Randomized comparative efficacy study of parent-mediated interventions for toddlers with autism

Abstract

Objective

This study compared effects of two parent-mediated interventions on joint engagement outcomes as augmentations of an early intervention program for toddlers with autism spectrum disorder (ASD).

Method

Participants included 86 toddlers (range 22 – 36 months) with ASD and their primary caregiver. Caregiver-child dyads were randomized to receive ten weeks of hands-on parent training in a naturalistic, developmental behavioral intervention (Joint Attention, Symbolic Play, Engagement and Regulation—JASPER) or a parent-only psychoeducational intervention (PEI). Dose was controlled in terms of researcher-parent contact and early intervention services received by the child.

Results

Results yielded significant effects of the JASPER intervention on the primary outcome of joint engagement. The treatment effect was large (Cohen’s f²=.69).and maintained over the six-month follow-up. JASPER effects were also found on secondary outcomes of play diversity, highest play level achieved, and generalization to the child’s classroom for child-initiated joint engagement. The PEI intervention was found to be effective in reducing parenting stress associated with child characteristics. All secondary effects were generally small to moderate.

Conclusions

These data highlight the benefit of a brief, targeted, parent-mediated intervention on child outcomes. Future studies may consider the combination of JASPER and PEI treatments for optimal parent and child outcomes. Trial registry # NCT00999778.

Young children with autism spectrum disorder (ASD) display significant impairments in early social communication skills. These include the initiation of joint attention gestures (e.g., showing toys to others, pointing to share, and coordinated eye gaze between objects and people) and the ability to jointly engage in social interactions with others (Adamson, Bakeman, & Deckner, 2004; Kasari, Freeman, & Paparella, 2006; Sigman, Mundy, Sherman, & Ungerer, 1986). These impairments uniquely discriminate children with ASD from children with other developmental delays and typical children of similar mental age (Mundy, Sigman, Ungerer, & Sherman, 1987). Given the large literature on the importance of these skills as predictors of later development, early interventions have increasingly targeted these skills, especially for very young children (Kasari, et al., 2005).

The foundational nature of these skills for the ability of children to develop relationships with others leads researchers to consider parents as important mediators of change and potential targets of intervention. However, the current evidence for parent-mediated interventions for children with ASD is mixed. For example, experimental low intensity, three-month, short-term parent-mediated interventions for very young children who are at risk for ASD have not demonstrated significantly greater change in parent and child outcomes relative to community-based, treatment-as-usual interventions (Carter et al., 2011; Rogers et al., 2012). Longer -term interventions of nine months have shown greater effects for children who begin intervention before age two years (Wetherby et al, 2014). However, for older children with confirmed diagnoses of ASD, these same types of interventions of 12–24 sessions over 3 to 6 months have improved parent responsiveness and child outcomes to a significantly greater extent when compared to treatment-as-usual community groups (Green et al., 2010; Kasari, Gulsrud, Wong, Kwon, & Locke, 2010) or an alternative treatment (Kasari et al, 2014).

What might account for these age-related differences? One notion is that older children display more readily apparent delays relative to other children. Thus, parents are better able to recognize the specific needs of their children. Another speculation is that children who have confirmed diagnoses are often receiving a range of intervention services in the community, thereby increasing the difficulty in identifying the augmenting effects of parent-mediated interventions against the background of more intensive treatments. Currently, we are unclear on the absolute dose needed and the best methods for teaching parents to achieve the most optimal child outcomes.

Other factors may also affect parent and child outcomes. Increased stress and worry have been well documented for parents of children with ASD and suggest the need for specific interventions to address parental mental health concerns (Schieve, Blumberg, Rice, Visser, & Boyle, 2007). The increased stress may result from many sources: distress from the impact of their child’s diagnosis, the strain of additional parenting roles and demands, including expectations that they deliver interventions to their young child, as well as time lost from work, and increased medical costs associated with caring for a child with ASD (Cidav, Marcus, & Mandell, 2012). Early interventions that provide parenting strategies through psychoeducational programs have significantly decreased parental stress in these families (Feinberg et al., 2014; Tonge et al., 2006). Although effects of psychoeducational approaches on child outcomes are rarely tested, a recent study found that a parent education and counseling program that included behavior management strategies helped to improve child adaptive behaviors (Tonge, Brereton, Kiomall, Mackinnon, & Rinehart, 2014). Thus, the combination of counseling and education for ASD-specific parenting strategies may be effective for improving both parent and child outcomes.

Individual child characteristics can also affect treatment outcomes. These individual characteristics may include a wide range of features such as language ability, cognition, and severity of ASD. This heterogeneity may account for findings that parent-mediated interventions have had very limited effects on child outcomes. Child treatment gains have been limited to particular subgroups, such as those entering the intervention with the greatest impairment, or those identified as displaying greater pre-intervention object interest (Carter et al, 2011; Siller, Hutman & Sigman, 2013). Overall, there is a need to identify other parent-mediated interventions for young children with ASD that are more robust in significantly improving parent and child outcomes.

