Measurement Invariance of the Child Behavior Checklist in a Large Sample of Children with Autism Spectrum Disorder with and without Intellectual Disability

Poster Presentation
Thursday, May 10, 2018: 5:30 PM-7:00 PM
Hall Grote Zaal (de Doelen ICC Rotterdam)
K. N. Medeiros, Health Psychology, University of Missouri, Columbia, MO
Background: Autism spectrum disorder (ASD) is characterized by core impairments in social communication and restricted and repetitive behaviors, with high rates of co-occurring emotional and behavioral problems. The Child Behavior Checklist (CBCL) is one of the most widely accepted rating scales used to assess childhood emotional and behavioral problems, and it has been used in many large-scale studies of children with ASD. Recently, conflicting research assessed whether the previously established factor model sufficiently accounts for symptom patterns in children with ASD. However, these samples have included a mix of children with and without concurrent intellectual disability (ID).

Objectives: The aim of this study was to determine whether the CBCL has the same validity in a sample of children with ASD who had ID compared to a sample of children with ASD who did not have ID. I aimed to decipher which elements of the factor models could confirm measurement invariance.

Methods: I used a multi-group confirmatory factor analysis of the CBCL subscales for young children (ages 1.5-5) and older children (ages 6-18) across those with (n=107) and without (n=301) ID. I used nested models to set progressively more constraints to test for various types of invariance across the two groups. The configural model had no constraints, and the groups were free to vary. The metric model constrained factor loadings to be equal across groups. The scalar model added that item intercepts were equal across groups. The residual model added that error variances were equal across groups. The structural model set factor variances equal across groups. Nested chi square tests assessed the null hypothesis that the more constrained model is correct under the assumption that the less constrained model is correct. I examined the seven CBCL subscales for young children: Emotionally Reactive, Anxious/Depressed, Somatic Complaints, Withdrawn, Sleep Problems, Attention Problems, and Aggressive Behavior. I examined the eight CBCL subscales for older children: Anxious/Depressed, Withdrawn/Depressed, Somatic Complaints, Social Problems, Thought Problems, Attention Problems, Rule-Breaking Behavior, and Aggressive Behavior.

Results: For young children, the configural model was the best fit for the emotional reactivity, anxiety/depression, and somatic complaints subscales. Sleep problems held metric invariance, and withdrawn behavior held residual invariance. The structural model was the best fit for both attention problems and aggression. For older children with ASD, the configural model was the best fit for the anxiety/depression, somatic complaints, social problems, attention problems, and rule-breaking behavior subscales. Whereas, withdrawn/depression, thought problems, and aggression held metric invariance.

Conclusions: The equality of factor loadings held for sleep problems in younger children, and for conduct/behavior regulation problems in older children (i.e., withdrawn, thought problems, aggression), suggesting that the CBCL measures the same latent traits in both groups, but with a bias in the manner in which trait levels are estimated. However, for young children with ASD, the CBCL measured attention and aggression problems equally across ID groups. Results of the current study suggest that cross-group comparisons of CBCL scores between children with ASD with and without ID may not always be appropriate.