The Validity of the Child Behavior Checklist in Identifying Anxiety Disorders in Children with ASD

Background:  While there is increased awareness of the impact of elevated anxiety rates in ASD, there are few well-validated screening measures for identifying DSM-based anxiety disorders in children or adolescents with ASD.  The Child Behavior Checklist (CBCL) is a parent-completed behavioral screening measure that has been used in several autism registry networks, including the Autism Speaks Autism Treatment Network (AS-ATN), the Center for Disease Control (CDC) Study to Explore Early Development (SEED), and the Simons Foundation Autism Research Initiative (SFARI).  It is also often used in clinical settings as a behavioral screening measure for children.  For these reasons, data on the validity of the CBCL in predicting true anxiety disorders in children with ASD could have a significant influence on clinical practice as well as ASD research initiatives.

Objectives:  The objective of this study was to examine the validity of the CBCL anxiety subscale in children and adolescents with ASD versus typically developing controls (TDCs).  Ratings on the CBCL were used to predict DSM-driven anxiety disorder diagnoses established via psychodiagnostic interviewing (the gold-standard Anxiety Disorder Interview Schedule [ADIS]).

Methods:  204 children were recruited through two studies following a two-by-two factorial design, including children with (ASD) and without ASD (TDC) and with (ANX) and without (NON) an anxiety disorder (ASDANX: N = 79; ASDNON: N = 44; TDCANX: N = 18; TDCNON: N = 30; mean age = 11 years).  ASD diagnosis was confirmed through research-reliable ADOS and parent interview.  Anxiety diagnoses were determined by the ADIS.  The ASDANX and TDCANX groups presented with a range of anxiety disorders, including social anxiety disorder, generalized anxiety disorder, separation anxiety disorder, and specific phobia.  None of the groups differed in general cognitive ability or age.  Subjects in the ASD groups did not differ in core symptoms of ASD (measured by the Social Responsiveness Scale [SRS]).  Sensitivity and specificity of CBCL in predicting anxiety disorder status was examined via multinomial logistic regression (MLR) and receiver operating characteristic (ROC) curves, carried out separately for the ASD and TDC groups.

Results:  CBCL Anxiety subscale t-score of 70 was highly predictive of anxiety disorder diagnosis among children with ASD (Chi Square = 61.97, p < .01) on MLR.  CBCL successfully classified ASD participants as ANX or NON with an overall accuracy of 79%.  Sensitivity (86% true positive rate) was superior to specificity (68% true negative rate).  ROC curve had AUC=0.783.  Although fairly robust, these predictive rates were lower than the predictive rates of anxiety disorder among TDCs for this study (94% and 93%, respectively).

Conclusions:  The CBCL Anxiety subscale has good convergent validity with gold standard diagnostic tools designed to measure common anxiety disorders in children and adolescents. The relatively higher rate of sensitivity versus specificity indicates that the CBCL may prove more valuable in identifying co-occurring anxiety disorder in children with ASD rather than in ruling anxiety out.  Further research is needed to determine if a different t-score or separate ASD-Anxiety subscale should be created to identify the unique features of Anxiety in children with ASD.