This comparative efficacy study attempts to advance intervention research by testing whether a parent-mediated intervention involving direct parent coaching with the child is more efficacious than a parent-education model (without direct involvement of the child) in improving parent and child outcomes. The study treatments were added to existing early intervention services, controlling for type and dose of these services. All of the children were toddlers with ASD, between the ages of 22 and 36 months, and attended the same 30-hour per week, early-intervention program.

Families attending the early-intervention program were randomized to receive either one hour per week of parent training based on a manualized parent psychoeducation program for parents of children with autism (Psychoeducational Intervention -PEI; Brereton &Tonge, 2005) or a hands-on manualized, parent-mediated program aimed at improving core deficits in their toddlers (Joint Attention Symbolic Play, Engagement, and Regulation-JASPER, UCLA, 2005, unpublished manual). Controlling for the overall dose of intervention, we compared the approaches on the primary outcome of joint engagement between parent and child. Joint engagement, as defined, requires the active and reciprocal engagement of both parent and child. Secondary measures included child-initiated joint attention, functional and symbolic play types (reflecting play flexibility and diversity), play level, and generalization of joint engagement to the child’s classroom. We hypothesized that improvements in child engagement with objects and people, joint attention gestures, and play skills would be significantly greater in the JASPER condition as compared to the PEI condition. Because one focus of the PEI condition is on reducing parent stress, we also hypothesized that stress would be reduced to a greater extent for parents in the PEI condition relative to the JASPER condition. Child and parent measures were collected prior to intervention (Pre-Treatment), at the end of intervention (Post-Treatment), and at 6 months follow-up.

Methods

Participants

Participants were recruited from the same outpatient early intervention (EI) program consisting of 30 hours per week of a combination of behavioral, speech, and occupational therapies for 10 weeks. A University Institutional Review Board approved the study and parents provided written consent. Children were included if they were younger than 36 months, had a clinical diagnosis of ASD confirmed by independent testers with the Autism Diagnostic Interview-Revised (ADI-R; Lord, Storoschuk, Rutter, & Pickles, 1993) and the Autism Diagnostic Observation Schedule (ADOS; Lord, Rutter, DiLavore, & Risi, 2001), had no significant physical disabilities, and parent and child were available for follow-up assessments (e.g., not international residents). A total of 86 parent-child dyads enrolled in the study. Children were on average 31.5 months. See Table 1 for participant characteristics.

Table 1.

Participant Characteristics

Child and Parent Characteristics: N (%)	JASPER N=43	PEI N=43	Total	Test	p
Chronological Age (Months): Mean (SD)	30.7 (3.5)	32.3 (2.7)	31.5 (3.2)	F(1,84)=6.3	.01 *
Gender
Female	8 (19%)	8 (19%)	16 (19%)	f2(1)=0.0	1.00
Race/Ethnicity
African American	0 (0%)	2 (5%)	2 (2%)
Caucasian	27 (63%)	26 (60%)	53 (61%)
Hispanic	3 (7%)	4 (9%)	7 (8%)
Asian/PI	4 (9%)	6 (14%)	10 (12%)	f2(4)=4.5	.34
Other	9 (21%)	5 (12%)	14 (17%)
Mullen Age Equivalency (Months): Mean (SD)
Developmental Quotient	68.0 (20.3)	68.1 (20.6)	68.0 (20.3)	F(1,84)=0.0	.98
Age of Mother	36.9 (4.4)	34.9(4.7)	35.9 (4.6)	F(1,83)=3.9	.05
Maternal Education
Years of Education	17.2(2.3)	16.4 (2.6)	16.8(2.4)	F(1,84)=2.6	.11

A random numbers list was used to randomize participants. Participants were randomized to one of two treatment conditions –JASPER-Parent-mediated or Psychoeducational Intervention (PEI) – in addition to the 30-hour early intervention program. Randomization and analyses were conducted by an independent data-coordinating center. Testing for the success of randomization showed that the two groups were matched on demographics (Table 1) except for age of entry, in which the JASPER group, with an average entry age of 31 months, was statistically significantly younger than the PEI group with an average entry age of 32 months. Three dyads discontinued treatment and another ten did not complete follow-up assessments. See Participant Flow Chart in Figure 1.

Figure 1.

Recruitment Flow Diagram

Due to the intensity of the EI program, families discontinued outside early intervention services during the 10-week program. During the follow-up period all children continued early intervention services. A large percentage (76%) maintained full time services (>=30 hrs per week). There was no significant difference in service utilization between the JASPER Group (73%) and the PEI Group (79%, p=.56) during the follow-up period.

Interventions

Each intervention model involved one hour of interventionist contact per week.

Psychoeducational Intervention (PEI; Brereton & Tonge, 2005)

The aim of the PEI intervention is to provide individual education and support to parents of young children with autism. This intervention provided 1:1 interventionist meetings with the parents in informational sessions of 1 hour per week for 10 weeks. Sessions covered specific topics each week and parents were able to ask questions specific to their own child’s development directly with their therapist. The content of the manualized intervention included information on autism, details of specific behavioral impairments, principles of managing behavior, strategies for teaching new skills, improving social interaction and communication, service availability, managing parental stress, and sibling, family, and community responses to autism.

