ABSTRACT
Understanding when and who is at greater risk of dropping out can help therapists and mental health professionals develop strategies to keep clients engaged, thereby improving the overall effectiveness of therapy. In the present study, we aimed to explore the hazard of dropout, both early and late in treatment, and evaluate the predictive power of both static (i.e., baseline client characteristics) and dynamic predictors (i.e., treatment process variables), in a sample of 97 clients treated with the Unified Protocol for Transdiagnostic Treatment of Emotional Disorders. Thirty-one clients (31.96%) discontinued treatment, with the highest dropout rate observed at the end of Session 4. Stratified Cox modelling showed a significant effect of younger age β1=−1.706,p=0.031$$ left({beta}_1=-1.706,p=0.031right) $$, non-student status β2=−1.715,p=0.031$$ left({beta}_2=-1.715,p=0.031right) $$ and higher baseline anxiety severity β3=1.806,p=0.022$$ left({beta}_3=1.806,p=0.022right) $$, on the risk of dropping out until Session 4 (i.e., the early phase of treatment). Beyond Session 4, only lower depressive severity measured at intake began to significantly affect the risk of dropout β8=−1.007,p=0.034$$ left({beta}_8=-1.007,p=0.034right) $$. Additionally, joint modelling (JM) of survival and longitudinal data for the early phase of treatment provided evidence of the predictive power of lower distress levels γ01=−0.181,p=0.093$$ left({gamma}_{01}=-0.181,p=0.093right) $$ and lower therapeutic alliance quality γ02=−0.137,p=0.05$$ left({gamma}_{02}=-0.137,p=0.05right) $$ on the clients’ decision to discontinue treatment prematurely. After Session 4, no factor showed a significant association with dropout risk within the JM approach. The results suggest that early sessions are critical for the client’s retention, and the JM approach is indispensable for studying psychotherapy dropout, especially in the context of endogenous time-varying predictors. Implications for clinical and research contexts are discussed.