Examining “Turning Points” in Trajectories of Symptom Severity in Children with Autism
Objectives: To extend our longitudinal investigation of symptom severity into the early school years and identify possible key “turning points” in the developmental trajectories of children with ASD.
Methods: Data were drawn from the Canadian Pathways in ASD study and included 360 children with ASD. Autistic symptom severity was indexed by the Autism Diagnostic Observation Schedule (ADOS) standardized severity score. Trajectories were plotted using data from four time points collected between the time of diagnosis and age 11. Three-way clustering methods consisting of variables and observations over time modelled as matrix variate data were used to identify homogeneous groups of children. Our matrix variate data comprised ADOS severity score and associated variables such as the child’s age at assessment, child’s sex, and adaptive functioning skills (indexed by the Vineland Adaptive Behavior Scales; VABS II). These variables were identified via boosted regression trees, a predictive modeling method. Missing data were imputed using median imputation combined with normal noise. A finite mixture of skewed matrix variate distributions (Gallaugher and McNicholas, 2017) was used for clustering.
Results: Using the Bayesian Information Criterion, a 3-group model was selected as the best fit to the data. Symptom severity scores for children in Group 1 (54% of the sample) and Group 3 (14% of the sample) decreased over time and then plateaued. However, symptom severity scores for children in Group 2 (32% of the sample) increased over time. A “turning point” (or “elbow”) at around 54 months of age – a time that for many children coincides with transitioning into the school system – was identified in two trajectory groups. Specifically, at that point, severity scores for children in Group 3 stopped decreasing and leveled out; severity scores for children in Group 2 stopped decreasing and started to increase. Groups differed by sex and had different linear associations with ADOS, VABS and age at diagnosis.
Conclusions: Study findings provide further evidence for the heterogeneous nature of symptom severity in children with ASD. The identification of key “turning points” in these trajectories supports the idea that symptom severity does not unfold in a linear way. Child, family and contextual factors associated with “turning points” in symptom severity will be discussed within the longitudinal research and clinical practice frameworks.