JASPER Parent-mediated model

The JASPER model included the parent and his/her child for one hour per week for 10 weeks (2 sessions of 30 minutes per week) with active coaching of the parent by a trained interventionist. A total of 20 sessions were completed delivering an equivalent amount of therapist contact as the PEI condition per week (one hour total of interventionist contact) to control for this variable.

JASPER is an empirically-supported and manualized treatment for toddlers and preschoolers with a primary focus on sustaining periods of joint engagement and increasing joint attention gestures and play skills (Kaale, Smith, & Sponheim, 2012; Kasari et al., 2006; Kasari et al, 2014; Kasari, Paparella, Freeman, & Jahromi, 2008; Kasari et al., 2010; Kasari, Gulsrud, Freeman, Paparella, & Hellemann, 2012; Lawton & Kasari, 2012). Intervention sessions were based on developmental and behavioral principles consistent with JASPER. Parents were first taught to recognize the child’s current developmental level of play and use of social-communication gestures. Capitalizing on the child’s current level of play and interests, caregivers provided opportunities for the child to initiate interest in a toy/activity and to establish jointly engaged play routines. Parents used a number of strategies to keep children engaged while also improving their frequency of social communication gestures, spoken words, and play acts. Parents were taught to use these strategies in a structured sequence that has been previously tested (Kasari et al., 2010; 2014).

Therapists, Supervision, and Treatment Integrity

A post-doctoral clinical psychologist supervised all interventionists, and the majority of therapists were PhD candidates in Human Development and Psychology. Three of the eight different therapists were BA level therapists, all in the JASPER arm. Prior to beginning study treatments, therapists were required to demonstrate >90% fidelity on a PEI or JASPER-specific measure of fidelity with practice children, typically requiring 2 to 6 months of training. Supervision of active cases was conducted weekly.

The project coordinator used a random numbers list to select sessions for fidelity ratings. To maintain blinding, the sessions were observed through a one-way observation window. Twenty percent of sessions were rated for treatment integrity with average ratings for PEI of 92.7% (78.6%–100%) and for JASPER, 91.4% (75%–100%).

Measures

Examiners blind to treatment status conducted all pre-treatment, post-treatment, and follow-up assessments. Separate video coders were also blind to treatment condition.

The primary outcome was a measure of joint engagement coded from the Parent Child Interaction. A 10-minute interaction recorded between parent and child was collected pre-treatment, post-treatment, and at the 6-month follow-up. Parents were asked to engage in free play with their toddler as they normally would at home using a standard set of toys (including dolls, dishes, puzzles, trucks, shape sorter, and blocks). Children’s joint engagement was coded using an adapted coding system from Adamson and colleagues (2009) where time jointly engaged included time in supported joint engagement and coordinated joint engagement, with or without symbols. Supported joint engagement was coded if the child demonstrated awareness of the parent’s participation (e.g., takes turns with the same object, follows parent suggestion in play with the object) while coordinated joint engagement was coded when the child directly acknowledged the partner through triadic eye gaze, language, and/or gesture. Instances of supported joint and coordinated joint engagement were coded as symbol infused if the child followed or showed evidence of attending to parent language or used language him/herself. For the purposes of this study, we combined supported joint and coordinated joint with and without symbols into a single variable of joint engagement consistent with previous studies (Kasari et al., 2010; 2014).

Secondary outcomes from the parent-child interaction included child play skills, play level, and joint attention. The child’s play behaviors recorded during the mother-child interaction were coded for types (i.e., diversity) of functional and symbolic play acts (Kasari et al., 2006). Functional play types involved counting the number of different novel forms of functional play from “relational” to “child directed play to dolls.” Symbolic play types included counting all novel symbolic play types from “substitutions” through “multiple schemes” (Lifter, Sulzer-Azaroff, Anderson, Edwards, & Cowdery, 1993). In addition, the highest play level that the child was observed to maintain during the interaction was assigned a numerical value for subsequent analyses. Highest play level achieved is a function of functional and symbolic play acts.

Child’s frequency of initiating joint attention skills was also coded in the parent child interaction (e.g., coordinated joint looks, pointing to share attention, and showing). We collapsed the frequency of joint attention skills that were spontaneously initiated into a summary variable of initiating joint attention.

Graduate students, not involved in other aspects of the study and blind to child treatment condition, coded the videotapes according to a protocol used in several other studies (Harris, Kasari, & Sigman, 1996; Kasari et al., 2006; Kasari et al., 2008). The reliability of the observational variables of interest, including total time jointly engaged (ICC=.95), initiating joint attention skills (ICC=.97), and number of functional (ICC= .95) and symbolic (ICC= .98) play types was excellent.

