Marianne Welmers

Chapter 5 128 more congruent in their perspective on the alliance with parents, and did lack a shared perspective on the alliance with youth. However, as in our present study, the number of included youth with self-reports on the alliance in this previous study was very small. Consequently, further research on the effect of alliances with children and adolescents in the context of conjoint family treatment is warranted. Limitations of this study include the small sample size and the relatively weak interrater reliability of observer rated alliance. The use of a small samplemay have limited the ability to identify associations at the 0.05 level significance. Insufficient statistical power also hindered the ability to apply multivariate analyses to study unique effects of SSP and unbalanced alliances. Several authors have pointed to the statistical disadvantages of using difference scores and have recommended other methods requiring larger samples, such as polynomial regression or the use of latent group models (Laird & De Los Reyes, 2013; Kivlighan, 2007). Another limitation is that changes in youth behavior problems frombaseline to follow upmeasures were rather small (yet significant), with one standard deviation or less improvement. It has been reasoned before that change should be interpreted as clinically significant if it is at least two standard deviations from baseline (e.g. Bachler et al., 2016, following Jacobson et al., 1984). Perhaps treatment effects might have been stronger whenmeasured end-of-treatment, and thus the lack of end-of- treatment measures in our study may have hindered the investigation of larger treatment effects. On the other hand, the use of follow-up measures enabled the investigation of sustainable change in youth problems after ending treatment. An important strength of our study is the use of observations in a clinical representative research context of home-based family treatment. Although observational methods are time-consuming and therefore often result in smaller sample sizes, they also enable the investigation of family members’ actual behaviors and interactions with each other and with the therapist. As Oka and Whiting (2013) point out, observing behaviors and interactions in a representative contextmay help inbridging thegapbetween research and everyday clinical practice of systemic family treatment. Our additional use of therapist- and self-reports of the alliance enabled a comprehensivemulti-informant perspective on the occurrence of unbalanced alliances. Furthermore, our study was the first to examine a disbalance in family members’ alliances regardless of family role, whereas previous studies only investigated alliance differences specifically between or within subsystems (parent – adolescent or father-mother respectively). Future research on unbalanced alliances and the family’s shared sense of purpose may benefit from several approaches. First, studying larger and more diverse samples could improve generalizability of the results. Studying larger samples also enables