The Brazilian-adapted Working Alliance Inventory: preliminary report on the psychometric properties of the original and short revised versions

Abstract Introduction Alliance is an essential component of all psychotherapies and a consistent predictor of its outcomes. The Working Alliance Inventory (WAI) is a widely used and psychometrically sound measure of alliance. It assesses three key aspects of the construct: a) agreement on the tasks of therapy; b) agreement on the therapeutic goals; and c) development of an affective bond. Objective To preliminarily analyze the psychometric properties of the Brazilian version of both the original, 36-item WAI, and of the short form revised, 16-item version (WAI-SR). Methods The sample comprised 201 psychodynamic psychotherapy patients. Alliance assessments were made after the 4th treatment session. Results The inventory adapted to Brazilian Portuguese, in both the original and short forms, appears to be reliable and valid to measure alliance and its dimensions by clients in psychotherapy. Further studies are needed to replicate and expand the findings.


Introduction
The patient-therapist relationship is a factor common to all psychotherapeutic processes. Relationship factors are at least as much relevant as the particular psychotherapy approach adopted to predict treatment outcome. 1 Therapeutic alliance (working alliance or simply alliance) is by far the most widely researched common factor in the context of psychotherapy research worldwide. 2 A recent meta-analysis of 295 independent studies (published between 1978 and 2017) that covered more than 30,000 patients for face-to-face and Internet-based psychotherapy confirmed the robustness of the positive relation between alliance and outcome.
These data remain consistent across raters (clients, therapists, observers, and others, such as family, group members or staff), alliance and outcome measures, treatment approaches, patient characteristics, and countries. The study findings indicated a trend towards slightly lower observer-rated effects when compared with client-rated effects in terms of alliance-outcome correlation (whereas therapist and other categories did not differ from client-rated alliance). 3 Thus, measuring alliance is fundamental for conducting process-outcome research of any kind, as well as for clinical assessments in routine practice. 1  It is noteworthy that the diversity of instruments designed to measure alliance reflects the variety of co-existing constructs. 4  We found limited studies, with few instruments: the California Psychotherapy Alliance Scale -Patient Version (CALPAS-P), the Helping Alliance Questionnaire (HAq II), and the WAI. 12 The Brazilian version of the CALPAS-P has shown limited to acceptable reliability, with Cronbach's alphas of subscales varying from 0.56 to 0.84 and a total scale score of 0.90. 13 The Brazilian versions of HAq II (patient and therapist) were used to explore the relationship between alliance and defense mechanism. Psychometric properties of these versions are not reported. 14  The WAI has three main forms: therapist (WAI-T), client (WAI-C), and observer (WAI-O). Other variants of the inventory have also been developed, e.g., the 12-item Working Alliance Inventory -Short (WAI-S) 16 and its revised version (WAI-SR). 17 A 6-item version of the instrument was developed for use repeatedly over treatment allowing to measure alliance change on a session-to-session basis. 18 The original inventory has 36 items, 12 for each subscale: emotional bond between patient and therapist; agreement of goals; and tasks.
Several studies confirmed the three factor-structure of the scale, but also noted a high correlation between the goal and task subscales. 10,19,20 Other authors 21,22 prefer a twofactor model that combines the goal and task subscales.
As for the development of the first short version Munder et al. 20 investigated and compared the psychometric properties of the WAI-SR in German outpatients (n = 88) and inpatients (n = 243). Their results were widely in accordance with those of Hatcher & Gillaspy, 17 supporting the three-factor model. In other words, the findings suggested that the WAI-SR is able to distinguish the task and goal aspects of the therapeutic alliance. Conversely, Falkenström et al. 23

Results
In order to examine whether the original Bordin The scales' Cronbach's alphas are presented in Table 2.
As expected, correlations between the subscales and the WAI total alliance score were all strong and significant (≤ 0.001): r = 0.86 (bond), r = 0.94 (task) and r = 0.95 (goal). A similar correlation pattern was found between the WAI-SR bond, task and goal subscales and the WAI-SR total score: r = 0.65, r = 0.87, and r = 0.84, respectively. Correlations between the subscales and the total score of the two versions of the measure were also estimated ( Table 3).
As can be observed in Table 3, all WAI and WAI-SR subscales were highly correlated. Therefore, both the WAI and the WAI-SR were considered suitable to measure alliance and its components in a very similar way.

Discussion
The process of cross-cultural adaptation of a measure intends to produce equivalence between the original and the adapted versions based on item content. A well-done cultural adaptation process, however, does not assure retention of psychometric properties such as validity and reliability. Thus, it is necessary to test the psychometric properties of the adapted measure as well. 24 Bordin 11 conceived that alliance is developed and sustained by an ongoing negotiation between patient and therapist, comprising three facets: agreement on goals, collaboration on tasks, and bond. The WAI was built based on this model. 10