Four perspectives on traumatized youths’ therapeutic alliance: Correspondence and outcome predictions

ABSTRACT

Objective

Does the rater-perspective of youths’ therapeutic alliance matter? To answer this, we evaluated the relationships between four perspectives of youths’ alliance, then, we examined whether each perspective and potential discordance between the perspectives predicted outcomes.

Method

Participants were 65 youth (M age = 15.11, SD = 2.14; 76.9% girls) undergoing trauma-focused cognitive behavioral therapy (TF-CBT) and their therapists (n = 24). Youths’ alliance was rated by youth, therapists and parents using the Therapeutic Alliance Scale for Children-revised and by observers using the Therapy Process Observational Coding System-Alliance scale. Posttraumatic stress symptoms (PTSS) were assessed with the Child PTSD Symptom Scale (CPSS) and the Clinician Administered PTSD Scale for Children and Adolescents (CAPS-CA).

Results

The alliance ratings by youth-parent, parent-therapist, and therapist-observer significantly correlated. Only a higher youth-rated alliance significantly predicted fewer PTSS. Furthermore, a higher therapist-rated than youth-rated alliance significantly predicted higher scores on CPSS and CAPS-CA, and a higher parent-rated than youth-rated alliance predicted significantly higher CPSS score.

Conclusion

Therapists should explicitly check in with youth clients about the alliance; because only youths’ evaluation of their alliance predicted the outcome and an overestimation of their alliance by therapists and parents predicted more PTSS.

Trial registration: ClinicalTrials.gov identifier: NCT00635752..

Keywords:

• alliance
• trauma
• outcome research
• child psychotherapy
• trauma-focused cognitive behavioral therapy

Clinical or methodological significance of this article: This study is the first to evaluate four perspectives of traumatized youths’ therapeutic alliance in TF-CBT, in which suggests that the rater-perspectives are not interchangeable. The results highlights the importance of directly assessing youths’ evaluation of their alliance; only a stronger alliance as rated from youths’ perspective predicted greater treatment improvements, furthermore, an overestimations of youths’ alliance by therapists and parents were associated with poorer outcomes. Thus, therapists should monitor and take steps to understand and tune into youths’ subjective experience of the alliance as the treatment progresses, also, it can be useful to consult parents about their evaluation of their child’s alliance.

Numerous studies show that the therapeutic alliance is an important predictor of treatment outcomes in youth and adult therapy (Flückiger et al., 2018; Karver et al., 2018; Murphy & Hutton, 2018). The therapeutic alliance, involving an emotional bond between the client and therapist and their agreement on the tasks and goals of the treatment (Bordin, 1979), has typically been treated as a dyadic construct. This is partly because most therapeutic alliance studies have evaluated adult therapy relationships that usually consist of one patient and one therapist. Therapy relationships differ in child therapy, where the child is commonly accompanied by a parent; thus, the therapeutic relationship becomes more complex. The therapist has to involve one or two parents in the therapeutic tasks and goals and secure good alliances in the triangulated relationship that involves both the child and his or her parents (Green, 2006; Karver et al., 2018; Kirsch et al., 2018). In youth therapy in particular, the question then becomes which perspective would be most helpful to assess the alliance: the child’s, the parent’s, the therapist’s or an observer-rated alliance. To answer this question, we need to know which perspective(s) predicts the treatment outcome, whether different perspectives are in accordance with each other, and whether discordance among perspectives is associated with the outcome.

For traumatized children, studies have found that both a strong alliance and trauma-specific interventions may be necessary to alleviate posttraumatic stress symptoms (PTSS) (Ormhaug et al., 2014; Zorzella et al., 2015). However, establishing and assessing the therapeutic alliance with youth exposed to trauma may be particularly challenging for several reasons. First, many traumatized youth, particularly those who have experienced complex trauma, struggle with interpersonal problems (Cloitre et al., 2013). Second, trauma commonly leads to maladaptive thoughts about self-worth, the world as a safe place and negative expectancy to change (Meiser-Stedman et al., 2009; Meiser-Stedman et al., 2019). Such negative appraisals may affect the establishment of mutual task agreement and goals for the future (e.g., “I will never be able to have normal feelings again,” “I can’t cope when things get tough,” “I am permanently damaged”). Third, the hallmark of trauma and posttraumatic stress symptoms is avoidance of trauma triggers (World Health Organization, 2018). Talking or even thinking about traumatizing experiences can lead to re-experiencing with subsequent avoidance of performing the therapeutic tasks—particularly the exposure work. Fourth, re-experiencing often leads to heightened stress responses. Observers may interpret this as something negative in the working relationship instead of constructive trauma processing. Lastly, many trauma related stress responses such as re-experiencing, dissociation, and avoidance of traumatic thoughts are not easily observable. If the child does not verbally express these feelings and responses, it may be very difficult for an observer to correctly assess how the therapeutic relationship is evolving. Even though some of these issues are not exclusive for trauma (i.e., avoidance for anxious children, negative foresight for depressed), there are some additional challenges when evaluating traumatized youths’ alliance

When it comes to rater perspectives on youth’s alliance on outcomes, youths’ subjective evaluation of their alliance has been found to be a reliable predictor of outcomes across studies and diagnoses. A systematic review and meta-analysis that solely included youths’ alliance perspective found a significant mean relationship of r = .29, suggesting that the youths’ alliance perspective accounts for an unprecedented 8% to 12% of explained variability in treatment outcomes across clinical populations (Murphy & Hutton, 2018). This effect size is larger than the results from earlier meta-studies (r = 0.11–0.22) that included multiple perspectives and therapeutic relationships (McLeod, 2011; Shirk et al., 2011). Additionally, studies that include traumatized youth have identified youths’ alliance perspective to be a significant outcome predictor in trauma treatment (Ormhaug et al., 2014; Zorzella et al., 2015).

