Confirmatory Factor Analysis of the Child Behavior Checklist—Preschool Version In a Sample of Children with Autism Spectrum Disorders

Background: The Child Behavior Checklist (CBCL; Achenbach & Rescorla, 2001) is an empirically-derived checklist completed by caregivers that measures behavior and emotional problems in children. In children with Autism Spectrum Disorders (ASDs), the CBCL has been used to identify problem behaviors and emotional problems, and measure autism symptomology (e.g., Hurtig et al., 2009). Additionally, the CBCL has been used to demonstrate the construct validity of other measures of autism symptoms (e.g., Bolte et al., 2008). Despite its widespread use in the ASD population, few studies have examined the validity of the CBCL in children with ASDs. Only one study has examined its factor structure in this population, yet suffered from methodological limitations (Pandolfi et al., 2009).

Objectives: The goal of this study was to evaluate the validity of the preschool version of the CBCL in children with ASD using confirmatory factor analysis (CFA). Separate analyses were performed on the factor-analytically-derived Syndrome Scales, and the DSM-Oriented Scales, which were derived originally using clinical judgment.

Methods: The current sample included 155 children with ASDs (n=98 Autism, n=52 PDD-NOS, n=5 Asperger Syndrome) between the ages of 18 and 73 months (M=43.8, SD=14.1). Polychoric correlation matrices and robust diagonally weighted least squares estimation procedures were employed using LISREL 8.8 (Jöreskog & Sörbom, 2007). Three indices of fit were of particular interest in evaluating the fit of the models: the Root Mean Square Error of Approximation (RMSEA), the Comparative Fit Index (CFI), and the Root Mean Square Residual (RMR).

Results: Indices of fit were mixed for the Syndrome Scales. The CFI was in the “good” range (0.98), the RMSEA was in the “acceptable” range (90% CI 0.047, 0.056; p = 0.287), and the RMR was outside of the “acceptable” range (0.11). The DSM-Oriented Scales yielded similar results: the RMSEA (90% CI 0.061, 0.072; p = 0.00) and CFI (0.96) were in the “acceptable” range, while the RMR was not (0.12). In general, factor loadings were moderate. On the Syndrome Scale, Somatic Complaints had the lowest mean loading (M = 0.55, SD = 0.17) and Aggressive Behavior had the highest (M = 0.73, SD = 0.08). Of the DSM Scale, Pervasive Developmental Problems had the lowest mean loading (M = 0.58, SD = 0.11), while Oppositional Defiant Problems had the highest (M = 0.79, SD = 0.05).

Conclusions: These results tentatively support the use of the CBCL Syndrome Scales in children with ASDs. Although the DSM-Oriented scales were not derived through factor analysis, the structure of these scales also performed reasonably well. Interestingly, the DSM scale tapping ASD symptoms had the lowest mean factor loadings. This may reflect less variability on this subscale in this sample than would be present in a typically-developing sample. The sample used in this study was quite small according to CFA standards. However, the largely “acceptable” indices of fit, despite the small sample size, suggest that the factor structure is robust. Future analyses with a larger sample are planned.