Relation of Efficacy Beliefs and Working Alliance to Psychotherapy Outcomes: A Multilevel Analysis

The purpose of the study is to investigate the contributions of several common factor variables to individual psychotherapy outcomes at both individual and dyad levels. Two dyad-level variables (i.e., the working alliance and collective counseling efficacy) were hypothesized to mediate the relation of client individual-level predictors (i.e., coping efficacy, adult attachment, positive and negative affect) to outcomes (i.e., client-rated progress and session quality). The study involved a naturalistic research design in which no variables were manipulated and psychotherapy occurred without alteration. A novel instrument - the Collective Counseling Efficacy Scale (CCES) - was developed to measure the concept of collective counseling efficacy. Data collected from 73 dyads of clients and therapists were analyzed using multilevel structural equation modeling (SEM) techniques. Results of principal axis factoring procedures revealed a 1-factor structure for the CCES, and internal consistency estimates of the total scores were .96 and .93 for clients and therapists, respectively. Based on data from both clients and therapists, intraclass correlations showed that 59% and 34% of the variances of the alliance and collective counseling efficacy, respectively, occurred at the dyad level as opposed to the individual level. To model them as dyad-level factors, 2nd-order intercepts and slopes were created to represent the midpoint and the discrepancy of clients' and therapists' ratings of the two constructs. The alliance and collective counseling efficacy were further integrated by a set of 3rd-order intercept and slope, denoted relational factors. Multilevel SEM analyses provided evidence for the mediating effect of relational factors on the relation of pre-therapy client predictors to post-therapy client-rated outcomes. Specifically, client coping efficacy indirectly predicted client-perceived progress and session quality through the 3rd-order intercept. Also, client coping efficacy (positively) and avoidance attachment (negatively) were indirectly predictive of session quality through the 3rd-order slope. Post-hoc analyses indicated that clients' ratings of the alliance and collective counseling efficacy were significantly higher than those of therapists, and this pattern was positively associated with session quality. In conclusion, the importance of modeling common factor variables at different levels and the inclusion of collective counseling efficacy in psychotherapy research were initially supported by the findings of the study