Thus far, the predictive value of therapists’ ratings of youth alliance on outcomes is unclear. On the one hand, an early meta-study by Shirk and Karver (2003), found that therapists’ rating of youths’ alliance was a better predictor than youths’ own rating. Since then, therapists’ rating of youths’ alliance is identified to predict treatment improvements across different youth populations (Kazdin et al., 2006; Marker et al., 2013; Shirk et al., 2008). On the other hand, McLeod (2011) did not find a significant difference between the effect sizes of therapists’ and youths’ evaluations of their alliance as outcome predictors. As noted above, trauma related aspects may affect the therapeutic alliance differently than other patient populations, and to date, evidence suggests that therapists’ perspective on traumatized youths’ alliance does not serve well as an outcome predictor (Ormhaug et al., 2015; Zorzella et al., 2015).

Research on parental reports of the alliance within youth treatment has mainly focused on parents’ own relationship with the therapist (McLeod, 2011), and to a lesser degree on parents’ perceptions of the youths’ alliance. Parents possess unique knowledge about their child’s emotional, social and cognitive development, which in combination with parents’ direct observations in therapy can inform therapy outcomes. Thus, the limited research on parents’ evaluation of their child’s alliance is surprising. In one of the few studies to include parental reports of their anxious child’s alliance, the results showed that mother-rated alliance predicted outcomes, but father-rated alliance did not (Marker et al., 2013). Whether parents’ perception of their child’s alliance predicts the outcome in the treatment of traumatized youth has yet to be evaluated.

Because the findings on the predictive value of therapist-, and parent-rated alliance on youths’ outcomes are inconsistent, some have argued that an observer’s rating of the alliance might be a more equitable methodological approach than the report from those subjectively involved in the treatment (Albaum et al., 2020; McLeod & Weisz, 2005; Shirk & Karver, 2003). However, few studies have examined this issue, and the available findings are mixed. Studies have found that a stronger observer-rated alliance predicted treatment improvements for depressed youth (Labouliere et al., 2017) and youth with autism (Albaum et al., 2020; Kang et al., 2021). Furthermore, McLeod and Weisz (2005) found that a higher observer-rated child alliance predicted greater reductions in anxiety symptomatology but not depressive symptomatology or other internalizing symptoms post-treatment. To evaluate the utility of observer-rated alliance, more studies of its relationship with outcomes are required, and a knowledge gap exists in the evaluation of the predictive value of observers’ alliance perspective on the outcome for traumatized youth.

Since youth therapy most often involves at least three participants (the youth, parent and therapist) one important question is whether these have a similar perception of the alliance (Zandberg et al., 2015). It may be helpful for therapists to know if they are on track and can lean on their own clinical perception of how their youth patients experience the alliance, or if they need to use other sources to evaluate its quality. The results from studies that investigate the level of concordance between youth and therapist alliance ratings are mixed; in some studies, youths’ and therapists’ evaluations of the alliance were found to be significantly associated (Bickman et al., 2012; Kazdin et al., 2006; McLeod et al., 2017), while other studies did not find this association (Fjermestad et al., 2016; van Benthem et al., 2020). One consistent finding is that therapists tend to underestimate the strength of the alliance compared to youths’ alliance reports (Hawley & Garland, 2008; Loos et al., 2020; Ormhaug et al., 2015; Zandberg et al., 2015). In regards to studies of traumatized youth, we are currently aware of only two studies that have investigated concordance between alliance ratings by youth and therapists. In the first study, Zorzella et al. (2015) found significant correlations between therapists’ and children’s (ages 7–12) ratings of the alliance at three-time points of trauma-focused cognitive behavioral therapy (TF-CBT). In the second study that included an older sample (ages 10–18), we found that therapists’ and youths’ alliance ratings across TF-CBT and therapy as usual (TAU) at session six were significantly correlated, r = .39 (Ormhaug et al., 2015).

Regarding the evaluation of parents’ ratings of their child’s alliance, the literature is scarce. In one study, the concordance between parents’ evaluations of their own alliance with the therapist significantly correlated with their traumatized child’s perception of the alliance with the same therapist (Kirsch et al., 2018). The concordance between traumatized youths’ alliance rated by self-report and parents, in addition to the concordance between parents’ and therapists’ perspectives of traumatized youths’ alliance, have yet to be evaluated.

Few studies have simultaneously included an evaluation of youth alliance from self- and observer-reports (McLeod, 2011). In one study, a significant correlation was found between the alliance rated by observers and anxious youth receiving CBT (Creed & Kendall, 2005). In line with this, Karver et al. (2008) found alliance reports from depressed youth receiving CBT to be significantly correlated with observers’ ratings of the alliance. In contrast, a recent study did not find observers’ alliance reports to be significantly correlated with alliance reports from youth with autism spectrum disorder (Kang et al., 2021). In the aforementioned study by McLeod et al. (2017), youths’ alliance scores were only weakly correlated with observers’ alliance scores. The concordance between observers’ and youths’ ratings of traumatized youths’ alliance remains to be examined.

The last question we examine is whether the level of discordance between different rater perspectives is related to outcomes. Several studies of the alliance within the dyadic relationship in adult treatments suggest that a mutual agreement of the strength of the alliance is associated with better treatment outcomes (e.g., Jennissen et al., 2020; Marmarosh & Kivlighan, 2012; Zilcha-Mano et al., 2017). Among the few studies to examine discrepancies in alliance ratings as outcome predictors within the child field, Zandberg et al. (2015) found that a larger discordance in client and therapist alliance ratings was not associated with treatment outcomes for anxious youth. Fjermestad et al. (2016) found that a greater concordance between therapists and anxious youth on changes of the alliance during CBT was related to greater treatment improvements. In our previous research, we investigated the concordance in alliance ratings between traumatized youth and their therapist—results showed that an overestimation of youths’ alliance by their therapists predicted poorer outcomes (Ormhaug et al., 2015). One could assume that a large discrepancy between the child’s perspective and the adults’ perspectives indicates that the child is not being accurately understood, and this could potentially affect outcome.

