Parent–Child Interaction Therapy for Children with Disruptive Behaviors and Autism: A Randomized Clinical Trial

The purpose of this study was to evaluate the effectiveness of PCIT in reducing oppositional behaviors and increasing positive parenting behavior among children (4 to 10 years) with ASD (without intellectual disabilities). This study expands the PCIT effectiveness research by including children with varying levels of ASD and who may fall outside the standard age-range for PCIT. The study included three hypotheses: (1) PCIT will result in a significant decrease in parent-reported disruptive behaviors; (2) parent and child interactions (child compliance rates and parenting skills) will significantly improve over the course of PCIT; and, (3) PCIT will result in significant improvements in parent efficacy and parent mental health (stress and depression). The second goal of the study was to perform exploratory analyses to assess the differential impact of the full PCIT protocol by autism severity, medication use, and language level on the disruptive behaviors of children across the autism spectrum. Finally, this study assessed parental satisfaction with PCIT.

The study design followed a step-wise model for conducting psychosocial interventions for ASD, as outlined by Smith et al. (2007) and in accordance with the guidelines adopted by the National Institute of Mental Health (NIMH). This research design adheres to the recommendation that when applying PCIT to a new population, it should first be empirically tested in its standard form to determine its efficacy before any modifications are made to the model (Masse et al., 2007; McCabe et al., 2005). Thus, our study evaluated the efficacy of PCIT in its manualized form (Eyberg & Funderburk, 2011) for children with ASD and behavioral problems.

Families were recruited from the eastern United States. The region included 10 cities and 6 counties in two states located in socio-economically and culturally diverse areas that ranged from rural to urban and suburban settings. The region consisted of 61% White, 31% Black, and 8% other ethnic groups. Fifty-five female and eight male caregivers (N = 55 families) and their 4- to 10-year-old children participated in the present study. All adult caregivers living in the home were encouraged to participate in this treatment as research suggests that dual-parent involvement (e.g., mother, father) leads to better maintenance of treatment gains (Bagner & Eyberg, 2003). Although eight fathers participated in treatment, the primary caregivers in each household were identified as the participating mothers. Therefore, data for this study only include information from the primary caregivers (i.e., mothers, in this case).

Children who participated in the study (Table 1) were mostly boys (85.5%) with a mean age of 7.15 years (SD 1.72). Their racial/ethnic composition was 65.5% White, 16.4% Black, 9.1% Latinx, 9.1% other ethnicity. Most children (80%) were referred by pediatric health care professionals, 12% were referred by teachers, and 8% were self-referred. At the time of intake, 56.4% were prescribed medication to address behavioral issues. Most children came from families with total household incomes between $50,000 and $99,000 (n = 29; 52.7%). To be included in the study, children had to demonstrate at-risk or clinically significant externalizing behaviors (Behavior Assessment Scale for Children [BASC]; Reynolds & Kamphaus, 1992), be diagnosed with ASD by a health professional prior to the study (confirmed and assessed for severity using the Child Autism Rating Scale [CARS]; Schopler et al., 1980), and obtain a receptive language age equivalent of 2 years or higher (Peabody Picture Vocabulary Test [PPVT]; Dunn & Dunn, 2007) to ensure the child was able to follow basic parental commands. Parents were asked to report if the child had been previously diagnosed with an intellectual disability (yes/no). No parents endorsed concerns which was confirmed through a check of the children’s medical records. However, no cognitive data was collected in the present study.

