Literature Review: Child Autistic Rating Scale (CARS) Factor Analytic Results

DiLalla and Rogers (1994) used an exploratory factor analysis and applied factor- based scales to diagnose children in a day treatment program, and they found three factors: Social Impairment, Negative Emotionality, and Distorted Sensory Response which accounted for 69% of the variance in the sample’s Child Autism Rating Scale (CARS) scores.  By using an exploratory factor analysis, DiLalla’s and Roger’s objective was to understand how the variables were related and to determine the underlying factors’ structure. The children were previously diagnosed with autism and other developmental and psychiatric disorders, and the sample size consisted of 69 children between two and six years old. Each child had a primary teacher, and DiLalla and Rogers (1994) videotaped 20 minute long semi-structured play activities between the children and their assigned teachers. 

DiLalla and Rogers (1994) supported three factors in their hypothesis and found that the CARS allowed accurate clinical diagnosis. Specifically, DiLalla and Rogers observed that Social Impairment defined a significant difference between autistic and nonautistic participants. After a six month follow up, DiLalla and Rogers (1994) discovered that “The CARS items on this major factor reflect the social-interpersonal deficits associated with autism.” (p. 121). In addition, DiLalla and Rogers (1994) found in their follow up that Distorted Sensory Response was stable over 6 months of treatment, while their sample’s Social Impairment decreased over time; however, Negative Emotionality varied amongst the children.  DiLalla and Rogers (1994) concluded that the CARS was able to separate autistic children from those children who suffered other developmental or psychological issues (p. 125). DiLalla and Rogers (1994) suggested further research should view the CARS “as a multidimensional instrument” (p. 126) because that will allow researchers to focus on “behaviors and emotions associated with autism and other developmental psychopathologies” (p. 126). DiLalla and Rogers (1994) have suggested that Negative Emotionality should be studied based on low and high levels because it was the most sensitive to treatment, and may improve an autistic child’s ability to function socially.  In sum, DiLalla and Rogers (1994) have noted further research should measure by CARS’ subdomains in order to further its diagnostic skills and develop treatment, but a larger sample size was necessary for optimal results. A larger sample  DiLalla’s and Rogers’ (1994) video raters were unaware of the child and teacher’s relationships, so while they offered unbiased findings, the research study may have had inherent bias due the child teacher relationship. 

Stella, Mundy, and Tuchman (1999) performed a factor analysis to identify subgroups based on CARS that suggested an Autism diagnosis was multidimensional and individualist. Stella et al., (1999) agreed with DiLalla’s and Roger’s (1994) in that a larger sample size was necessary; however, DiLalla’s and Roger’s (1994) factor analysis did not account for the symptom clusters in autistic children. Therefore, Stella et al.’s (1999) sample consisted of archival records from 90 children diagnosed with autism or PDDNOS from the ages 2 to 17 over the years 1989 through 1995.Utilizing an oblique factor rotation, Stella et al. (1999) found five factors: Emotional Reactivity (29.4%), Social Communication, Social Orienting, Odd Sensory Exploration, and Cognitive & Behavioral Consistency (6.9%).  In addition, Stella et al. (1999) used a varimax rotation and it also produced five factors, but “the Social Communication factor (Factor 1) accounted for 29% of the variance…[and] Social Orienting at 9.5% (Factor 3) accounted for an additional 9.5%” (p. 311). The varimax rotation was ultimately used for Stella et al.’s analysis (1999) because it maximized variances (p. 311).  Based on Stella et al.’s (1999) results, the five-factor structure was uniform with the DSM-IV.   

Ultimately, the autistic group was found to have more disturbances on the “Cognitive and Behavioral Consistency factor and the Emotional Reactivity factor. However, [Stella et al. (1999) found that] the autistic and the PDDNOS groups did not differ on the Social Communication, the Social Orienting, and the Odd Sensory Exploration” (p. 313). Because DiLalla’s and Roger’s factor analysis (1994) identified social skills as the main differences, Stella et al.’s (1999) initial concern appeared to be correct. Essentially, because Stella et al. (1999) initially used varimax and oblique factor rotations within their study; they were able to measure un-correlation and correlation between emotional attitude and social skills. Significantly, the two social factors that were extracted accounted for 38.5% of the 64%. 

In an exploratory factor analysis, Magyar and Pandolfi (2007) hypothesized that a larger sample would “generate more stable factor analytic solutions” (p. 1789) and their sample consisted of 164 children, mostly toddler and preschool age, from Western New York hospital’s 1995 to 2002 database records.  Magyar and Pandolfi (2007) orthogonally and obliquely rotated the extracted components and found four factors: Social Communication, Social Interaction, Stereotypies and Sensory Abnormalities, and Emotional Regulation. Magyar and Pandolfi (2007) replicated DiLalla’s and Rogers’ (1994) and Stella et al.’s (1999) original procedures, but they also included “a second set of factor analysis procedures to help identify the CARS internal structure” (p. 1789). Former principal components analysis (PCA) determined that CARS was able to assess multiple constructs. Because PCA would extract the maximum variances with each component, Magyar and Pandolfi (2007) relied on principal axis factor analysis. They specified three and five factor solutions and performed several extractions based on Kaiser and scree test (Magyar and Pandolfi, 2007). In order to improve interpretability, Magyar and Pandolfi (2007) used varimax/quatimax and promax/direct oblimin rotations. 

