Journal of Consulting and Clinical Psychology, Vol 94(1), Jan 2026, 49-62; doi:10.1037/ccp0000991
Objective: Couple relationship education (CRE) seeks to enhance relationship functioning and prevent deterioration of relationship quality over time. However, impacts of CRE are mixed and often appear to be influenced by the characteristics of the couples receiving the intervention. To provide effective interventions, a better understanding of the couples who are most likely to benefit from CRE is needed. Unfortunately, the existing literature has failed to account for the complex and interdependent nature of pretreatment risk factors, leading to inconsistent and inconclusive results. Method: The present study addresses this issue by applying causal forest, a machine learning technique, to two randomized controlled trials of CRE to determine the pretreatment characteristics that are most predictive of treatment outcomes. In Study 1, data from 6,298 couples were used to train causal forest algorithms, and in Study 2, data from 1,595 couples were used to test the accuracy and generalizability of the trained models. Results: Causal forest models indicated that pretreatment characteristics predicted 12-month treatment effects, such that participants with higher psychological distress and lower baseline relationship happiness experienced greater improvements in relationship happiness, while those with higher psychological distress and perceived stress had greater reductions in negative emotions and behaviors within the relationship. These results were robust when tested in a novel data set. Conclusions: This research highlights the underlying heterogeneity in CRE treatment effects and demonstrates the ability of machine learning methods to identify who may benefit most from CRE and can inform efforts to improve targeting of these interventions. (PsycInfo Database Record (c) 2026 APA, all rights reserved)