Other secondary outcome measures included cognitive and language assessments, parenting stress, and an observational measure of joint engagement in the classroom. The Mullen Scales of Early Learning (MSEL; Mullen, 1989) was used to assess general cognitive ability. The MSEL yields an early learning composite score based on scores for visual reception, gross motor, fine motor, and receptive and expressive language. This measure was collected pre-treatment and at the 6-month follow-up. The Reynell Developmental Language Scales (Reynell, 1977) were used to assess the receptive and expressive language abilities of children in the sample. The scales yield raw scores on Expressive Language and Verbal Comprehension, and these raw scores were transformed into age equivalencies. The Reynell was administered pre-treatment and at the 6-month follow-up.

The Parenting Stress Index

(PSI; Loyd & Abidin, 1985) was used to obtain a measure of parent-reported stress. The PSI consists of two domains: one associated with parent characteristics and the other with child characteristics. The parent domain consists of items targeting sources of stress in the parent-child system related to parental functioning and consists of seven subscales (e.g., parental attachment, sense of competence, relationship with spouse, and depression). The child domain consists of items reflecting perceptions of child characteristics that make it difficult for parents to fulfill their parenting role and consists of six subscales including child distractibility, demandingness, and child mood. Items are rated on a Likert-type scale and summed with higher scores reflecting greater dysfunction. The PSI was administered pre- and post-treatment and at the 6-month follow-up.

Classroom Observations

Children were observed in their classrooms in play with their teacher for 15 minutes at pre- and post-treatment. During the 15-minute play interaction, the classroom teacher, who was blind to treatment status, was instructed to engage in structured play with the child. As videotaping was not permitted in the classroom, independent observers coded the classroom play interactions in 1-minute intervals and coded the child’s predominant engagement state. The engagement states consisted of six mutually exclusive categories: unengaged, onlooking, object engaged, person engaged, supported joint engagement, and coordinated joint engagement (Adamson et al., 2004). The variable of interest was time in joint engagement, and consistent with coding from the parent child interaction, supported joint engagement and coordinated joint engagement with and without symbols were collapsed into one variable of joint engagement. Six observers were trained to conduct the classroom observations over the course of the study (average Kappa=.81, range .73–.97).

Statistical Methods

One of the challenges of this study was the inherent structure of the behavioral measures. The majority of the variables of interest were right skewed, and some of these behaviors were comparatively rare in this population (e.g., only 6% of the children in the sample showed any joint attention skills at baseline). To avoid potential bias or inflation of Type-I errors, we used a conservative approach. First, we determined whether the variable was zero inflated as suggested by Min and Agresti (2005). Then, using the Heilbrons (1994) approach, we tested if there was a strong enough floor effect to suggest that the measure was too difficult for part of the population. If this was the case, the variable was estimated using a Poisson hurdle model, in which the effect of the intervention was estimated simultaneously, but separately for the participants who were and were not yet in the range of ability that is covered by the scale. If there was no significant floor effect, we analyzed the data using a generalized linear mixed model (GLMM) with time, treatment assignment, and the time × treatment interaction as fixed effects and participants as random effects to account for individual differences. The main effect of interest is the interaction between time and treatment in order to test for differences in the degree of change over time associated with the treatment condition. We chose either a Poisson GLMM or a linear GLMM depending on which model fit the data better based on the BIC.

To identify maintenance of, or changes in, treatment gains (i.e., if there are significant differences at the follow-up point), we used the same model previously employed to analyze the primary outcome point in order to maximize comparability of the results. In all follow-up assessments, the main outcome of interest (the time × treatment interaction) was reported. In cases lacking an interaction effect and interpretable main effects, the main effect of time was also reported (i.e., if participants changed overall from baseline to the measurement point).

In every analysis we controlled for age to account for the difference between the JASPER and PEI groups at baseline. Age was not a significant factor in any of the models tested.

Lastly, we reported the effect size using Cohen’s f² where effect sizes of 0.02, 0.15, and 0.35 are generally regarded as small, moderate, and large respectively (Cohen, 1988).

Results

Means, standard deviations, and effect sizes are presented in Table 2 for primary and secondary outcomes.

Table 2.