Overall, 181 interested families responded to study recruitment. Please see Fig. 1 for an overview of the screening and recruitment process. Recruitment took place over 10 months, and assessment and treatment completion took 18 months. In total, 163 phone screens were conducted and 92 families qualified and were scheduled for a clinical intake assessment. Of those families that completed the intake process, 55 met criteria for inclusion. Families were excluded if based on the screening measures the child (a) had limited receptive language (n = 8), (b) lacked severe behavior problems (n = 10), (c) were not previously diagnosed with ASD by a health professional (n = 4), or (d) the family did not complete the intake process (n = 9). There were three families screened as eligible for the study that cancelled their participation from the original group. Families not meeting the inclusion criteria (n = 37) were given feedback and appropriate recommendations for services. Upon completing the clinical intake, children were stratified based on the dichotomized variable of psychiatric medication use (yes/no), then matched using the continuous variables of externalizing behaviors (BASC), severity of autism (as measured using the CARS), and age. Families were assigned using stratified randomization to the control or treatment (PCIT) group using a research randomizer program. Randomization was determined by the lead researchers (first and second authors). Two children did not match and were not included in the randomization process, yielding a total of 30 treatment group (TG) families and 25 control group (CG) families. Treatment for children with ASD in the study location was limited. Many families were on a wait-list of 12–18 months for ABA. Families in both the TG and CG continued routine community care throughout the study, which primarily consisted of speech therapy services (TG = 38%, CG = 40%).

The study was approved by the Eastern Virginia Medical School and the Children's Hospital of the King's Daughters Institutional Review Boards. Children were recruited by providing study information packets to pediatric offices, the Tidewater Autism Society of America, Community Service Boards, and school systems. Children were screened prior to enrollment on level of disruptive behavior (BASC), autism severity (CARS), and receptive language (PPVT). Enrolled families completed a baseline assessment (Time 1: pre-treatment) that included parental report questionnaires and parent–child observations which were videotaped and coded by research personnel. Approximately 8 weeks after the Time 1 assessment, all TG and CG families were scheduled to complete an interim-treatment assessment (Time 2). This interim assessment marked the transition from CDI to PDI for treatment families. The Time 3 assessment occurred post-treatment (upon graduation for TG), 8 weeks after Time 2. All families were compensated for study participation at Time 1 and Time 3 assessment visits. Data collection and treatment delivery was conducted in a medical center. Enrollment, assessments, and treatment delivery were conducted by the research team.

Primary outcome measures included measures of externalizing behavior and observational measures. Secondary outcome measures included assessments of parenting stress, depression, and locus of control.

PCIT sessions were conducted by a clinical psychologist once a week and lasted between 60 and 90 min. Families in the TG condition received the entire protocol of PCIT (i.e., both CDI and PDI phases) unless families terminated before treatment completion. In both phases of treatment (CDI, PDI), therapists actively coached parents toward understanding of the therapeutic interaction skills as assessed during a 5-min parent–child observation (DPICS) at the start of the session. On average, families achieved CDI skills competencies in 6.2 sessions and PDI in 5.9 sessions. Throughout treatment, parents were asked to practice the skills at home daily in 5–10 min sessions, initially focusing on CDI skills and then incorporating PDI skills at times when a command was necessary. The therapists included a licensed clinical psychologist and a supervised post-doctoral clinical psychology fellow, each of whom attended a 40-h PCIT training conducted by a PCIT Global Trainer. All therapy sessions were videotaped, and 50% of the session tapes from each family were randomly selected and checked independently by two coders for integrity using the PCIT treatment manual checklist. Accuracy was 95% with the treatment protocol. In addition, supervision from a PCIT Global Trainer was received regularly.

Mean scores on parent-report questionnaires as well as behavioral observation counts and ratios derived from the DPICS were primarily analyzed using Repeated Measures MANOVAs. This is a suitable technique given the uniformity of the assessment schedule. Additionally, it is preferable to the alternative of multiple univariate tests as it detects patterns between multiple dependent variables which may not otherwise arise in univariate tests. Additional univariate comparisons are presented as follow-ups to the MANOVAs, as are graphical displays of observed effects.