In sum, little is known about the best way to monitor the alliance in youth therapy (Bickman et al., 2012), and very few studies have examined ways to monitor traumatized youths’ alliance in therapy. On the one hand, it may be helpful to evaluate multiple perspectives of the alliance, as this could help avoid demand characteristics and help to control for common rater variance that may occur because the rater of the alliance and outcome is the same (Hawley & Garland, 2008; Kazdin & Durbin, 2012; McLeod & Weisz, 2005). On the other hand, it may not be practical in a clinician’s everyday practice to administer several assessments, and observer-rated alliance ratings are especially resource demanding. More knowledge on whose perspective to monitor may improve evidence-based practice and help clinicians to be more efficient in their clinical work.

Research Aims and Hypotheses

We aimed to extend existing studies by examining four perspectives on traumatized youths’ alliance (self-report, therapist, parent and observer) during a trauma specific treatment (TF-CBT), building on secondary analysis from a randomized controlled trial showing better treatment improvements for youth receiving TF-CBT compared to those receiving therapy as usual (Jensen et al., 2014). We examined three research questions and formed hypotheses based on the reviewed literature. (1) Which perspectives predict outcomes? We hypothesized that a strong youth-rated alliance would predict larger decline in PTS symptoms, while therapists’ ratings of the alliance would not predict decline in PTS symptoms. Given the mixed findings on observers’ ratings, the limited studies on parents’ ratings, and no studies within a trauma population, we had no predefined hypothesis regarding their predictive value. (2) To what extent is there a concordance between youth, parent, therapist and observer ratings of youths’ alliance? We hypothesized that the four perspectives would be moderately positively correlated. (3) Does the level of discordance predict poorer outcomes? We hypothesized that an overestimation of youths’ alliance by parents and therapist would predict poorer treatment outcomes. We had no hypothesis as to whether discordance between parents’ and therapists’ alliance reports would predict the outcome.

Methods

Participants

Clients

Participants were 65 youth (M age = 15.11, SD = 2.14, range 10–18 years, 76.9% girls) who received TF-CBT as part of a clinical trial in Norway. Information about the study was given both verbally and in written form, and written consent was obtained from both caretakers and youth. Procedures were reviewed and approved by the Regional Committee for Medical and Health Research Ethics. For a full description of the source study, see (Jensen et al., 2014). Inclusion criteria for this trial were referral to one of the eight participating community clinics, age 10–18 years, exposure to at least one traumatic event and significant symptoms of PTS (i.e., a score of 15 or higher on the Child PTSD Symptom Scale (CPSS); Foa et al., 2001). The majority of the sample (76.92%) fulfilled the diagnostic criteria for PTSD as assessed with the Clinician Administered PTSD Scale for Children and Adolescents (CAPS-CA; Nader et al., 2004). In addition, 69.2% scored above the clinical cut-off for depression (Mood and Feelings Questionnaire [MFQ]; Angold et al., 1995), 52.4% over clinical cut-off for anxiety (Screen for Child Anxiety Related Disorders [SCARED]; Birmaher et al., 1999), and 47.7% on other general mental health problems (Strengths and Difficulties Questionnaire [SDQ]; Goodman, 2001). Pretreatment traumatic experiences were assessed using an adapted version of the Traumatic Events Screening Inventory for Children (TESI-C; Ribbe, 1996). Participants reported on average exposure to 3.63 different types of trauma (SD = 1.62, range 1–8). When asked to identify the worst traumatic experience, the majority reported exposure to family violence (38.4%), followed by violence outside the family (16.9%), sudden death of a person close to the participant or involved in a severe accident (18.5%), inter-familial sexual abuse (12.3%), or sexual abuse outside the family (13.8%). Participants’ background was classified into those with at least one Norwegian-born parent (n = 53) or those with non-Norwegian-born parents (n = 12). Most participants lived primarily with one parent (55.4%), 32.3% lived with or spent equal time with both parents, and 10.8% had other living arrangements (1.5% did not report their living situation). For the majority (41.5%) the mean level income was > 83,300 USD, and 37% had a mean level income ≤ 83,300 USD (21.6% did not report income; mean level income in Norway in 2012 = USD 79,800; https://www.ssb.no/).

Caregivers

The majority of caregivers were a biological parent (89.2%); 4.6% were foster parents. Most parents came from the study country (n = 50, 76.9%). Approximately one-half (50.8%) had completed high school/vocational school; 36.8% had attended college/university and 6.2% had completed junior high school. The majority was working full/part time (69.2%), 18.5% were welfare recipients, 6.2% were job seekers/students. Demographic data for caregivers were missing in four cases.

Therapists

The therapists (n = 24, 91.7% female) consisted of 19 psychologists, 2 psychiatrists, 2 clinical educational therapists and 1 clinical social worker. Years of clinical experience ranged from 3 to 28 (M = 9.63, SD = 5.72), and on average, therapists treated 2.71 youth (SD = 1.40, range 1–6). The theoretical orientation of the therapists were CBT (n = 16), psychodynamic (n = 5), or systemic/family therapy (n = 2; one therapist did not report a theoretical orientation). All therapists volunteered to receive TF-CBT training and participate in the study (for further details, see Jensen et al., 2014).

