 55 (4):520–537, 2018; Shimokawa in J Consult Clin Psychol 78:298–311, 2010. https://doi.org/10.1037/a0019247 ). One way to improve the effect sizes of ROM systems is the usage of Clinical Support Tools (CSTs) (e.g. Barkham et al., 2023) based on the relations between therapeutic processes and the development of symptoms. This study aims to apply Bayesian Networks which have the ability to identify admissible causal relationships (Maathuis et al. in Handbook of graphical models, CRC, 2018) to uncover associations between intersession processes and symptom severity for the development of a CST. Methods The routine data set consists of N = 262 outpatients receiving cognitive behavioral therapy with weekly assessments of symptom severity, intersession processes, insight, and the therapeutic alliance via the Greifswald Psychotherapy Navigation System (GPNS). Using Bayesian Networks, we identify process-outcome relations and propose a CST. Results The multilevel Bayesian Network showed a negative link between the evaluation of therapeutic strategies that are applied between sessions to solve problems with β = − 0.19 and the symptom outcome of the following session as well as a positive link with β = 0.22 for negative therapy-related emotions and β = 0.018 for the frequency of the recalling therapy during difficult situations. Based on these findings, a CST incorporating a categorical decision tree was developed. Conclusion These results indicate that the application of Bayesian Networks offers a promising approach to enhance data-driven personalization in psychotherapy that should be validated in future research.