Reliability generalization for childhood autism rating scale

摘要

The Childhood Autism Rating Scale (CARS) is a popular behavior-observation instrument that was developed more than 34 years ago and has since been adopted in a wide variety of contexts for assessing the presence and severity of autism symptomatology in both children and adolescents. This investigation of the reliability of CARS scores involves meta-analysis and meta-regression of empirical data from reports of original research that made use of CARS between 1980 and 2012. Findings of good internal consistency (.896, 95 % CI.877-.913) and good interrater reliability (.796, 95 % CI.736-.844) support use of CARS at least in early-phase, exploratory research. Evidence of heterogeneity among literature data indicates that reliability is a property of CARS scores and is not intrinsic to the instrument itself. As the first of its kind pertaining to autism, this investigation provides guidance for reviews of other instruments' ratings.