(PS12-A12) The Therapeutic Alliance and Symptom Change in Behavioral Activation Therapy for Anhedonic Adolescents: Comparing Within vs. Between Patient Effects

Background: The alliance between patient and therapist, often measured using the Working Alliance Inventory (WAI; Horvath & Greenberg, 1986), has been shown to correlate with treatment outcome. Still, the nature of the relationship between the alliance and symptom change is not fully understood, and these two constructs may influence each other reciprocally over time (Flückiger et al., 2020). Moreover, research is needed to clarify whether symptom change can be attributed more to trait-like (between-person) differences in the alliance or state-like (within-person) differences over time (Zilcha-Mano & Webb, 2021). The current study examined the relations between the alliance and symptoms (depression and anhedonia) on both between- and within- patient levels across a 12-week trial of Behavioral Activation (BA) therapy for anhedonic adolescents.

Methods: Adolescents (n = 60; Mage = 15.63) with elevated baseline anhedonia completed 12 sessions of weekly BA therapy. Participants rated the alliance every three sessions and reported anhedonia (Snaith–Hamilton Pleasure Scale [SHAPS; Snaith et al., 1995]) and depression (Center for Epidemiological Studies Depression Scale [CES-D; Radloff, 1977]) symptoms weekly. Multilevel models were used to examine the relations between the alliance at each session and symptoms at the following session. Specifically, two separate models were constructed with lagged SHAPS or CES-D scores as the dependent variable. The predictors were alliance scores at the present session as well as a variable representing the difference between the alliance at the present session and the patient-specific mean alliance across all sessions (i.e., capturing within-patient variability in the alliance over time). The models control for age, sex, SHAPS/CES-D scores at the present session, change in SHAPS/CES-D scores since baseline, and the linear effect of time. Additionally, a model was constructed to assess whether change in symptoms from baseline would predict alliance scores at the following session, controlling for age, sex, and symptoms at the present session.

Results: An intraclass correlation coefficient showed that 77% of the variance in the alliance was due to differences between patients, whereas the other 23% could be attributed to changes in the alliance within patients over time. Between-patient differences in the alliance significantly predicted lower anhedonia (b = -.20, t(23.97) = -3.55, p < 0.01) and depression scores (b = -.35, t(16.62) = -4.35, p < 0.001). Meanwhile, within-person changes in the alliance did not significantly predict symptoms at the next session. Prior change in symptoms did not significantly predict changes in the alliance at the following session.

Conclusions: Overall, stronger alliances were related to improvements in symptoms. By comparing within-patient vs. between-patient components of this relationship, analyses suggest that this effect may be due primarily to trait-like differences in the alliance between patients, as opposed to state-like changes in the alliance within patients over time. Although the alliance tended to strengthen and symptoms tended to decrease, improvements in symptoms since baseline did not appear to directly drive changes in the alliance.