	JASPER N=43		PEI N=43		Effect sizes for group differences in change from baseline
	Proportion not on Scale	Mean (SD)	Proportion not on scale	Mean (SD)	f2 for proportions	f2 for means
Time Joint Engaged
Entry		147.49 (99.93)		161.02 (123.84)
Exit		325.17 (120.20)		159.72 (113.51)		.69
Follow-up		363.66 (132.50)		266.71 (119.90)		.26
IJA
Entry	23/42 (55%)	4.84 (4.62)	19/42 (45%)	4.61 (4.40)
Exit	17/41 (41%)	8.08 (7.23)	7/41 (17%)	7.00 (6.12)	.03	.00
Follow-up	9/38 (24%)	8.66 (8.58)	10/34 (29%)	6.79 (5.72)	.02	.03
Functional Play
Entry		2.00 (2.01)		2.24 (2.03)
Exit		4.05 (3.25)		2.85 (2.79)		.06
Follow-up		2.79 (2.17)		2.61 (2.36)		.01
Symbolic Play
Entry	38/42 (90%)	2.25 (2.50)	35/42 (83%)	2.00 (1.29)
Exit	31/41 (76%)	3.60 (1.95)	31/41 (76%)	1.80 (1.55)	.01	.01
Follow-up	26/38 (68%)	4.75 (2.83)	20/33 (61%)	3.08 (1.98)	.00	.00
Highest PlayLevel
Entry		5.77 (3.41)		7.20 (3.11)
Exit		7.94 (3.43)		7.00 (3.77)		.11
Follow-up		8.18 (3.92)		8.18 (4.05)		.03
Receptive Language
Entry		16.09 (9.12)		16.47 (9.58)
Exit		20.87 (11.85)		23.17 (13.02)		.02
Follow-up		32.74 (15.24)		33.38 (16.00)		.03
Expressive Language
Entry		14.09 (6.84)		14.98 (7.02)
Exit		18.42 (8.03)		19.83 (7.84)		.01
Follow-up		24.26 (9.34)		24.59 (8.82)		.01
PSI Child
Entry	30/43 (70%)	62.31 (24.80)	30/41 (73%)	65.00 (24.70)
Exit	22/39 (56%)	61.23 (25.62)	25/37 (68%)	47.08 (27.91)	.01	.05
Follow-up	17/37 (46%)	59.30 (24.33)	15/32 (47%)	62.41 (24.73)	.00	.00
PSI Parent
Entry	6/43 (14%)	47.35 (28.13)	9/41 (22%)	42.56 (24.16)
Exit	4/39 (10%)	47.74 (27.93)	5/37 (13%)	39.31 (30.68)	.00	00
Follow-up	7/37 (19%)	50.73 (27.60)	10/32 (31%)	35.86 (27.63)	.00	.01
Classroom Observation
Entry		5.77 (3.97)		6.81 (3.51)
Exit		8.14 (4.03)		7.33 (4.32)		0.06

Guidelines for f² (Cohen, 1988): small =.02, medium= .15, large=.35

Primary Outcome

Duration of Joint Engagement

The distribution of this variable was sufficiently close to normal to allow the use of a standard GLMM (AIC of the linear GLMM=1964, AIC of the Poisson GLMM=4599). There was a significant group by time interaction [187.42, 95% CI: (138.26, 236.58), F(1,83)=57.50, p<.01], showing that there was a significant increase in the length of time spent jointly engaged for the JASPER treatment compared to the PEI condition. Joint engagement more than doubled from entry to week 10 for the JASPER group, with a large effect size (Cohen’s f²=.69).

The increase in the length of time spent jointly engaged was maintained at the 6-month follow-up and significant for the JASPER group compared to the PEI group [63.79, 95% CI: (36.12, 91.45), F(1,83)=21.03, p<.01] with a moderate effect size (Cohen’s f²=.26). See Figure 2.

Figure 2.

Time in Joint Engagement (in seconds)

Secondary Outcomes

Initiating Joint Attention

This variable was zero inflated as there were participants whose initiations of joint attention fell below the measurement range of the measure [F(1,85)=30.37, p<.01]. Using the hurdle model, there was no difference in the rate participants who were below the measurement range entered onto the measurement range [−1.59, 95% CI (−4.55, 1.36), F(1,83)=2.86, p=.28], and there was no difference in the rate that participants who were in the measurement range improved their initiations of joint attention after treatment [0.098, 95% CI: (−0.268, 0.465), F(1,84)=0.28, p=.60]. There was an overall effect of time, in that the number of participants within the measurement range increased over time [3.80, 95% CI (0.40, 6.91), F(1,83)=5.94, p=.02], and the mean score of the participants within the measurement range also increased [0.54, 95% CI (0.29, 0.80), F(1,83=18.38), p<.01].

At follow-up, there was no difference in the rate participants who were below the measurement range entered onto the measurement range [0.76, 95% CI (−0.45, 1.98), F(1,83)=1.56, p=.22], and there was no difference in the rate that participants who were in the measurement range improved their initiations of joint attention after treatment [0.15, 95% CI (−0.04, 0.34), F(1,83)=2.52, p=.12]. The number of participants crossing the hurdle did not increase significantly from baseline to follow-up [0.73, 95% CI (−0.14, 1.61), F(1,83)=2.76, p=.10], but the mean score increased significantly for participants who crossed the hurdle [0.28, 95% CI (0.14, 0.42), F(1,83)=16.57, p<.01].

Number of Functional Play Types

The number of functional play types did not show zero inflation [F(1,84)=0.25, p=.61]; however, due to the variable’s skew, it was modeled best using a Poisson GLMM (AIC=694) over a linear GLMM (AIC=789). There was a significant group by treatment interaction indicating that the JASPER group increased more in types of functional play than the PEI group [0.45, 95% CI (0.06, 0.83), F(1,83)=5.35, p=.02] with a small effect size (Cohen’s f=.06).

At the follow-up time point, there was no significant interaction effect of group and treatment [0.12, 95% CI (−0.09, 0.33), F(1,83)=1.23, p=.27], and there was no increase from baseline [0.03, 95% CI (−0.12, 0.19), F(1,83)=0.16, p=.69].

