Multimethod Longitudinal Analysis of Repetitive Behavior Measures in Children with ASD

Background:  There are no systematic investigations of the consistency of results among the various methods used to measure restricted and repetitive behavior (RRB) in ASD, such as the Autism Diagnostic Interview, Revised (ADI-R), the Autism Diagnostic Observation Schedule (ADOS), and the Repetitive Behavior Scale, Revised (RBS-R) (Leekam, Prior, & Uljarevic, 2011). Further, there have been few longitudinal studies examining the unique predictive utility of these popular instruments (Honey, Rodgers, & McConachie, 2012). 

Objectives:  In the current study, we explore ADI-R, ADOS, and RBS-R scores in children with ASD, aged 2-7 years at the initial evaluation. Our first goal was to explore forms of measurement (observation, interview, questionnaire) using cross-sectional data. Our second goal was to characterize the relationships among initial ADI-R, ADOS, and RBS-R scores and follow-up scores on the same measures, 2 to 5 years later.   

Methods: Participants were part of a natural history study of ASD (N=100; 81% male; mean age=4.5±1.6; mean nonverbal developmental quotient=58.9±18.6). Screening evaluation confirmed DSM-IV-TR Autistic Disorder, using the ADI-R, ADOS, and clinical judgment by expert clinicians. Data from screening and follow up were used in these analyses. Cross-sectional and longitudinal relationships between ADI-R (current algorithm), ADOS, and RBS-R scores in subjects with both timepoints (n=58) were first examined in a multimethod correlation matrix, and will be examined in a structural equation modeling framework to quantify the shared and unique measurement of RRB among methods (n=100) (Figure 1).  

Results:  The pattern of correlations suggested differential relationships between methods and across time (Table 1). Cross-sectionally, the ADOS was related to few other subscales, but many subscales of the RBS-R and ADI-R were moderately-to-strongly correlated. Longitudinally, the ADOS calibrated severity score and raw score were unrelated to themselves and to the other scales. Generally, the correlations observed cross-sectionally between the RBS-R and ADI-R persisted longitudinally. 

Conclusions: The RBS-R, ADI-R, and ADOS appear to measure different aspects of RRB, both cross-sectionally and longitudinally. Our results may be partially explained by method variance, as the scales with parental input were most strongly related. The longitudinal bifactor decomposition that we will present will shed light on this possibility. The content of the scales must also be considered, given that the early ADOS modules assess mostly lower-order RRB. Finally, it is possible that the behaviors did actually change over time; however, the relative stability of the RBS-R and ADI-R suggest that this is not sufficient explanation. In summary, the method used to quantify RRB appears to strongly influence characterization of symptom severity and change.