The two parent-report measures that were used to assess disruptive and oppositional behavior observed by the parent in the home included the ECBI (Table 2; Fig. 2) and the BASC (Table 2). On the ECBI Intensity scale, a significant Time X Group interaction was observed, such that children in the TG demonstrated a much steeper decline in behavioral intensity (Wilk’s λ (2, 39) = 47.28, p < .001; Partial η² = .48). Univariate tests reveal that children receiving PCIT demonstrated lower intensity of behavior problems at Time 2 (F(1,39) = 8.21, p = .006) and significantly increased that difference (F(1,39) = 30.76, p < .001) at Time 3 (see Fig. 2) compared to CG families. A similar interaction was observed for the ECBI Problems scale (Wilk’s λ (2, 37) = 5.77, p = .007; Partial η² = .238), revealing that children in the TG demonstrated significantly fewer problems at Time 3 (F(1,37) = 8.41, p = .006), but not at Time 2. Overall, each child in the treatment group scored in the non-clinical level of the ECBI on both the Problems and Intensity scale at the conclusion of the treatment.

When measured with the BASC Externalizing Problems subscale (Table 2), a similar Time X Group interaction was observed, such that children in the TG demonstrated continued decline in externalizing problems at Times 2 and 3, whereas the CG leveled off after Time 2 (Wilk’s λ (2, 36) = 11.79, p < .001; Partial η² = .40). Univariate tests revealed that children in the TG demonstrated lower levels of Externalizing Problems at Time 3 (F(1,37) = 6.61, p = .014) but not at Time 2.

In terms of PCIT “Do” skills (measured by DPICS labeled praises, behavior descriptions, reflections), parents in the TG showed significant increases over the three assessment periods (Wilk’s λ (2,37) = 49.81, p < .001; Partial η² = .73). Univariate tests indicated no statistically significant differences between groups at baseline with TG parents showing significantly more “Do” behaviors at Time 2 (F(1,38) = 49.07, p < .001) and maintaining this change (F(1,48) = 299.19, p < .001) through Time 3 (See Fig. 3).

PCIT “Don’t” skills (DPICS commands, questions, criticism) showed a similar but inverse pattern over time. Parents in the TG showed significant decreases in “Don’t” behaviors over the three assessment periods (Wilk’s λ (2, 37) = 21.33, p = .001; Partial η² = .54). Univariate tests indicated no statistically significant differences between groups at Time 1, with TG parents showing significantly fewer “Don’t” behaviors at Time 2 (F(1,38) = 36.68, p < .001), and maintaining this change (F(1,38) = 63.53, p < .001) through Time 3 (Fig. 3).

Parents and children were observed for compliance/command ratio during the PLP and CU periods of the DPICS (Table 2). Treatment families showed a statistically significant Time X Compliance ratio change during the PLP activity (Wilk’s λ (2, 36) = 15.85, p < .001; Partial η² = .47) with significant relative improvement occurring at Time 3 (F(1,37) = 21.22, p < .001). A very similar pattern was observed during the CU activity with the Time X Compliance being significant (Wilk’s λ (2, 37) = 7.99, p = .001; Partial η² = .30) and the major improvement being observed for Time 3 (F(1, 38) = 10.07, p = .003).

The parent-level variables revealed a large variation, with only parental stress demonstrating a statistically significant change between the TG and CG. On the PSI-SF Total Stress scale, a significant Time X Group interaction was observed, such that parents in the TG demonstrated a steeper decline in parenting stress over the course of the study (Fig. 4; Wilk’s λ (2, 37) = 4.67, p = .016; Partial η² = .20). No significant difference were found between the TG and the CG on the PLOC-SF (p = .072) or BDI-II (p = .208; Table 2).

Findings on the TAI show that TG parents found the techniques helpful in regards to disciplining their child (91%) and teaching their child new skills (93%). All parents of children in the TG felt their relationship with their child was better than before the program and that their child’s behavior problems and compliance with parental commands and requests had improved. Ninety-three percent of the parents reported satisfaction with the progress their child had made in regards to their general behavior and many (86%) felt that the program helped with other general personal and family problems. These findings suggest that families of children with a range of ASD presentations were highly satisfied with PCIT.