Treatment

TF-CBT is a component-based manualized treatment specifically developed to target PTSS (for a full description of the model, see Cohen et al., 2017). The model is based on theoretical principles from cognitive, behavioral, interpersonal and family therapy. The components are organized into three treatment phases: stabilization and skills building (psychoeducation, relaxation and stress-management skills, affective modulation skills, and cognitive coping), exposure and cognitive processing (creating a trauma narrative, alteration of posttraumatic cognitions), and finally consolidation and closure (in vivo mastery of trauma reminders, enhancing safety and future development). Gradual exposure to the youth’s traumatic experience(s) is a central part of the model and therapists are encouraged to continuously focus on building and maintaining a strong alliance by validating the youth’s experiences, being supportive and trustworthy, and using gradual exposure techniques before the trauma narration to not overwhelm the youth. Youth and their non-offending parents are involved in the treatment, and both parallel and conjoint sessions are provided. The treatment is typically delivered over 12–15 weekly sessions. In this study, youth had, on average, 16.26 (SD = 8.76) sessions before the case was discharged from the clinic. Parents attended, on average, 7.94 sessions (SD = 4.93), and a majority attended ≥ 3 sessions (81.5%). All therapy sessions were audio-recorded and coded for fidelity by at least one trained TF-CBT therapist using the treatment adherence checklist for TF-CBT (Deblinger et al., 2008). All included cases were in accordance with the treatment manual.

Instruments

Youth-, therapist- and parent-rated youth alliance

The Therapeutic Alliance Scale for Children-revised (TASC-r; Shirk, 2003, november; Shirk & Saiz, 1992) was used to measure the perceived quality of youths’ alliance as rated by youth, therapists and parents. The TASC-r measure consists of 12 items, of which half assess the therapeutic bond (e.g., I like spending time with my therapist) and the other half of the items assess task collaboration (e.g., I work with my therapist on solving my problems). Each item is rated on a 4-point scale (0 = not at all to 4 = very much). The scale was translated and back translated according to recommended procedures, and the scales’ first author approved the final Norwegian version. Reliability analyses with the current sample showed that the scale had good internal consistency on scores from youth (n = 58, α = .92), therapists (n = 56, α = .92) and parents (n = 40, α = .93).

Observer-rated youth alliance

The Therapy Process Observational Coding System for Child Psychotherapy—Alliance Scale (TPOCS-A; McLeod, 2001) was used to obtain an observer-rated measure of youths’ alliance. The TPOCS-A is rated by independent evaluators on nine items on a 6-point scale (0 = not at all to 5 = a great deal), in which six items assess bond elements of the client–therapist relationship (e.g., to what extent does the client demonstrate positive affect toward the therapist), and three task items assess client participation in therapeutic activities (to what extent does the client not comply with tasks). Two studies have demonstrated that TPOCS-A has adequate interrater reliability, good internal consistency, and mixed convergent validity with other alliance measures (see McLeod et al., 2021; McLeod & Weisz, 2005). In the present study, the internal consistency between items was excellent (α = .95).

Two graduate students in psychology coded the sessions from audiotapes using the TPOCS-A coding manual. One of the scale’s authors trained the coders, and the coders practiced coding on 22 TF-CBT patients from another study, in which they reached a coder agreement of α = .89, which is excellent (Cicchetti, 1994). Then, the coders used the TPOCS-A to code the alliance for youth in the present study. The patients were randomly selected for the coders, and the coders were blinded to the treatment outcome. A random selection of nine patients (18%) was double coded, and the mean score was used from the patients that were double coded to prevent coders from drifting; also, the coders had weekly meetings. Interrater agreement on the included cases was α = .90, which is excellent (Cicchetti, 1994).

Youth-rated PTSS

PTSS were assessed using the self-completion questionnaire CPSS (Foa et al., 2001), which covers 17 symptoms of PTS disorder (PTSD) defined in the Diagnostic and Statistical Manual of Mental Disorders-IV (American Psychiatric Association, 1994). This includes the three factors re-experiencing, avoidance, and hyperarousal. Symptom frequency in the last two weeks is rated on a 4-point scale (0 = never to 3 = almost every day), yielding a total score ranging from 0 to 51. The scale was translated and back translated, and the developers of the scale approved the Norwegian version. The measure is appropriate for children aged 8–18 years and has demonstrated excellent internal consistency, test-retest reliability, and convergent validity (Foa et al., 2001; Gillihan et al., 2013). Within the larger RCT study, in which the present study is a subsample, the scale showed good internal consistency (total scale: α = .91, re-experience: α = .84, avoidance: α = .80, hyperarousal: α = .75; for further details, see Jensen et al., 2014).

Clinician-rated PTSS

PTSD diagnosis was assessed with the CAPS-CA (Nader et al., 2004). The CAPS-CA is a structured interview that assesses the frequency and intensity of the 17 DSM–IV defined symptoms of PTSD. Items are scored on 5-point frequency scales (e.g., from 0 = none of the time to 4 = most of the time) and 5-point intensity rating scales (e.g., from 0 = not a problem to 4 = a big problem, I have to stop what I am doing), assessing the past month. Items are scored based on both the youths’ answers and clinical judgment during the interview. The interview was translated and back translated, and the first author of the CAPS-CA approved the translation. CAPS-CA has shown good internal consistency (Cronbach’s alpha = .75-.82), excellent interrater reliability (ICC = .97), and adequate convergent validity (Leigh et al., 2016). In the larger RCT study in which the present study is a subsample, the scale showed satisfactory internal consistency (total scale: α = .90, re-experience: α = .87, avoidance: α = .77, hyperarousal: α = .79; for further details, see Jensen et al., 2014).

Procedure

All measurements were administered by licensed psychologists that were blinded to the treatment condition. The CAPS-CA was administered at pre-treatment and post-treatment (after completion or the 15th session). The self-completion instrument CPSS was administered pre-, and post-treatment, and TASC-r was administered at mid-treatment (around session six, M session number = 6.47, SD = 1.25, range 3–9). In order to provide consistency, TPOCS-A was coded during the first narrative session (M session number = 7.22, SD = 1.34, range 5–12).