In their results, Magyar and Pandolfi (2007) noted that the Principal Components Analysis was comparable, but it did not produce the same results as DiLalla and Rogers (1994) and Stella et al. (1999). However, Magyar and Pandolfi (2007) have noted that they based their research on archives, so their methodology was similar to Stella et al. (1999), and DiLalla’s and Roger’s (1984) structured play inclusions may have accounted for the differences. While using archives is a valid method in which to conduct research, the consensus is social skills are the primary deficiency in autistic children. Therefore, accurate findings may rely on a large sample size and a physical population. In sum, Magyar and Pandolfi (2007) recognized that levels of autism vary amongst individuals, and they agreed with the former researchers that Social Impairment, Negative Emotionality, and Distorted Sensory Response were the most pervasive in autistic children. 

Using a retrospective design, Mick (2005) attempted to determine whether CARS or Autism Diagnostic Observation Schedule (ADOS) was more accurate in its diagnosis. Mick (2005) included 320 children in her research and they were all under 6 years old. Participants were divided into two groups and each group identified children diagnosed with autism spectrum disorder or a disorder other than autism. Mick (2005) hypothesized that the CARS And ADOS were similar in their diagnoses. Mick’s (2005) “correlational analysis between the CARS and ADOS…result[ed] in a correlational matrix” (p. 28). Conbrach’s alpha established the internal consistency within the CARS And ADOS modules and factor analyses reduced the number of constructs in order to simplify the instruments’ data. 

Based on CARS, Mick (2005) found three factors: Meaningful Communication, Emotional Adaptability, and Sensory and Intellectual Response. Rotating the three factors for the ADOS factor analyses reflected Social Interaction, Social Language, and Joint Attention that accounted for 59.6% of the variance. In regards to the ADOS, Impaired Communication loaded at .830, the CARS loaded Meaningful Communication at .808. Because of the high scores of the five weighted factors, Mick (2005) speculated that the five factors, especially communication skills, were valid in an autism diagnosis. 

Mick (2005) based her study on a physical population, whereas Stella et al. (1999) Magyar and Pandolfi (2007) research studies utilized archives. In this way, Mick’s (2005) study added further evidence to the DiLalla’s and Rogers’ (1994) original findings. Most importantly, Mick’s (2005) findings were consistent with DiLalla’s and Roger’s (1994). For example, Mick’s (2005) first factor identified by Meaningful Communication is based on social and communication impairments and DiLalla’s and Roger’s (1994) first factor was Social Impairment. Secondly, Mick’s (2005) second factor Emotional Adaptability is based on an autistic child’s ability to adapt to changes in schedule, environment, or activity. DiLalla’s and Rogers’ (1994) second factor Negative Emotionality “taps anxiety and adaptation to change” (p. 126). Lastly, Mick’s (2005) third factor Sensory and Intellectual Responses is similar to DiLalla’s and Roger’s (1994) Distorted Sensory Response because it is a behavior that describes an autistic child’s peculiar behavior in response to stimuli. 

 It is worth noting Mick’s (2005) second level used a “factor analysis [that] was conducted on each instrument that resulted in weighted factor scores and a correlation matrix of factors” (p. 40). Because Mick (2005) used correlations between 15 CARS items, she demonstrated that autism is a multi-level disorder. However, her objective was to note any prevailing factors and to recommend clinicians use both the CARS and the ADOS in their diagnoses. Specifically, Mick (2005) hypothesized that the CARS and the ADOS in Module 1 would vary depending on the variable than Module 2. As the prior researchers in this literature review, Mick (2005) concluded the CARS was a viable and steady measure that identified social and communication problems as the most prevalent amongst autistic children.

References

DiLalla, D. L., & Rogers, S. J. (1994). Domains of the childhood autism rating scale: Relevance for diagnosis and treatment. Journal of Autism and Developmental Disorders, 24(2), 115-128. doi: 10.1007/BF02172092

Magyar, C. I., & Pandolfi, V. (2007). Factor structure evaluation of the childhood Autism rating scale. Journal of Autism and Developmental Disorders, 37(9), 1787-1794. doi: 10.1007/s10803-006-0313-9

Mick, K. A. (2005). Diagnosing Autism: Comparison of the childhood Autism rating scale (CARS) and the Autism diagnostic observation schedule (ADOS) (Doctoral dissertation, Wichita State University) (pp. 1-71). Wichita State University.

Stella, J., Mundy, P., & Tuchman, R. (1999). Social and nonsocial factors in the childhood autism rating scale. Journal of the Autism and Developmental Disorders, 29(4), 307-317.