Number of Symbolic Play Types

The number of symbolic play types revealed zero inflation [F(1,84)=4.48, p=.04]. Consequently, we modeled it using a hurdle model. The hurdle model did not reveal group differences in the change from below the hurdle to over the hurdle [0.14, 95% CI (−.84, 1.14), F(1,82)=0.57, p=.45], nor did it reveal group differences in the change of the observed level of symbolic play from baseline to the end of the study for participants who were within the measurement range [0.07, 95% CI (−0.56, 0.69), F(1,82)=0.84, p=.36]. In addition, we tested for overall changes over time as the interactions were not significant. There were no overall changes for the participants below the hurdle [0.97, 95% CI (−0.61, 2.57), F(1,82)=1.49, p=.22], and the mean score for participants above the mean did not change [−0.04, 95% CI (−1.03, 0.95), F(1,82)=0.01, p=.93].

At the follow-up time point, there was no difference between the groups in the number of children that moved from out of the measurement range into the measurement range [0.14, 95% CI (−.0.84, 1.13), F(1,82)=0.08, p=.77]; however, there was a significant increase over time of the proportion of children in the measurement range [0.84, 95% CI (0.10, 1.59), F(1,82)=5.10, p=.02]. Although there was no overall difference between groups in the increase of the scores for children who were within the range of the measurement [0.07, 95% CI (−0.56, 0.69), F(1,82)=.05, p=.82], some results revealed increased scores between baseline and follow-up [0.35, 95% CI (−.06, 0.76), F(1,82)=2.85, p=.09].

Highest Play Level Achieved

Highest play level achieved did not show significant zero inflation [F(1,84)=0.39, p=.53], but the skew of the data led to a best-fit model of a Poisson distribution (AIC=833) rather than a normal distribution (AIC=839). Using the Poisson model, there was a significant treatment by time interaction for the highest play level achieved [0.36, 95% CI (0.11, 0.61), F(1,84)=9.07, p<.01], such that the JASPER group increased more than the PEI group in highest play level achieved, although the effect size was small (Cohen’s f²=.11).

At follow-up, the difference between the treatment groups was no longer significant [0.10, 95% CI (−.02, 0.22), F(1,83)=2.91, p=.09], and there were no overall sustained treatment gains from baseline across the groups [0.07, 95%CI (−0.02, 0.16), F(1,83)=2.54, p=.11].

Reynell Receptive Language

This variable did not show zero inflation [F(1,84)=0.00, p=.98]. The skew of the data led to a best-fit model of a Poisson distribution (AIC=1137) rather than a normal distribution (AIC=1793). There was no significant treatment effect on the rate of change [−0.11, 95% CI (−0.31, 0.08), F(1,84)=1.35, p=.25], but there was an overall increase of receptive language over time across both groups [1.23, 95% CI (1.09, 1.38), F(1,84)=278.59, p<.01].

This same pattern was obtained at follow-up, with an overall significant increase from baseline [0.62, 95% CI (0.54, 0.69), F(1,84)=278.59, p<.01], but no differences were found between the treatment groups in the degree of change [−0.06, 95% CI (−0.16, 0.04), F(1,84)=1.35, p=.25]. Both groups increased in receptive language by nearly 17 months over the 9-month study.

Reynell Expressive Language

This variable did not show zero inflation [F(1,84)=0.00, p=.99]. A Poisson distribution (AIC=1047), rather than a normal distribution (AIC=1067), was used to address the skew of the data. There was no significant treatment effect on the rate of change [0.10, 95%CI (−0.13, 0.33), F(1,84)=0.74, p=.39], but there was an overall increase in rate of change over time across both groups [0.91, 95% CI (0.75, 1.08), F(1,84)=122.90, p<.01].

At follow-up, a similar pattern emerged with a significant increase from baseline in expressive language {0.46, 95% CI (0.37, 0.54), F(1,85)=122.90, p<.001] and no significant difference between the treatment groups in the degree of change [0.05, 95% CI (−0.06, 0.16), F(1,85)=0.74, p=.39]. Both groups increased an average of 10 months in expressive language over the 9-month study.

Parenting Stress Index (PSI), Child domain

As the stress variables were notably left-skewed with a large proportion of the respondents having extremely high stress scores (>90), we decided to invert this scale to use the zero-inflation framework that requires the data to be right-skewed. The inverted variable indicated significant zero inflation [F(1,84)=15.5, p<.01], a significant overrepresentation of extremely highly stressed respondents, which led to analyses using a zero-inflated Poisson model. The amount of change in the child-domain stress variable significantly differed across treatment groups [−0.76, 95%CI (−1.52, −0.00), F(1,82)=3.99, p=.049] in that parents in the PEI group experienced a larger reduction in child-related stress over time as compared to the parents in the JASPER group but with a small effect size (Cohen’s f²=.05). There was no difference in the ratio of participants on the scale over time [0.72, 95% CI (−1.22, 2.67), F(1,82)=0.46, p=.46].