Data Analysis Plan

We adopted an eight-step approach to data analysis. First, we estimated and inspected skewness, kurtosis, means and standard deviations for all scores on the CPSS, CAPS-CA, TASC-r, and TPOCS-A. Second, we examined the relationships between pre-treatment symptom scores (from CAPS-CA and CPSS) and the alliance scores (from TASC-r and TPOCS-A) using bivariate correlations. Third, we used bivariate correlations to examine the relationships between the variables in step one and the continuous variables; youths’ age and therapists’ level of experience. We then used independent sample t-tests to examine potential differences in gender and background (at least one Norwegian-born parent versus non-Norwegian-born parents) on the variables in step one. Fourth, missing data on the alliance (youth with vs. youth without an alliance score on each of the four alliance measures) were examined with nonparametric tests because of the unequal sample sizes. Potential group differences in continuous variables (age and pretreatment symptoms of PTS) were assessed using Mann–Whitney U tests. Potential differences in categorical variables (gender and background) were assessed using chi-squared tests. Fifth, each alliance perspective was separately entered as an independent variable (IV) in two linear regression models: (1) posttreatment CPSS score as the dependent variable (DV) and pretreatment CPSS score as the IV and (2) posttreatment CAPS-CA score as the DV and pretreatment CAPS-CA as the IV. Sixth, bivariate correlations were used to examine associations between TASC-r and TPOCS-A scores. Seventh, to examine discordance between the rater perspectives on youths’ alliance, we computed three scores: the (1) therapist- minus youth-rated alliance score, (2) parent- minus youth-rated alliance score, and (3) therapist- minus parent-rated alliance score. Then, each score was separately entered as an IV in the two linear regression models as described in step five. Finally, in order to adjust for potential therapist effects we recomputed all models from step five and seven by entering a single multi-category categorical level for therapists as an additional IV, and used the Akaike information criterion (AIC) to compare model fit between models.

Given the nested data structure, we first attempted to estimate mixed-effects models with random effects for therapists with R version 3.6.1 (The R Foundation for Statistical Computation, Vienna, Austria) and the R package nlme (Pinheiro & Bates, 2000). Results showed that the models came out unstable, probably due to the small number of youth treated for some therapists, thus, we followed the advice by Pinheiro and Bates (2000) and performed single-level analyses. The level of statistical significance was set at p < .05. Analyses were conducted using IBM SPSS, version 22 (IBM, 2013). AIC values were computed using R (The R Foundation for Statistical Computation, Vienna, Austria)

Results

Preliminary Analyses

The means and standard deviations for the scores on CPSS (pre- and posttreatment), CAPS-CA (pre- and posttreatment), TASC-r (rated by youth, therapists, and parents) and TPOCS-A showed substantial variability (See Table I), and skew and kurtosis were within acceptable ranges (1.36 — 1.32; −0.85–1.82, respectively), indicating a non-problematic deviation from the assumption of the normality. Correlations between pretreatment symptom scores (CPSS and CAPS-CA) and the alliance measures (TASC-r and TPOCS-A) were all nonsignificant (see Table I). The variables in step one were not associated with youths’ age or therapists’ level of experience (Table I), neither did any differences appeared on these variables across youth’ gender and background. Missing data analyses showed no significant differences between groups of participants with or without TASC-r scores from youth or therapists. Participants with a TASC-r parent score were significantly younger (M = 14.20) than those without this score (M = 16.52, U = 165, z = −4.60, p < .001, r = −0.57). Participants with a TPOCS-A score were significantly younger (M = 14.82) than those without this score (M = 16.07, U = 229, z = −2.31, p = 0.21, r = −2.29). Participants with a TPOCS-A score were more likely to have at least one Norwegian-born parent than those without a TPOCS-A score, X² (1, N = 65) = 6.00, p = 0.14.

Table I. Means, standard deviations and bivariate correlations for youths’ age, therapists’ years of experience, PTSS and alliance scores.

Variables	n	M	SD	1	2	3	4	5	6	7	8	9	10
1. Youths’ age	65	15.09	2.19	-	.19	.11	.12	-.02	.12	-.07	.19	.09	.09
2. Therapists’ yrs. experience	65	9.42	5.33		-	-.07	.14	.02	.08	-.13	.24^†	-.17	.06
3. Pretreatment CPSS	65	27.40	7.41			-	.26^†	.60***	.36*	.04	.15	.24	.03
4. Posttreatment CPSS	53	11.72	10.73				-	.39**	.93***	-.37**	.24^†	.19	-.06
5. Pretreatment CAPS	65	58.95	19.08					-	.47**	.01	.11	.20	-.04
6. Posttreatment CAPS	50	31.66	25.79						-	-.28^†	.18	.16	-.05
7. Youth TASC-r	58	37.83	7.83							-	.15	.46**	.23
8. Therapist TASC-r	56	36.13	5.63								-	.55***	.39**
9. Parent TASC-r	40	39.42	7.78									-	.16
10. Observer TPOCS-A	50	29.11	6.80										-

Note: n = number of participants; M = mean; SD = standard deviation; Therapists’ yrs. Experience = Therapists years of clinical experience. CPSS = Child PTS Symptom Scale; CAPS-CA = Clinical Diagnostic Interview of Adolescents’ PTSS; TASC-r = Therapeutic Alliance Scale for Children-revised; TPOCS-A = Therapy Process Observational Coding System-Alliance scale. ^†p < 0.10, *p < 0.05, **p < 0.010, ***p < 0.001.

Primary Analyses

First, we examined the four perspective(s) of youths ‘alliance as PTS outcome predictors in TF-CBT (see Table II). As hypothesized, higher alliance scores by youth significantly predicted lower PTS scores on CPSS (Est. = −0.52, p = .004, 95% CI [−0.86, −0.17]) and CAPS-CA (Est. = −0.84, p = .046, 95% CI [−1.66, −0.02]), while therapists’ alliance scores were not significantly associated with PTS outcomes. Furthermore, our exploratory analyses showed that neither parents’ nor observers’ alliance scores significantly predicted PTS outcomes.

