Exploring the Behavioral Profiles of Preschoolers with ASD Using Cluster Analysis within the Context of An Intervention Efficacy Trial

Background:   As evidence emerges to establish the efficacy of interventions, concurrent explorations of variables that may explain variability in individual responses to intervention may provide important guidance to the iterative process of tailoring practices to enhance efficacy.

Objectives:   To determine if meaningful clusters would emerge based on the examination of nine variables that are theoretically linked to potentially moderating/mediating children’s responses to intervention.

Methods:   Of the 205 preschool aged children with ASD who participated in the efficacy trial, 163 were included in this study given the availability of data for all nine clustering variables based on first wave data collection at pretest.  A hierarchical cluster analysis based on Ward’s method and squared Euclidean distance was performed to explore a four cluster solution.  Nine variables were included in the model: (1) calibrated severity of ASD measured by the ADOS; 2) PLS-4 auditory comprehension and (3) PLS-4 expressive communication; two subscales of the RBS-R—(4) RBS-R ritualistic and sameness behavior (5) RBS-R restricted interest behavior, 6) Vineland II (V-II) interpersonal relationships, (7) V-II play and leisure time, (8) V-II problem behavior internalizing, and (9) V-II problem behavior externalizing.  Correlations and analysis of variance were also utilized to explore relations between variables and clusters.

Results: Four unique behavioral profiles emerged.  Between cluster differences were observed for ASD severity [F (3, 159) = 7.21; p <.001] with, in general, Cluster 1 (n= 30) and Cluster 2 (n= 57) including preschoolers with more severe ASD symptomology.  Auditory comprehension and expressive language skills also differentiated the clusters [F (3, 159) = 211.50; p <.001 and F (3,159) = 145.99; p <.001 respectively], with Cluster 3 (n= 43) and Cluster 4 (n= 33) including children with higher skills than Clusters 1 and 2. The four subscales of the Vineland offered significant contributions to differentiating the clusters (p values were all <.001). Cluster 1 exhibited stronger interpersonal skills and play and leisure skills than Cluster 2; however, Cluster 1 also exhibited more externalizing problem behavior.  Similarly, Clusters 3 and 4 were also differentiated by externalizing problem behavior, with Cluster 4 exhibiting higher rates than Cluster 3.  Interestingly, despite higher rates of problem behavior, Cluster 4 also exhibited the highest interpersonal skills of all clusters and play and leisure skills that were similar to Cluster 1.  Cluster 3 was differentiated from Cluster 4 by lower externalizing behavior problems as well as play and leisure skills. Ritualistic/sameness behavior and restricted interest behavior provided descriptively different patterns.

Conclusions: Severity of ASD symptomology, auditory comprehension, and expressive communication skills provide important insight when exploring child characteristics within the context of intervention outcomes.  Beyond those characteristics, interpersonal skills, play and leisure skills, and externalizing problem behavior appear to offer unique contributions to different behavioral profiles for this particular group of preschoolers with ASD.