Therapeutic alliance in youth with autism spectrum disorder receiving cognitive-behavioral treatment for anxiety

Participants

Participants were 64 children (ages 7–16 years) with ASD and their parents. Participants were recruited as part of two randomized controlled trials evaluating the efficacy of a modular CBT for anxiety in children (n = 34) and adolescents (n = 30) with ASD at the University of South Florida. Inclusion criteria were as follows: (a) diagnosis of autistic disorder, Asperger’s syndrome, or pervasive developmental-disorder, not otherwise specified (PDD-NOS) and (b) diagnosis of separation anxiety disorder, social phobia, generalized anxiety disorder, or obsessive-compulsive disorder per the Anxiety Disorders Interview Schedule–Child and Parent Versions (ADIS-IV-C/P). Exclusion criteria were as follows: (a) full scale or verbal comprehension IQ <70 on standardized testing; (b) bipolar, psychotic disorder, or suicidality in past 6 months; and (c) currently receiving psychotherapy for anxiety. See Storch et al. (2013) and Storch et al. (2015) for more details regarding participant selection and Table 1 for participant characteristics.

OPEN IN VIEWER

Table 1.

Participant pretreatment characteristics (N = 64) and TASC ratings.

				N (%)
Male				52 (81)
Ethnicity
White				54 (84)
Asian				5 (5)
Latino/Hispanic				7 (11)
Taking psychotropic medication				40 (63)
				M (SD)
Age in years				10.81 (2.25)
PARS 5-item Total				15.53 (2.69)
CGI-S				3.59 (0.77)
CBCL Externalizing Problems Raw Total				16.00 (11.38)
CBCL Internalizing Problems Raw Total				21.00 (10.14)
SRS Raw Total				144.85 (21.93)
	Cronbach’s α	M (SD)	Possible range	Observed range
Child TASC	0.88	37.95 (8.07)	12–48	15–48
Therapist TASC	0.90	35.06 (7.09)	12–48	21–47
Parent TASC	0.76	26.97 (1.89)	7–28	20–28

SD: standard deviation; PARS: Pediatric Anxiety Rating Scale, 5-item total score; CGI-S: Clinical Global Impression–Severity; CBCL: Child Behavior Checklist; SRS: Social Responsiveness Scale; TASC: Therapeutic Alliance Scale for Children.

Procedure

Both clinical trials received approval from the local institutional review board and obtained written, informed consent, and assent from participating parents and children. Participants completed a phone prescreen and then were scheduled for an in-person assessment with an independent evaluator (a post-doctoral- or doctoral-level clinician not involved in the child’s treatment). Independent evaluators were reliably trained, blind to group assignment, and completed pre/post-assessments for the same child 90% of the time. Pre/post-assessments included semi-structured interviews (ADIS-IV-C/P; Pediatric Anxiety Rating Scale (PARS)) and clinician ratings of anxiety severity (Clinical Global Impression–Severity (CGI-S)), as well as parent ratings of autism-related symptoms (Social Responsiveness Scale (SRS)) and child internalizing and externalizing behavior (Child Behavior Checklist (CBCL)). ASD diagnosis was confirmed using combined information from the Autism Diagnostic Observation Schedule–Generic, Autism Diagnostic Interview–Revised (results from prior evaluations were accepted to reduce burden), and clinical judgment. Although efforts to refine anxiety assessment in ASD are underway, the PARS and ADIS-IV-C/P have shown adequate reliability and validity in children with ASD here (Storch et al., 2013, 2015)) and in prior studies (Kerns et al., 2016). Eligible participants were randomized by computer-generated algorithm in a 1:1 ratio to CBT or treatment as usual (TAU). The CBT intervention involved 16-weekly, 60- to 90-min individual and family sessions of the Behavioral Intervention Anxiety in Children with Autism (BIACA), a modular CBT designed to target anxiety and related difficulties (e.g. social skills) in youth with ASD (Sze and Wood, 2007). Modular programs tailor intervention strategies to each child rather than following a uniform protocol. Parents were included for a portion of each session to learn CBT concepts and facilitate child application of these concepts in daily life. In TAU, families could begin, continue, or change any psychosocial or pharmacological treatments for 16 weeks before being offered free BIACA. Post-treatment assessments, completed within 1 week of the last CBT session, included all screening measures plus parent, therapist, and child versions of the Therapeutic Alliance Rating Scale for Children–Revised (TASC-R) and the Clinical Global Impression–Improvement scale (CGI-I). The CGI-I is a 6-point Likert rating of treatment-related improvement (anchors range from “very much worse” to “very much improved”) based on the clinician’s impression of treatment response after considering all information from the post-treatment assessment. Participants who received a CGI-I rating of “very much” or “much improved” were considered “treatment responders.” The primary continuous outcome was change in anxiety severity on the PARS. Participants were all youth who received CBT, including those originally randomized to TAU (n = 26). For further details regarding the administration and psychometric characteristics of study measures and other study procedures, see Storch et al. (2013) and Storch et al. (2015).