Table II. Posttreatment PTSS outcomes predicted by Pretreatment PTSS outcomes and alliance measures.

Predicting Posttreatment CPSS scores				Predicting Posttreatment CAPS scores
Variables	Est.	95% CI	p	Variables	Est.	95% CI	p
Pretreatment CPSS scores	0.38	[0.02, 0.75]	.039	Pretreatment CAPS scores	0.58	[0.26, 0.90]	.001
Youth TASC-r	−0.52	[−0.86, −0.17]	.004	Youth TASC-r	−0.84	[−1.66, −0.02]	.046
Pretreatment CPSS scores	0.34	[−0.08, 0.75]	.110	Pretreatment CAPS scores	0.65	[0.29, 1.01]	.001
Therapist TASC-r	0.44	[−0.14, 1.03]	.134	Therapist TASC-r	0.66	[−0.66, 1.98]	.317
Pretreatment CPSS scores	0.60	[0.12, 1.09]	.016	Pretreatment CAPS scores	0.67	[0.27,1.06]	.002
Parent TASC-r	0.13	[−0.32, 0.57]	.568	Parent TASC-r	0.24	[−0.83, 1.30]	.655
Pretreatment CPSS scores	0.40	[−0.01, 0.81]	.055	Pretreatment CAPS scores	0.50	[0.15, 0.86]	.007
Observer TPOCS-A	−0.12	[−0.61, 0.38]	.629	Observer TPOCS-A	−0.10	[−1.24, 1.04]	.860

Note. Est. = estimate; CI = confidence interval. CPSS = Child PTSD Symptom Scale; CAPS-CA = Clinician-Administered PTSD Scale for Children and Adolescents; TASC-r = Therapeutic Alliance Scale for Children-revised; TPOCS-A = Therapy Process Observational Coding System-Alliance scale.

Secondly, we examined the relationships between the four perspectives of youths’ alliance (see Table I). As predicted, youths’ and parents’ alliance scores were significantly positively correlated (n = 40, r = .46, p = .003), with medium strength. Furthermore, significantly positive correlations were found between the parent-therapist rated alliance scores (n = 39, r = .55, p = < .001) and observer-therapist-rated alliance scores (n = 48, r = .39, p = .006), both with medium strengths. Contrary to our prediction, youths’ alliance scores were not significantly correlated with the alliance scores by therapists or observers (n = 54, r = .15, p = .276 and n = 50, r = .23, p = .107, respectively), neither was parent-rated alliance scores significantly correlated with observers’ ratings (n = 37, r = .16, p = .349). To further dismantle the discrepancies from youths’ alliance ratings by the other respondents we did post-hoc analyses to examine the correlations between youths’ bond scores with the other informants’ bond scores; and youths’ task scores with the other informants’ task scores. Results showed that youths’ bond scores significantly correlated with parents’ (r = .53, p < .001) but not therapists’ or observers’ task scores (r = .25, p = .069 and r = .26, p = .069, respectively). Youths’ task scores significantly correlated with parents’ (r = .35, p = .027) but not with therapists’ or observers’ task scores (r = .02, p = .882 and r = .19, p = .175, respectively).

Third, we examined discrepancies between youths’, therapists’ and parents’ alliance scores as PTS outcome predictors controlling for pretreatment PTS (see Table III). As predicted, a higher therapist- than youth-rated alliance score was a significant predictor of higher posttreatment PTSS scores on CPSS (Est. = 0.57, p < .001, CI [0.27, 0.87]) and CAPS-CA (Est. = 0.90, p = .017, CI [0.17, 1.63]). Furthermore, our hypothesis that a higher parent- than youth-rated alliance score would predict poorer PTSS outcomes was partially supported; a larger discordance in terms of parental overrating of the alliance compared with youths’ own reports significantly predicted higher scores on CPSS (Est. = 0.54, p = .008, 95% CI [0.15, 0.93]), but it did not predict scores on CAPS-CA (Est. = 0.81, p = .116, CI [−0.21, 1.83]). Our last exploratory analysis showed that the discrepancy between therapist-parent alliance scores did not predict outcomes on CPSS (Est. = 0.52, p = .004, 95% CI [−0.86, −0.17]) or CAPS-CA (Est. = 0.52, p = .004, 95% CI [−0.86, −0.17]).

Table III. Posttreatment PTSS outcomes predicted by Pretreatment PTSS outcomes and discrepancies in alliance scores by youth, therapists and parents.

Predicting Posttreatment CPSS scores				Predicting Posttreatment CAPS scores
Variable	Est.	95% CI	p	Variable	Est.	95% CI	p
Pretreatment CPSS score	0.35	[−0.02–0.72]	.063	Pretreatment CAPS score	0.62	[0.28–0.96]	.001
Therapist minus Youth TASC-r	0.57	[0.27–0.87]	<.001	Therapist minus Youth TASC-r	0.90	[0.17–1.63]	.017
Pretreatment CPSS score	0.51	[0.06–0.94]	.027	Pretreatment CAPS scores	0.59	[0.20–0.99]	.004
Parent minus Youth TASC-r	0.54	[0.15–0.93]	.009	Parent minus Youth TASC-r	0.81	[−0.21–1.83]	.116
Pretreatment CPSS score	0.63	[0.14–1.11]	.013	Pretreatment CAPS score	0.74	[0.34–1.14]	.001
Therapist minus Parent TASC-r	−0.07	[−0.59–0.46]	.799	Therapist minus Parent TASC-r	−0.08	[−1.31–1.14]	.890

Note. Est. = estimate. CI = confidence interval; CPSS = Child PTS Symptom Scale; CAPS = Clinician-Administered PTSD Scale for Children and Adolescents; TASC-r = Therapeutic Alliance Scale for Children-revised.

