General and specific factors of working alliance in parent training: A bifactor exploratory structural equation modelling approach

Abstract

The Working Alliance Inventory Short form (WAI-S) comprises 12 items that measure 3 subdomains (goal, task, and bond). In the present study, we evaluated the factor structure of WAI-S in a parent management training (PMT) context, by investigating a series of different factor models, including standard confirmatory factor analyses (CFA) models and more recent alternatives, like the exploratory structural equation model (ESEM), the bifactor-CFA, and the bifactor exploratory structural equation model (B-ESEM). The study sample consisted of 259 Norwegian parents receiving PMT - the Oregon model (PMTO). Alliance was rated by parents of children with emerging or present conduct problems after the first therapy session. Results showed that the B-ESEM model provided best model fit to the data. Estimated sources of variance and omega reliabilities supported a strong general alliance factor, but revealed poor quality of the specific factors. Overall, the present study implies that specific factors of working alliance should be interpreted with caution; rather one should rely on a general working alliance construct. Clinical or methodological significance of this article: Findings suggest that working alliance, as measured by WAI-S in a PMT context, is best reflected by a general construct that also take into account multidimensionality. However, only the general factors provide acceptable reliability. Consequently, practitioners should use the specific factors with caution. The indicators of the specific factors should be improved.

The Working Alliance Inventory's Measurement Properties: A Systematic Review

“…In total, four out of 51 studies reported an adequate model fit, according to the COSMIN criteria for sufficient structural validity (Hatcher et al, 1995;Toste et al, 2015;Hukkelberg and Ogden, 2017;Knowles et al, 2020). The study of Hatcher et al (1995) confirmed a three-factor structure (bond, task, and goal) as the best-fitting model; however, the methodological quality of this study was inadequate because of its small sample size.…”

“…The criteria for sufficient internal consistency are a Cronbach's alpha ≥ 0.70 and at least low evidence for sufficient structural validity (sufficient validity = sufficient model fit, tested with confirmatory factor analyses or IRT/RASH analyses). Only one study met both criteria for internal consistency (Hukkelberg and Ogden, 2017). In 64 different analyses of the WAI and its adapted versions a Cronbach's alpha > 0.70 was found.…”

“…As a result, these findings are rated as inconsistent. Only four studies were in accordance with the COSMIN criteria for sufficient model fit for structural validity (Hatcher et al, 1995, Hukkelberg andOgden 2017;Toste et al, 2015;Knowles et al, 2020). Three of these studies met all measurement criteria and found a bi-factor structure (hierarchical structure) and a two-factor structure to be the best-fitting model (Toste et al, 2015;Hukkelberg and Ogden, 2017;Knowles et al, 2020).…”

“…The other three studies with an adequate fit confirmed a bi-factor structure (hierarchical model) or a two-factor structure (n = 2) as the best-fitting model. The methodological quality of these studies was very good (Toste et al, 2015;Hukkelberg and Ogden, 2017;Knowles et al, 2020). The methodological quality of six studies concerning structural validity was rated indeterminate (in Supplementary Table 3 indicated with a ?…”

“…Only four studies were in accordance with the COSMIN criteria for sufficient model fit for structural validity (Hatcher et al, 1995, Hukkelberg andOgden 2017;Toste et al, 2015;Knowles et al, 2020). Three of these studies met all measurement criteria and found a bi-factor structure (hierarchical structure) and a two-factor structure to be the best-fitting model (Toste et al, 2015;Hukkelberg and Ogden, 2017;Knowles et al, 2020). Within the included studies no <15 different criteria were applied to assess goodness of fit, which illustrates the heterogeneity regarding interpretation of outcomes concerning confirmatory factor analyses.…”