Using the Childhood Autism Rating Scale (CARS) to Diagnose Autism Spectrum Disorders

Background: The Childhood Autism Rating Scale (CARS) is a widely used rating scale for the detection and diagnosis of autism. Although it is highly sensitive, the CARS may over-diagnose young intellectually disabled children as having autism (Lord, 1995).  Another limitation of the CARS is the lack of an empirically based Autism Spectrum Disorder (ASD) cutoff.  A cutoff of 30 is recommended for Autistic Disorder, but there is no generally accepted or validated cutoff for ASD. Although significant group differences on CARS total scores have been reported among clinical groups (Perry, Condillac, Freeman, Dunn-Geier, & Belair, 2005) the CARS is not designed to distinguish PDD-NOS from Autistic Disorder, or the ASD spectrum from non-spectrum (Perry et al., 2005). Identifying an ASD cutoff on the CARS that would include children with PDD-NOS may make the measure more useful.

Objectives: To investigate the ideal CARS cutoff for a diagnosis of Autistic Disorder and to investigate the utility of the CARS as a tool for ASD diagnoses in samples of 2-year-old and 4-year-old children referred for possible autism.
Methods: Participants were 376 two-year-old and 230 four-year-old children who were screened with the M-CHAT (Robins, et al., 2001), and evaluated at age 2  (mean age=26 months) and/or age 4 (mean age=54 months) using the ADI-R, ADOS, and the CARS, and were given diagnoses from clinical judgment based on DSM-IV criteria. Sensitivity, specificity, and positive and negative predictive values were calculated to determine optimal CARS cutoff scores for a diagnosis of Autistic Disorder as well as for an ASD diagnosis as compared to a gold standard diagnosis based on clinical judgment.
Results: Consistent with Lord (1995), the optimal cut-off score to distinguish Autistic Disorder from PDD-NOS in the two-year-old sample was 32, with a sensitivity (relative to Clinical Best Estimate diagnosis) of 0.79 and a specificity of .81. The optimal cut-off score to distinguish Autistic Disorder from PDD-NOS in the four-year-old sample was 30 with sensitivity of .86 and a specificity of .80. The optimal cut-off score to distinguish ASD from non-ASD in both samples was 25.5, with a sensitivity of .92 and a specificity of .89 in the two-year-old sample and a sensitivity of .82 and a specificity of .95 in the four-year-old sample.
Conclusions: Results suggest that, in a two-year-old sample, a cutoff of score of 32 more accurately distinguishes Autistic Disorder from PDD-NOS than the current cutoff of 30. Findings confirm the utility of the CARS in distinguishing Autistic Disorder from PDD-NOS, and distinguishing ASD from other developmental disorders and typical development and suggest that an ASD cutoff around 25, which is in common clinical use, is valid.