Overall, more scores moved onto the measurement scale at follow-up, indicating decreased stress [0.70, 95% CI (0.05, 1.35), F(1,82)=4.65, p=.03], yet there was no group difference in this decrease [0.01, 95% CI (−0.82, 0.79), F(1,82)=.0, p=.97]. For respondents on the measurement scale, there was no significant change in stress over time [0.42, 95% CI (−0.64, 1.49), F(1,82)=.41, p=.52] and no difference between the treatment groups in this change [−0.03, 95% CI (−0.11, 0.05), F(1,82)=.34, p=.55].

Parental Stress Index (PSI), Parent domain

This variable also displayed zero-inflation [F(1,84)=12.9, p<.01]. There was no difference between the groups in the change over time [−0.11, 95%CI (−0.44, 0.22), F(1,82)=0.44, p=.51], and no overall change over time [0.13, 95% CI (−0.10, 0.36), F(1,83)=1.26, p=.26]. The proportion of parents on the scale also did not change overall [0.75, 95% CI (−0.85, 2.36), F(1,82)=0.87, p=.35], nor was it group dependent [0.00, 95% CI (−2.29, 2.29), F(1,82)=0.00, p=.99]. Similarly at follow-up, there was no difference between groups in change over time [−0.08, 95% CI (−0.27, 0.10), F(1,82=.85), p=.36], nor an overall change over time [0.04, 95% CI (−0.09, 0.17), F(1,82)=.35, p=.55]. There was also no difference between groups in the rate that people moved off the scale [0.40, 95% CI (−0.19, 0.62), F(1,82)=.34, p=.56], nor was there an overall trend of people moving off the scale [−0.70, 95% CI (−1.76, 0.35), F(1,82)=1.75, p=.19].

Classroom Observations

Results indicated a significant interaction effect for joint engagement [2.62, 95% CI (0.38, 4.86), F(1,84)=5.4, p=.02]. The children in the JASPER group improved significantly more than the PEI group in the amount of time spent jointly engaged with the teacher in the classroom environment. The time spent jointly engaged increased for the JASPER group from 5.7 minutes to 8.9 minutes within the 15-minute observation interval, while the PEI group only increased from 6.9 minutes to 7.4 minutes within the same observation period. This difference translated to a small effect size (Cohen’s f²=.06). See Figure 3.

Figure 3.

Classroom Joint Engagement (in minutes)

Discussion

The aim of this study was to determine whether parent-mediated interventions would significantly affect parenting stress and child behaviors above and beyond the early intervention services children were receiving. There were four main findings. First, consistent with our hypotheses, a hands-on, parent-training program (JASPER) resulted in significantly more treatment gains in dyadic joint engagement than a parent education program (PEI). Maintaining joint engagement requires parent strategies that capture the child’s interest and engagement with activities at the child’s developmental level, consistent with tenets of JASPER. Children also have to be active participants in treatment as it is not sufficient for parents to simply keep the child’s attention on the task at hand. In this study, the measure of joint engagement was coded from an independent assessment of the parent playing with child using a standard set of novel toys not previously used in intervention sessions. Parents coached in specific JASPER strategies were significantly more effective at engaging their children in play at post-treatment and follow-up than parents who received information about specific strategies through the PEI. Effect sizes were moderate to large.

Second, independent observations found that children in the JASPER condition engaged with their teachers more in their early intervention classroom. These findings may be among the first indicating generalization of joint engagement skills from a parent-mediated intervention to new partners and contexts. Teachers and classroom observers were blinded to the child’s treatment assignment and had not received any training regarding the contrasting treatment conditions. Other studies using adaptations of JASPER have found similar generalization of engagement. Kasari et al (2006) found increased joint engagement between mothers and children from therapist mediated interventions, and Kaale et al (2012) found generalization from preschool teacher mediated intervention to parent-child interactions. Longer durations of joint engagement over time facilitate greater opportunity to communicate with others and to learn additional skills from them. Thus, these findings lend greater support to joint engagement as an important treatment target for young children with ASD.

Third, mixed results were found for our secondary child outcomes in play and joint attention. Compared to children in the PEI condition, children in the JASPER condition demonstrated significantly greater improvements in functional-play diversity and overall play level at the end of treatment; however, these skills did not maintain at follow-up. The lack of maintenance may indicate the need for continued booster sessions for parents in order to maintain the strategies learned in the intervention. In regards to joint attention, children showed very few initiations of joint attention skills at the start of treatment, with more than half of all children showing no joint attention at all on independent assessments. Given this situation, we used a conservative analytic technique in order to model change in these skills across treatment and follow-up. Few children crossed the “hurdle” onto the measurement scale, and if they were on the scale, they did not show significant gains in joint attention skills over the course of treatment and follow-up. In contrast to findings with preschool-aged children with ASD, we did not find treatment effects on our measure of joint attention initiations, despite targeting initiations of joint attention (Kasari et al., 2006). Initiating joint attention is difficult for children with ASD and children may have needed more time to learn these skills than allotted in the present study. At the same time, we cannot rule out that another approach may have been more effective.

