Growth Mixture Modeling of the Repetitive Behavior Scale-Revised in Young Children with ASD

Background:  There are few data on how restricted and repetitive behaviors (RRB) emerge and change over time in young children with ASD. A handful of longitudinal studies provide some evidence that these patterns may differ based on type of behavior (e.g., higher order versus lower order) and age range.

Objectives:  The goal of this analysis was to evaluate the trajectories of subtypes of repetitive behavior, as measured by the Repetitive Behavior Scale-Revised (RBS-R), in a sample of children with ASD and low cognitive and adaptive function. We sought to determine whether the data were best represented by a mixture of trajectory classes, rather than a single, average trajectory, using growth mixture models.

Methods:  Data were drawn from a longitudinal natural history study of 106 children with ASD (DSM-IV autistic disorder) assessed at up to five approximately one-year intervals. The age range for the longitudinal data thus spanned 2 to 7 years. We first evaluated the standard growth models for each RBS-R subscale and the total score, followed by growth mixture models of increasing complexity. Where clustering around zero was common, models for censored data were used. Solutions with up to 4 classes were compared on a range of fit indices. Due to sample size constraints, joint trajectories were not modeled, but the joint distributions of most likely class membership across trajectories were calculated. Finally, correlates of class membership, such as level of cognitive impairment, were explored.

Results:  For all subscales, the mixture model was better fit to the data than the standard growth model, indicating that the growth model was not adequately accounting for heterogeneity in trajectory. For the purposes of space, we present in this abstract only the best-fitting solution to the Self-Injurious Behavior (SIB) subscale. This was a 4-class model (Figure 1). 41% of the sample was most likely assigned to the No SIB class and 49% to the Mild/Stable class. The remaining classes were small but clinically meaningful, and were therefore retained: Moderate/Valley (6%) and High/Peak (4%). Data on correlates, other subscales, and their joint distributions, will be provided.

Conclusions:  These analyses revealed considerable heterogeneity in the developmental trajectories of repetitive behavior in children with autism, as well as across types of repetitive behavior, providing more evidence for the notion that developmental trajectories may be required for ASD phenotyping.