Previous studies evaluating the impact of autism treatment approaches have been conducted with middle- to upper-middle-class families with an estimated annual treatment cost ranging from $25,000 to $60,000 per child and requiring up to 25 h per week (Solomon et al., 2007). For many families, there is a lack of both treatment availability and financial resources (Mackintosh et al., 2012). This study demonstrates that PCIT offers an innovative, more cost-effective approach to delivering an evidence-based therapy to a diverse population. Specific benefits of this model of treatment include: (a) a family-based approach that addresses caregivers’ capacity to manage ASD-related behaviors; (b) direct-coaching to maximize parental learning and retention; (c) a time-limited model; and (d) a treatment model that can be wildly disseminated. Historically, few studies of behavioral treatment of ASD employed an experimental design (2 of 68 studies per a 2001 meta-analysis; Lord & McGee, 2001); however, even though recent studies for various early interventions are utilizing randomized controlled trials, only a few of these are behaviorally-based treatments. Additionally, many studies have substantial bias and limitations preventing robust findings for this population (French & Kennedy, 2018; Tachibana et al., 2017). Therefore, the use of a controlled randomized design for this study addresses a significant gap in the research for this treatment and population.

Treatment research for children with ASD has primarily focused on the benefits of early intensive behavioral intervention (e.g., Remington et al., 2007; Rogers & Vismara, 2008). Many children receive some form of intensive behavioral training after being diagnosed between the ages of 3 to 4; by age 5, the treatment options begin to significantly decrease. However, the majority of children with ASD continue to experience language, social, and behavioral difficulties throughout their school years (Marsh et al., 2017; McKean et al., 2017). Additionally, for families receiving later diagnoses or who are limited by accessibility of resources (Godon-Lipkin et al., 2016), PCIT may be a gateway to reduce disruptive behaviors, improve the effectiveness of other interventions, and increase accessibility due to wide availability of PCIT providers in the United States (Scudder et al., 2017; Soke et al., 2018). A research review (Solomon et al., 2008) concluded that children with autism are significantly at-risk for problematic behaviors which, without intervention, are more likely to worsen than improve. Despite this problem, our understanding of effective behavioral treatment of children with ASD is limited. PCIT may be one answer.

Future studies may strive to overcome the present study’s limitations. For example, the study did not include an alternative treatment control; therefore, the results must still be considered provisional. Additionally, future research on PCIT may benefit from further analyses regarding how treatment may differentially impact children at various levels of autism functioning. Researchers should focus on PCIT adaptations for children at lower levels of functioning. Moreover, while the present study was unique in that it included youth classified as severely autistic, it excluded youth with a comorbid intellectual disability. Future studies may also benefit from measuring, reporting, and controlling for youth’s cognitive functioning (e.g., IQ) to lend insight in PCIT’s effectiveness of youth with ASD and diverse intellectual levels.

Although this study did not measure non-disruptive behaviors of autism, such as self-stimulation, eye contact, language, and social engagement, positive changes in these behaviors were both observed by therapists and reported by parents and teachers. Next steps include assessing these behaviors in the context of other child characteristics including autism severity, language level, and cognitive functioning.

As PCIT did not significantly decrease parent depression, future studies should explore ways to improve these measures of parent-functioning, possibly through the addition of a parent psycho-education/treatment module or referral for individual parental therapy. A larger sample size would allow for more thorough analyses of other possible correlates, such as age and gender.

In families with typically-developing children, PCIT has been found to provide treatment effects lasting up to two years (Boggs et al., 2004). In order to determine whether the treatment gains demonstrated by families completing PCIT will be maintained, follow up data has been collected three months after treatment completion. The results of these data will be reported in future publications. Furthermore, in-depth analyses of PCIT’s impact on child language from this study will also be presented in future publications.

Based on the results of this study, PCIT should be considered a viable treatment for children with ASD and behavioral problems (without intellectual disabilities). In addition, the therapy can effectively prepare children for other intense and focused ASD therapies requiring cooperation and attention (Masse et al., 2007; Williford et al., 2019) and take advantage of needed therapies (e.g., speech, occupational therapy) when behavioral problems limit their ability to engage in the therapeutic process. Parents were satisfied with their experience with PCIT and reported significant improvement in child behavior, compliance, and parent–child interactions. This study also expands the limited research on PCIT among children with autism by improving the generalizability to children of varying autism severity.