Advancing Methodologies to Improve RRB Outcome Measures in Autism Research: Evaluation of the RBS-R

This study evaluates the psychometric properties (dimensionality, item bias, reliability) of the Repetitive Behavior Scale-Revised (RBS-R), provides scoring guidelines for the dimensional measure, and makes recommendations for future RRB measure development. Participants included individuals from three large autism data repositories; Simon Foundation Powering Autism Research for Knowledge (SPARK), Simons Simplex Collection (SSC), and National Database for Autism Research (NDAR). The total sample included N = 15,318 autistic individuals ages 3-18. Confirmatory factor analysis was used to evaluate competing theoretical factor structures. Item response theory (IRT) was used to evaluate differential item functioning, estimate the reliability of each RBS-R subdomain, and score the subdomains. A unidimensional factor structure demonstrated clearly inadequate model fit, calling into question the practice of reporting a total score on the RBS-R. A five-dimensional factor structure was supported by the theoretical and empirical evidence, though the fifth factor (restricted interests) was not sufficiently reliable for use. IRT-based scoring tools were generated for use in research. The present study illustrates the promise in the future development of measures for RRBs, particularly in the development of measures to separately and specifically assess RRB constructs using rigorous methodological guidelines. Public Significance Statement Accurate measurement of restricted and repetitive behaviors (RRBs) has proven difficult. The present study provides a recommended scoring approach that improves precision of scores on an established measure of RRBs, the Repetitive Behavior Scale-Revised (RBS-R), and provides recommendations to guide future RRB measure development.

Automated summary

This study evaluated the psychometric properties of the Repetitive Behavior Scale-Revised (RBS-R) in autistic individuals and recommended breaking down the total score into different dimensions for improved accuracy. A five-dimensional factor structure was proposed, with recommendations for future RRB measure development.