Predictive Utility of CBCL Subscale Scores on Autism Diagnoses in Preschool Aged Children

Background:

The Child Behavior Checklist (CBCL; Achenbach & Rescorla, 2000) is a parent-report measure used to assess for a range of social, emotional, and behavioral disorders. The measure includes the Pervasive Developmental Problems (PDD) and Withdrawn Symptoms DSM-oriented scales, which have been suggested by some to be appropriate for the diagnosis of Autism Spectrum Disorders (ASD) in children (Pandolfi, Magyar, & Dill, 2009). However, others have indicated that the CBCL is not sensitive enough to identify ASD in young children (Kanne, Abbacchi, Constantino, 2009). However, to date, no research has examined whether caregivers’ ratings on the CBCL are predictive of a child’s final diagnosis of an ASD. 

Objectives:

The goal of the current study was to determine the clinical utility of the CBCL in diagnostic evaluations of preschool-aged children with possible ASD. More specifically, this study evaluated whether the PDD and Withdrawn scales were predictive of the final diagnosis given to the child.

Methods:

The present study was composed of 172 clinically-referred children between 25 and 71 months of age (M=47.1 months, SD= 12.52). All children were referred for psychological assessment to rule out an ASD. Evaluations were completed between November 1, 2010 and November 1, 2011 and all data were collected via retrospective record review. Clinical assessments typically included a parent interview, assessment of cognitive and adaptive abilities, ASD assessment using the Autism Diagnostic Observation Schedule, Revised, and completion of parent-report measures, including the CBCL.

Children were divided into two groups based on whether or not they received an ASD diagnosis, which included Autistic Disorder, Pervasive Developmental Disorder, NOS, and Asperger’s Disorder. Data were analyzed using multiple linear regression, with diagnosis as the criterion variable and CBCL subscales (i.e., scores on the PDD and Withdrawn subscales) as the predictor variables.

Results:

Of the 172 children included in the study, 62% were given an ASD diagnosis. The results indicated that the PDD scale scores (b = 0.09, t(170) = 1.17, p = 0.24) were not predictive of an ASD diagnosis. The PDD scale had a sensitivity of 0.65, but a specificity of only 0.42 in this sample. In contrast, scores on the Withdrawn subscale were predictive of a final ASD diagnosis (b = 0.18, t(170) = 2.41, p < .05). The sensitivity of the Withdrawn scale in this sample was 0.61, while the specificity was 0.55.

Conclusions:

Data from the current study suggests that the PDD subscale of the CBCL is not a valid predictor of ASD diagnoses in clinically referred preschool-aged children; however, the Withdrawn subscale shows some promise in differentiating children with an ASD from children without an ASD. Clinicians are cautioned against using this data outside of a comprehensive assessment to determine diagnoses.

Future research should examine the diagnostic utility of the other CBCL subscales, in addition to the possibility of gender and ethnic group differences. Examination of caregiver-teacher report forms as well as other more specific ASD parent report forms may also provide informative findings.