Finally, results showed that the recomputed models from step five with the entrance of a single multi-category categorical level for therapists as an additional IV provided poorer fit according to AIC for all models except for the model predicting CPSS T3 from the discrepancy in parent-youth alliance score in which provided a better model fit. The results from the recomputed models showed that higher alliance scores from youth significantly predicted lower outcome scores on CPSS (Est. = −0.75 CI [−1.26, −0.24], p = .005) and CAPS (Est. = - 1.60, CI [−2.68, −0.51], p = .006), while neither parents’, therapists’ nor observers’ alliance scores significantly predicted outcomes. Furthermore, the discrepancy between therapist- youth alliance scores, and the discrepancy between parent-youth alliance scores significantly predicted higher outcomes on CPSS (Est. = 0.75, CI [0.24, 1.26], p = .005 and Est. = 0.80, CI [0.08, 1.53], p = .032, respectively). The discrepancy between therapist-youth alliance scores significantly predicted outcome scores on CAPS-CA (Est. = 1.54, CI [0.40, 2.68], p = .010), while the discrepancy between parent-youth alliance scores was a trending predictor of outcome scores on CAPS-CA (Est. = 1.60, CI [−0.10–3.30], p = .063). The discrepancy between therapist-parent alliance scores did not predict outcomes. Of note, the single multi-category categorical level for therapists was not a significant outcome predictor in any of the models. The recomputed models from step five and seven that included the interactional terms between the primary IVs in each model with the single multi-category categorical level for therapists on outcome indicated a better model fit according to AIC for all models. The only significant interactional effect was between the discrepancy between therapist-parent alliance scores and the single multi-category categorical level for therapists variable on CPSS outcomes (p = .026). Overall, the interactional models provided higher p-values and increased the standard error for all models as was expected given the increased number of degrees of freedom in the models when entering the single multi-category categorical level for therapists in the primary models from step five and seven.

Discussion

It is widely recognized that a strong therapeutic alliance is important for optimizing treatment outcomes. However, the attention given to rater perspective has been scant and it is uncertain which rater perspective is best to monitor the alliance in youth treatment (Bickman et al., 2012). In response to this gap, we first investigated the predictive role of four perspectives of youths’ alliance on PTS outcomes from TF-CBT. In line with our hypothesis, a stronger youth perceived alliance predicted less PTSS, while neither therapists’, observers’ nor parents’ perspectives of youths’ alliance predicted outcomes. In terms of concordance, results showed that youth and parent ratings were significantly related, as well as parent—therapist, and therapist—observers. Lastly, we investigated the predictive role of rater discordance on youths’ alliance on outcome. As hypothesized, an overestimation of youths’ alliance by therapists and parents was associated with poorer treatment response, however, discordance between therapists’ and parents’ perception of youths’ alliance was not related to treatment outcomes.

Our finding that a stronger youth-rated alliance predicted better treatment response is in accordance with results from previous studies involving traumatized youth (Ormhaug et al., 2014; Zorzella et al., 2015). For traumatized youth to experience their therapist as trustworthy, caring and knowledgeable can help reduce an initial fear of talking about their traumatic experiences and help to establish a strong working relationship. Thus, perhaps the feeling of being in an emotionally containing and collaborative working relationship with a therapist during the mid-treatment phase is associated with a more beneficial trauma processing. For youth to experience that they are on the same page as their therapist on the treatment’s task, along with having a mutually strong emotional bond, may be curative on its own but may also be a prerequisite for treatment involvement in therapeutic activities (Kendall & Ollendick, 2004; Shirk & Karver, 2003). Future studies are encouraged to examine the associations between youths’ alliance and in-session involvement behaviors in trauma treatment.

According to the current study’s results, professionals’ evaluations of traumatized youths’ alliance, either based on direct or indirect observations of the youth, do neither predict traumatized youths’ treatment response nor significantly correspond with youths’ own view of their alliance. There may be several reasons why it may be challenging for therapists and observers to capture both facets of traumatized youths’ alliance. First, assessing traumatized youths’ alliance may be more complicated compared to other clinical populations due to the typical PTS symptoms of re-experiencing, avoidance and hypervigilance, and the role trauma reminders play in activating these responses. Results from a qualitatively study by Dittmann and Jensen (2014) indicated that many youth in TF-CBT experience that working with their trauma history is among the most challenging but also among the most helpful part of TF-CBT. However, this in-session activation/stress and avoidance during the trauma exposure work could be interpreted as poor task collaboration by the observers and therapists. Second, many symptoms of PTS are internalizing and commonly include negative appraisals of self and the world (Meiser-Stedman et al., 2009; Meiser-Stedman et al., 2019), that may affect traumatized youths’ working models and schemas of interplays in close relationships. Thus, it might be necessary to directly ask youth to describe their emotional bond with a therapist as well as frequently assess symptom development given that these are inner experiences that are not easily observed (Smith et al., 2019). Third, therapists’ and observers’ evaluations of the alliance were significantly associated, indicating that clinical training may be associated with common conceptualizations of the alliance. It is possible that professionals that are trained to evaluate youths’ alliance do not consider that the conceptualization of a therapeutic alliance may be differently perceived by profesionals and youth. Research suggests that therapists’ alliance scores center around the bond and tasks dimensions while it seems that youth rather perceive the alliance as a unidimensional affective construct (Accurso et al., 2013; Ormhaug et al., 2015). Thus, it is necessary that therapists also directly adress youths’ perception of the mutual task cooperation, and weighten the focusing on this alliance facet up agains the importance of a strong therapeutic bond.

