Comparison of Therapeutic Alliance across Informants: Therapist Versus Independent-Observer Ratings

Background: The working relationship between a therapist and a client, known as therapeutic alliance, is recognized as an important contributor to treatment change, with a stronger therapeutic alliance predicting better outcomes following therapy (Karver, Handelsman, Fields & Bickman, 2005). However, the association between therapeutic alliance and treatment outcomes has been found to vary depending upon the informant source (McLeod, Southam-Gerow & Kendall, 2017). Research has only recently begun to explore the role of therapeutic alliance in treatment for children with autism. To date, only one study using a multi-informant approach to assess therapeutic alliance exists (Kerns, Collier, Lewin, & Storch, 2017), which included ratings from those directly involved in the therapy process (i.e., child, parent and therapist). Research has yet to examine the association between independent-observer and treatment-involved informants on ratings of therapeutic alliance for children with autism.

Objectives: To examine association between therapist-rated and observer-rated therapeutic alliance in cognitive behaviour therapy (CBT) for school-aged children with autism.

Methods: Data were collected from 48 children with autism (91.7% male) and their therapists (90.9% female) who participated in a 10-session CBT program to improve emotion regulation. Children were 8 to 12 years of age (M = 9.60 years, SD = 1.25) with at least average IQ. Therapist-rated and observer-rated therapeutic alliances were assessed for four sessions (two early and two late). Therapists rated the alliance on a single-item using a 7-point Likert scale (1 = Very poor; 7 = Very good) at the end of each session. The Therapy Process Observational Coding System for Child Psychotherapy-Alliance Scale (TPOCS-A; McLeod & Weisz, 2005) was used to assess alliance through independent observation. Nine items measuring key aspects of the therapeutic alliance were rated on a 6-point Likert scale (0 = Not at all; 5 = A great deal). Observers reached excellent reliability for all items (ICC = .92, p < .001).

Results: Spearman’s correlations showed that therapist-rated alliance across early and late sessions were moderately correlated (rs= .46, p = .001) and observer-rated alliance across early and late sessions was strongly related (rs= .65, p < .001). Spearman’s correlations also indicated moderate, positive relations among therapist ratings and independent-observer ratings of early and late alliance: Session 2 (rs= .44, p = .003), Session 3 (rs= .44, p = .004), Session 8 (rs= .54, p < .001), and Session 9 (rs= .44, p = .003).

Conclusions: There is a moderate degree of consistency between therapist and independent-observer ratings of the quality of early and late alliance in CBT for children with autism. Differences may in part be related to therapist ratings reflecting subjective and motivational aspects of their alliance and observer ratings not taking this into account. In addition, therapist impression of child improvement during therapy could impact ratings. Future research should explore how parent-reported and child-reported ratings of alliance are associated with observational coding, designed for children with autism, and how these factors account for variability in child outcome.