Individual Treatment Change during the Early Social Interaction (ESI) Project Predicts Long-Term Behavioral Outcomes

Background: Early intervention has been shown to improve language, cognitive, and/or social outcomes for toddlers diagnosed with ASD (Dawson et al.,2010; Kasari et al.,2010; Wetherby et al.,2014), as well as reductions in autism symptoms and adaptive behavior at school-age (Estes et al.,2015; Pickles et al.,2016). However, given substantial variability in children’s responses to early intervention (Guthrie et al.,2016), it is also important to examine individual treatment change to elucidate maintenance of improvements and impact on long-term outcome.

Objectives: To examine how individual-level change in social interaction skills during treatment predicted child functioning at follow-up. Specifically, this study investigated the relationship between individual trajectories of children who received 18-months of parent-implemented early intervention and outcomes at school-age.

Methods: Toddlers with ASD (N=82) participated in a crossover RCT design in which they were randomized to receive either Individual-ESI or Group-ESI for 9-months (Condition I, ~18-27-months of age) and then crossed over to receive the other treatment for an additional 9-months (Condition II, ~27-36-months of age). The CSBS was administered at baseline and every other month until the end of treatment. Sixty-five children returned for a long-term follow-up assessment between 5-8-years (M=6.79y), which included the ADOS-2, ADI-R, DAS-II (Differential Ability Scales-II), WJ (Woodcock-Johnson Achievement), Vineland-II, and BASC-3 (Behavior Assessment System for Children).

Results: At follow-up, children demonstrated average IQ, academic achievement, and behavioral functioning, with mild impairment in adaptive behavior and autism symptoms.

A piecewise hierarchical linear model was fit to model change during treatment; the final model included random intercepts, random slopes, fixed quadratic terms, and condition-by-slopes interactions. The resulting slopes were then used to predict school-aged outcomes via linear regression. Greater increases (i.e., steeper slope) in CSBS-Social composite scores during Condition I, but not Condition II, predicted lower ADOS-Social Affect scores at follow-up (t=-2.33, p.=.02). In contrast, greater increases in CSBS-Social during Condition II predicted higher ADOS-RRB scores at follow-up (t=3.11, p=.003). Finally, greater increases in CSBS-Social during Condition I, but not Condition II, predicted moderately fewer externalizing symptoms on the BASC at follow-up (t=-1.93, p=.06). Changes in CSBS-Social during treatment were not significant predictors of intellectual or academic functioning (DAS-II, WJ), adaptive behavior (Vineland-II), or parent-reported social interaction symptoms (ADI-R).

Conclusions: On average, children who participated in early ESI treatment as toddlers demonstrated age-appropriate intellectual and academic functioning at school-age follow-up. The degree of change in social interaction skills during treatment was a significant predictor of behavioral outcomes (i.e., autism symptoms, externalizing symptoms) approximately 3.5-years after treatment ended, but not of intellectual, academic, or adaptive functioning. While change during Condition I predicted ADOS and BASC scores, change during Condition II was generally not predictive of outcomes, likely due to the more limited gains observed in Condition II. The exception to this was the unexpected finding regarding the negative relationship between change in social skills during Condition II and school-age RRBs; we plan to examine this relationship further via parent-report (i.e., RBS-R). In sum, this study demonstrates the importance of examining individual changes during early treatment in predicting long-term outcomes.