Alliance assessment

The TASC-R (Shirk and Saiz, 1992) is a 7-item (parent version) to 12-item (therapist and child versions) measure of therapist–client agreement about therapy goals and feelings about the therapeutic relationship. The TASC total score is based on a composite of all items. Child/therapist versions pertain to child–therapist alliance and the parent version pertains to parent–therapist alliance. The TASC has internal consistency in samples of youth with (α = 0.84–0.93; Klebanoff, 2015) and without ASD (α = 0.81–0.88; Shirk and Saiz, 1992). Social desirability is not explicitly assessed.

Therapeutic alliance ratings and inter-rater agreement

Table 1 presents the internal consistency, means, standard deviations, and range of TASC ratings per therapist, child, and parent reports. Ratings by children, therapists, and parents had acceptable internal consistency (α > 0.70) and suggested relatively strong alliance between parents and therapists as well as children and therapists. Parent ratings were positively skewed.

Pearson correlations evaluated inter-rater agreement. For to inter-rater agreement regarding alliance, child and therapist ratings of alliance were significantly associated (r = 0.27, p = 0.05). Agreement was not significantly different between youth 7–11 years (r = 0.21) and 12–16 years (r = 0.39, z = −0.69, p = 0.49). Parent ratings of parent–therapist alliance were not significantly associated with child ratings (r = 0.02) or therapist ratings (r = −0.07) of child–therapist alliance.

Pretreatment characteristics and alliance

One-way analysis of variance and Pearson correlations assessed relationships between pretreatment child characteristics and alliance (Table 2). Age, gender, ethnicity, initial anxiety severity (per PARS and CGI-S), primary anxiety diagnosis, CBCL externalizing and internalizing behaviors, and pretreatment ASD severity (per SRS) were not significantly associated with therapist, parent, or child ratings of alliance.

OPEN IN VIEWER

Table 2.

Relationship of alliance with pretreatment characteristics and treatment response.

Child characteristics	Child TASC	Therapist TASC	Parent TASC
	Pearson’s R
Age in years	0.17	0.13	0.03
PARS 5-item Total	−0.03	−0.06	−0.12
Pretreatment CGI-S	0.05	−0.09	0.02
CBCL Externalizing Raw Total	−0.10	−0.19	−0.02
CBCL Internalizing Raw Total	0.14	−0.20	0.05
SRS Raw Total	0.01	−0.12	0.01
Treatment response	M (SD)
Responder	40.15 (6.61)	37.21 (6.55) a	26.82 (2.07)
Non-responder	37.14 (9.98)	30.06 (5.75) a	27.64 (0.63)
	Pearson’s R
Change in PARS	0.20	0.40**	−0.22
Post-treatment CGI-S	−0.15	−0.30*	−0.02

TASC: Therapeutic Alliance Scale for Children; PARS: Pediatric Anxiety Rating Scale, 5-item total score; CGI-S: Clinical Global Impression–Severity, rating based on results of all measures collected at that time point; CBCL: Child Behavior Checklist; SRS: Social Responsiveness Scale; SD: standard deviation.

a

Significantly different at p < 0.01 level: *p < 0.05; **p < 0.01.

Association of alliance with treatment outcome

Therapist-rated alliance was significantly stronger for treatment responders M (SD) = 37.21 (6.55) versus non-responders M(SD)=30.06 (5.75), t(51) = −3.77, p < 0.01, and associated with reduced global severity ratings (CGI-S; r = −0.30, p < 0.05) and greater reductions in anxiety (PARS change; r = 0.40, p < 0.01) at post-treatment; however, these relationships were not found for parent and child ratings of alliance (Table 2).

Limitations

The association of alliance with treatment outcome in this study is difficult to interpret because alliance was only assessed post-treatment. Therapeutic alliance may play a role in children’s participation and ability to benefit from treatment; alternatively, therapists and youth may rate alliance higher when youth respond positively to treatment (Marker et al., 2013). The conflation of therapeutic progress and process seen here is a common issue in alliance research, which finds alliance ratings collected later as opposed to earlier in treatment more predictive of outcome (Shirk and Karver, 2003). Furthermore, although alliance ratings were positive, particularly for parents, this may reflect social desirability, which was not directly assessed by the TASC-R. Repeated, longitudinal and observational alliance assessment by independent raters over the course of therapy may help address these issues in future research (McLeod, 2011). Another limitation of this study was the lack of a comparison group. Direct comparisons of youth with and without ASD in extant (see Klebanoff, 2015) and future studies may suggest differences in the ways in which youth with and without ASD build and use alliances with their therapist to benefit from CBT.