Utility of three N-Item scales of the child behavior checklist 6–18 in autism diagnosis

Background: There is a need for well-validated questionnaire measures as an adjuvant to autism diagnosis. Three past research studies have each delineated a subsample of questions (5-, 9-, and 10-items) from the Child Behavior Checklist 6–18 (CBCL-6-18) that differentiate children with autism from those without (Achenbach & Rescorla, 2001; Duarte, Bordin, Oliveira, & Bird, 2003; Ooi, Rescorla, Ang, Woo, & Fung, 2011; So et al., 2012). Despite the potential of these subscales, no research aside from the initial investigations has been published, and the initial studies had methodological limitations. Method: This study investigated the criterion validity of the 5-, 9-, and 10-item autism subscales in a well-characterized sample of 483 community-referred children (X̅_age = 10.11, SD_age = 2.99, Autism n = 127; Not-Autism n = 356). Autism diagnosis was made using the Autism Diagnostic Observation Schedule-2 and children were diagnosed according to DSM-5 criteria. Results: Receiver Operating Characteristic analyses indicated Area Under the Curve in the poor range (0.633–.684 [95%CIs = 0.578–.734], ps < 0.001) for the three subscales. Sensitivity and specificity could not be optimized to acceptable levels with any cutoff value on any subscale. Conclusion: These n-item subscales are not suitable for identification of autism symptomatology. While the difference between current and past research may be related to sample differences, it is also likely that these differences are the result of methodological improvements. These results underscore the need for use of gold-standard measures, not parent report questionnaires, for autism diagnosis and additional well-designed research into the psychometric properties of autism questionnaires.