According to the current study’s results parent-rated alliance does not predict youths’ treatment outcomes from TF-CBT, even though there was a high concordance between parents’ and youths’ alliance ratings. There may be several reasons that parents seem to be particularly good at tuning in to their child’s alliance. For one, parents are often well informed about their child’s referral process and the child’s attitude and expectations regarding therapy. Secondly, parents possess unique insights into their child’s lifespan, developmental stage and relational response patterns. Lastly, parents observe their child across the therapeutic context and in private situations. This enables parents to evaluate their child’s alliance based on the combination of direct observations of the child’s interaction with the therapist with relevant information of the therapeutic relationship provided from the child between sessions (e.g., verbal expressions about the therapist or about attending therapy). Despite this, we did not find parents’ evaluation of their child’s alliance to predict youths’ treatment response. Again, this underscores that an evaluation of traumatized youths’ alliance through an adult’s point of view may not be sufficient to capture the working ingredient(s) of the alliance, thus, youths’ own voices must be heard when evaluating their alliance. Evidence suggests however that parents’ perspective of their child’s alliance is linked to other aspects of treatment, such as optimizing treatment retention and fostering engagement in treatment-related strategies (Marker et al., 2013). According to McLeod’s (2011) study, parents’ own alliance is a stronger outcome predictor in youth treatment than youths’ alliance rated by self-report and observers. Also, parents’ own alliance seems to be a good predictor of the outcome for youth receiving TF-CBT (Kirsch et al., 2018). Thus, we encourage future studies to simultaneously assess parents’ and traumatized youths’ alliances. In sum, it may be fruitful for therapists to attend to parents’ views of their traumatized child’s alliance in clinical work. Parents’ evaluation of their child’s alliance also deserves more attention in future youth alliance studies in general.

In line with our last hypothesis, we found that when therapists and parents overestimate the alliance compared with youths’ own alliance ratings that this is associated with poorer treatment response. This implies that when adults are not mindful of traumatized youths’ experience of the therapy process, undesired treatment outcomes can occur. For youth to perceive that the adults involved in the tripartite therapeutic relationship do not have a common understanding of whether they like their therapist or agree on the treatment tasks could result in a feeling of not being understood or that they lack control of their own treatment process. Perhaps traumatized youth are particularly vulnerable to such feelings. A lack of understanding during therapy could serve to confirm internal maladaptive schemas related to their trauma experiences (e.g., “my feelings do not matter”; “nobody can be trusted”; “nobody understands”), thus, negatively affect their healing process from trauma related wounds.

Strengths and Limitations

The current study was the first to assess four perspectives of traumatized youths’ alliance. The study’s sample represents a natural sample from an ordinary mental health clinic, however, with an overweight of girls and youth with at least one Norwegian-born parent. The alliance was assessed by two alliance measures specifically designed to evaluate youths’ alliance; PTSS were assessed by self- and clinician reports, which helped to reduce shared method variance. However, some study limitations must be mentioned. First, the relatively low n is a limitation; the risk of incurring a Type II error would have been reduced if the sample size were larger. Second, our attempt to control for therapist-effects by the use of hierarchical analyses resulted in unstable models, so we only used single-level analyses. Although an inclusion of therapists as a control variable in the recomputed models did not change the overall findings, the improvement in model fit from entering this variable in the model predicting CPSS T3 from the discrepancy in parent-youth alliance score, and from the significant interaction effect between the discrepancy of the therapist-parents alliance scores and the therapist variable on CPSS outcomes indicates that the relationships between the alliance perspectives and PTS outcomes may differ between therapists. Thus, we cannot rule out that therapist effects may have affected the results and a reliable investigation of therapist effects would require a substantially larger data set. Third, each alliance perspective was assessed at only one time point and not from the exact same treatment session, which is problematic given that the alliance may vary somewhat between sessions. On the other hand, McLeod et al. (2021) found that TPOCS-A and TASC-r scored remained stable across treatment. Forth, we cannot rule out that different trauma experiences have different effects on alliance formation; e.g., the exposure to a sexual abuse versus being involved in a serious accident may relate differently to youths’ internal schemas that may be activated in the therapeutic relationship. Thus, future studies should examine the relationships between different alliance perspectives, and how they pertain to treatment outcomes, within different trauma populations. Last, the study included traumatized youth receiving TF-CBT in a mental health outpatient clinic, and the results may not generalize to other conditions, interventions, or treatment settings.

Conclusion

A main clinical implication from the current study is that therapists should treat traumatized youths’ own alliance ratings as the gold standard for monitoring and optimizing the outcomes from TF-CBT. We found that when youth perceived their alliance to be strong in the mid-treatment phase that includes the trauma narrative work they had better outcomes. This may be related to greater involvement by youth in this task followed by more profitable processing of their trauma experiences. Assessing youth perceived alliance can also help therapists uncover motivational problems that hinder client engagement in therapy (Zandberg et al., 2015). Given the importance of having a strong alliance, it was encouraging that the majority of the youth rated their alliance as high. In general, therapists seem to have succeeded in developing the youths’ sense of being in a trusting and collaborative relationship during TF-CBT. Nonetheless, from the current study’s results, it seems that therapists are not very precise at assessing the therapeutic relationship thus they cannot lean solely on their own clinical expertise. It may also be helpful for the therapist to consult parents about their perspective of the child’s alliance since they seem to be able to capture their child’s therapeutic experiences. Furthermore, therapists should continuously monitor the youths’ perceived alliance and symptom development since an overestimation of the alliance was related to poorer treatment responses. Therapists’ use of rapport-building behaviors seems to strengthen traumatized youths’ experience of the alliance and attending to trauma-related aspects does not seem to hinder the alliance-building process (Ovenstad et al., 2020). Thus, providing and soliciting continuous feedback from traumatized youth on the therapeutic bond and agreement on the treatment tasks may be important for maintaining an optimal agreement and collaboration during the therapeutic process, and facilitate greater treatment response.

Acknowledgements

We acknowledge Tore Wentzel-Larsen for his help and guidance with the statistical analyses.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This research was supported by the Norwegian Research Council (Project no. 190756/H10); the Norwegian Directorate of Health; and the Norwegian ExtraFoundation for Health and Rehabilitation. ClinicalTrials.gov Identifier: NCT00635752.