Adaptive Intervention for Peer-Related Social Skills for Children with Autism Spectrum Disorders: Identifying Patterns Indicating Need for Change in Treatment

Background:  Social challenges are a significant concern for children with Autism Spectrum Disorder (ASD) across a wide range of abilities and ages.  These challenges may be most evident at school where mainstreamed children with ASD report significant difficulty in developing positive peer relationships.  For example, some children appear to be unaware of their peers on the yard, while others attempt to play and join in but their social initiations may be awkward or ineffective.  Still others are popular with their peers. This wide variation in the social characteristics of children with ASD suggests that there are likely different intervention needs.  Adapting interventions based on children’s response to intervention is a necessary next step that is currently limited in the autism research literature.

Objectives:  The purpose of this study was to explore methods for understanding the trajectories of children’s response to treatment prior to end of treatment in order to inform adaptive treatment models for future studies.

Methods:  Participants with ASD were drawn from a randomized controlled trial comparing two different social skills interventions at children’s schools.  We explored whether playground engagement scores measured at entry and midpoint of treatment predicted their engagement scores at exit using the Classification and Regression Tree (CART) method.  The CART method defines splits in the data that can then help professionals make data-based decisions about the individualization and adaptation of evidence-based social skills interventions.

Results:  Using the CART approach, four meaningful subgroups based on children’s playground engagement scores measured at entry and changes from entry to midpoint were identified using three splits.  All the splits were determined recursively by the CART algorithm.  The first split was based on how much the children’s percent time engaged changed from entry to midpoint by at least 14.01% from entry to midpoint.  Among those who did not increase at least 14.01% from entry to midpoint, a second split was conducted.  The second split was based on whether the children’s total percent time engaged with peers was greater than 51% at entry and this split separated theses children into two subgroups.  Lastly, among those who did increase from 14.01% from entry to midpoint, again a third split was applied.  For this group, the third split was based on whether the children’s total percent time engaged with peers was greater than 19.38% at entry and separated them into another two subgroups.

Conclusions:  This study illustrates the substantial heterogeneity in children’s response to treatment with multiple clinically salient subgroups embedded within the larger group.  The data suggest that measurements of children’s behavior mid-study can be used to predict children’s treatment outcomes.  Such data may be used to inform decisions to augment or alter programming prior to treatment exit in order to tailor intervention to best meet the needs of individual children and the CART method can be useful in defining metrics that could be used to build an adaptive treatment sequences for children.