Abstract Objective The incidence of autism is increasingly encountered in children, and more and more autism rating scales have come into use. This study clinically compared three autism rating scales: Autism Behavior Checklist (ABC), Clancy Autism Behavior Scale (CABS) and Childhood Autism Rating Scale (CARS), in order to provide a basis for selecting assessment method. Methods Twenty-eight cases with autism (Autism group), who all met the diagnostic standard of Diagnostic and Statistical Manual of Mental Disorders Ⅳ(DSM-Ⅳ, USA), were evaluated by ABC, CABS and CARS. The best cut-off score, specificity, sensitivity, accuracy, positive prognostic value (PPV) and negative prognostic value (NPV) of each rating scale in the diagnosis of autism were calculated. Thirty-four patients without autism served as the Control group. Results The scores evaluated by ABC, CARS and CABS in the Autism group were all significantly different from the Control group (P<0.01). The result of the CARS test had the highest coincidence to criteria of DSM-Ⅳ (Kappa=1), followed by that of the ABC test (Kappa=0.87) and that of the CABS test (Kappa=0.60). According to the receiver operator characteristic curve (ROC), the best cut-off score of ABC was 31, and its specificity, sensitivity, accuracy, PPV and NPV were 0.97, 0.89, 0.94, 0.96 and 0.92 respectively. It was more effective for children over 3 years. The best cut-off point of CARS was 30, and its specificity, sensitivity, accuracy, PPV and NPV were all 1.0. Its assessment efficacy was not associated with age. The best cut-off score of CABS was 6, and its specificity, sensitivity, accuracy, PPV and NPV were 0.91, 0.82, 0.87, 0.88 and 0.86 respectively. It was more effective for children over 3 years.Conclusions There was a high coincidence among ABC, CARS and CABS. CARS seems to be the best autism rating scale, followed by ABC and CABS.