While children demonstrated mixed progress in joint attention and play skills, they did make significant developmental gains in language skills over the study with 17 months gain in receptive language and 10 months gain in expressive language over the 9-month study. These data provide further support for the disassociation between core deficits of children with ASD and general developmental gains. Most children with ASD appear to make significant developmental gains when provided with early intervention, but improvements in core deficits of social communication require targeted and specific interventions (Kasari et al., 2008).

Finally, results indicated reduction in parenting stress for families in the PEI condition. There is no question that raising a child with ASD increases parenting stress related to the disorder (Osborne, McHugh, Saunders, & Reed, 2008; Schieve et al., 2007). In parent-mediated models of intervention, parents must assume an additional role as therapist with their child causing increased stress for some parents (Osborne et al., 2008). In this study, nearly all parents reported very high levels of parenting stress, with over half of the parents above the ceiling of the measure at the beginning of the study. However, all children were simultaneously enrolled in an intensive early intervention (EI) program where children had access to a variety of professionals. Thus, stress related to trying to obtain services should have been alleviated. Results revealed that parents in the PEI condition, who consulted with an expert about their children and gained greater knowledge about autism, reduced their levels of stress as a result of the treatment. In contrast, parents in the JASPER condition, who provided direct intervention to their child, maintained their previously-elevated levels of parenting stress. There may be several explanations for these findings. One is that parents may have preferred a counseling approach over a hands-on approach because of the high dose of direct services their children were already receiving. Another possibility is that parents’ worries increase when they take on an interventionist role with their child and are directly faced with their child’s progress, or lack thereof. Additional research is warranted including the potential of combining JASPER and PEI, or the testing of therapist plus parent interventions for effects on the well being of parents. Future studies will also want to compare interventions that control for the amount of supervised parent-child play in order to isolate differences related to type of directed parent support.

The current study is both consistent and inconsistent with previous parent-mediated interventions with young children with ASD. While it is one of the larger parent-mediated interventions for children with autism, a few others are notable. Rogers et al. (2012) compared a parent-mediated version of the Early Start Denver Model to treatment-as-usual for 98 parents and toddlers (15 to 24 months old) with 12 hours of treatment over 3 months. No differences were observed on parent or child outcomes, but it warrants noting that the children in the Rogers study were 12 months younger on average than the children in the current study. However in a similar age range (16 to 20 months), Wetherby et al (2014) noted significant improvement in social communication and receptive language scores for children who received parent coaching versus parent education over nine months. Thus dose and length of intervention may be significant factors in increasing outcomes for children less than two years. However, neither study controlled experimenter contact between conditions or provided follow up data; thus differences due to dose or maintenance of gains are unknown.

Green et al. (2010) compared a parent-mediated intervention, PACT, to treatment-as-usual for 150 parents and preschoolers with autism who were on average 13 months older than the children in the current study. This one-year-long study, delivering about 18 hours of intervention, resulted in significantly greater parent responsiveness and child initiations of social communication for participants in the PACT group as compared to community controls. The data in the current study are consistent with the PACT trial for parent child outcomes, as well as with studies demonstrating greater gains from parent coaching models over parent education ones (Kasari et al, 2014; Wetherby et al, 2014). A notable strength of the current study is the comparison between two active, evidence-based, parent-mediated interventions. JASPER is an empirically-validated, targeted, and modular treatment for young children with ASD with significant treatment effects noted when tested against treatment-as-usual control groups using expert therapists (Goods et al., 2012; Kasari et al., 2006, 2008), teacher-delivered interventions (Kaale et al., 2012; Lawton & Kasari, 2012; Wong, 2013), and parent-mediated interventions (Kasari et al., 2010; Kasari et al, 2014). The current study highlights the effects of JASPER when compared to an active comparator, PEI, that has also been empirically -validated (Brereton & Tonge, 2006). Another notable strength of the current study is the ability to isolate the contribution of the experimental treatments against other treatments the participants were receiving. Most studies of parent-implemented interventions vary on dose of intervention between experimental and comparison conditions. While commonly tracking other services by parent report, it is difficult to control for reporting error and variability in types and dose of services (e.g., Rogers et al, 2012; Wetherby et al, 2014). In the current study, all children received the same early intervention program with equivalent dose and type of intervention.

While these data suggest that focusing on core deficits can result in important changes in parent and child outcomes, they also raise questions about individual differences in child and parent outcomes. Overall, clinical effect sizes were large for the primary outcome, and small to moderate on secondary outcomes. Sustainability (i.e., maintenance of treatment gains) was limited to the primary outcome measure of joint engagement. Future studies should strive for achieving a better understanding of the active ingredients of the treatment and profiles of children and parents most likely to benefit from specific early interventions.

Public Health Statement.

To improve outcomes associated with core impairments of toddlers with ASD, this study highlights the impact of direct, hands-on parent coaching of techniques to facilitate child social development. Parents also benefit from expert-delivered educational consultation, as shown by reduced parenting stress, but this treatment is less likely to improve child outcomes.