Bias in follow-up assessment of psychotherapy: Are certain patients more likely to participate?

Follow-up (FU) assessments are a key component in evaluating the long-term effectiveness of psychotherapy, but participation is often incomplete. Therefore, with this study, we aimed to examine what patient characteristics predict follow-up participation and to estimate the potential bias this may introduce into treatment effectiveness estimates. We analyzed data from 586 patients who received cognitive-behavioral therapy. A random forest feature selection algorithm (Boruta) was applied in order to identify particularly important predictors of follow-up participation, 1-year posttreatment. To approximate the bias introduced by nonrandom follow-up participation, we conducted a post hoc propensity score matching analysis using the identified important predictors. Follow-up participation (38.2%) was associated with higher treatment satisfaction, a stronger therapeutic alliance at the end of treatment, and greater reductions of interpersonal sensitivity over the course of treatment. Older age was also linked to greater follow-up participation, while education, comorbidity, and baseline symptom severity were not. Patients with less improvement in somatization and interpersonal difficulties were more likely to respond. Propensity score matching revealed that treatment effectiveness (as measured by symptom reduction) may be overestimated by approximately 21% when relying solely on follow-up completers, relative to patients resembling nonparticipants. Our findings highlight a potential bias in psychotherapy outcome evaluations due to nonrandom follow-up participation. Strategies that may help improve follow-up response rates in routine care are discussed. (PsycInfo Database Record (c) 2026 APA, all rights reserved).