Childhood Autism Rating Scale
Childhood Autism Rating Scale (CARS) is a behavior rating scale intended to help diagnose autism. CARS was developed by Eric Schopler, Robert J. Reichier, and Barbara Rochen Renner. CARS was designed to help differentiate children with autism from those with other developmental delays, such as mental retardation.
Although there is no gold standard among rating scales in detecting autism, CARS is frequently used as part of the diagnostic process. Development of the CARS began in 1966 with the production of a scale that incorporated the criteria of Leo Kanner (1943) and Creak (1964), and characteristic symptoms of childhood autism.
CARS evaluation criteria
CARS is a diagnostic assessment method that rates children on a scale from one to four for various criteria, ranging from normal to severe, and yields a composite score ranging from non-autistic to mildly autistic, moderately autistic, or severely autistic. The scale is used to observe and subjectively rate fifteen items.
- relationship to people
- emotional response
- object use
- adaptation to change
- visual response
- listening response
- taste-smell-touch response and use
- fear and nervousness
- verbal communication
- non-verbal communication
- activity level
- level and consistency of intellectual response
- general impressions
Childhood Autism Rating Scale
This scale can be completed by a clinician or teacher or parent, based on subjective observations of the child's behavior. Each of the fifteen criteria listed above is rated with a score of:
- 1 normal for child’s age
- 2 mildly abnormal
- 3 moderately abnormal
- 4 severely abnormal
- Midpoint scores of 1.5, 2.5, and 3.5 are also used
Total CARS scores range from a fifteen to 60, with a minimum score of thirty serving as the cutoff for a diagnosis of autism on the mild end of the autism spectrum.
Initial psychometrics for the CARS were determined using 537 children enrolled in the Treatment and Education of Autistic and related Communication handicapped Children (TEACCH) program over a ten-year period (Schopler et al., 1980). Fifty-one percent of the children studied scored above the cutoff score of thirty.
Schopler et al. (1980) observed the existence of a bimodal distribution among these scores, leading them to develop criteria to differentiate between those with mild to moderate autism and those with severe autism. Children with a score exceeding 36 and who received a rating of three or greater on at least five subscales were categorized as being severely autistic.
Internal consistency of the CARS was high, with a coefficient alpha of .94 (Schopler et al., 1988), indicating the degree to which all of the fifteen scale criteria scores constitute a unitary phenomenon, rather than several individual behaviors. Inter-rater reliability was established using two raters for 280 cases. The average reliability of .71 indicated good overall agreement between raters.
Twelve-month test-retest data was also collected, with a finding that the means were not significantly different from the first testing to the second.
Criterion-related validity was determined by comparing CARS diagnoses to diagnoses made independently by child psychologists and psychiatrists. Diagnoses correlated at r = .80, which indicated that the CARS diagnosis was in agreement with clinical judgments.
CARS has also been shown to have 100% predictive accuracy when distinguishing between groups of autistic and intellectually disabled children, which was superior to the ABC and Diagnostic Checklist (Teal & Wiebe, 1986).
Validity of the CARS under different settings is of particular importance to the present study. CARS scores of 41 children taken through parent interview were compared to scores derived from direct observation. Mean scores under the two conditions were not significantly different and the correlation of r = .83 further indicated good agreement.
In addition, diagnoses based on parent interview and direct observation agreed in 90% of the cases. The authors suggest that valid CARS ratings and diagnoses can be achieved through parent interview (Schopler et al., 1988).
Of the autism rating scales discussed, the CARS is the only scale used in research with adolescents or adults. Mesibov, Schopler, Schaffer, and Michal (1989) examined the diagnostic ability of the CARS in adolescents and adults with autism. This study compared the CARS scores of 89 individuals before the age of ten (mean age of 8.7 years), with their scores after the age of thirteen (mean age of 15.9 years). Fifty-nine of the original 89 participants were diagnosed with autism before the age of ten, with a mean CARS score of 38.47. For those 59 participants, adolescent CARS scores revealed a significant decrease, with a mean score of 35.54.
Significant improvement in adolescent scores (i.e. significant decreases in abnormal behavior) were found on the imitation, body use, object use, adaptation to change, listening response, sensory response and use, verbal communication, nonverbal communication, and activity level scales. The only score with a significant increase over time was the general impression scale.
It was suggested that a cutoff score of 27, as opposed to 30, be used when administering the CARS to adolescents and adults. This recommendation stemmed from the observation that when the cutoff score was lowered by three points (corresponding to the mean difference in scores between the two age groups), the percentage of individuals accurately diagnosed as autistic before age ten and after age thirteen jumped from 81% to 92%.
Overall, Mesibov et al. (1989) suggested that the CARS is a good screening instrument for adolescents and adults. While Mesibov et al.’s (1989) study provided valuable insight into the course of the disorder over time; the generalization of the use of the CARS on adults based on its results is premature. The mean age of participants in the “adolescent and adult” age group was only 15.9 years. As a result, the study failed to demonstrate the ability of the CARS to diagnosis autism in adults (i.e. individuals over 21 years old). The CARS has been used with adults in clinical settings, but researchers have yet to verify its diagnostic ability with this population.
- Ozonoff, S, Boodlin-Jones, B, & Solomon, M. (2005). Evidence-based assessment of Autism Spectrum Disorder in children and adolescents. Journal of Clinical Child and Adolescent Psychology, 34, 523-540.
- Schopler E, Reichler RJ, DeVellis RF, Daly K (1980). "Toward objective classification of childhood autism: Childhood Autism Rating Scale (CARS)". J Autism Dev Disord 10 (1): 91–103. doi:10.1007/BF02408436. PMID 6927682.