Assessing Anxiety Symptomatology in Children with ASD Using Existing Measures: Is the Spence Children's Anxiety Scale- Parent Version a Useful Tool?

Background:  Assessing anxiety in ASD is challenging, because of core ASD features and anxiety symptom overlap and both traditional as well as varied manifestations of anxiety in ASD (Ozsivadjian et al., 2012; Kerns et al., 2014). Kerns and Kendall (2012) called for the validation of anxiety measures in ASD, but it is currently unclear whether ASD-specific measures are needed or existing ones are reliable and valid. There is to date no informant- or self-report anxiety checklist developed or validated specifically to assess anxiety in this population, but a growing number of studies in the last few years have begun to examine the use of existing measures. 

Objectives: To examine the internal consistency and convergent, divergent and discriminant validity and to explore the factor structure and clinical usefulness of the Spence Children Anxiety Scale-Parent Version (SCAS-P). This paper presents final analyses of a larger sample than previously reported in an earlier preliminary presentation (Magiati et al., 2015).

Methods:  Data from 12 studies from the UK, Singapore and the USA were pooled together. Participants were caregivers of 870 youth with a professional diagnosis of ASD, Autism, Asperger’s syndrome or PDD-NOS (87.7% males; 95% had a diagnosis of ASD; mean age=11.6 years, SD = 2.77) most of whom had at least one measure of autism symptomatology available. IQ data were available for 278 participants and another 238 had adaptive functioning scores. All but 21 clinically anxious help-seeking youth were recruited from special schools, other community or ASD diagnostic settings.

Results: Items from the total SCAS-P scale and from all but the Physical Injury subscales (α=.55) were internally reliable (all α>.75). The SCAS-P total and subscale scores had convergent validity with the Developmental Behavior Checklist (DBC) anxiety subscale (r=.32-.64; n=238). The clinically anxious subsample (n=21) had significantly higher total, generalized and social anxiety scores compared to the remaining participants. The existing SCAS-P six-factor structure was not a good fit for data obtained from 435 randomly selected participants in this sample and thus a principal components analysis (PCA) was carried out. Thirty of the 38 items loaded on five factors (social/ generalized anxiety, separation, obsessive compulsive, somatic symptoms and specific phobias) explaining 57.9% of the variance. A confirmatory factor analysis of the revised structure was re-run for the other half of the sample, but it was not a good fit.

Conclusions: The original SCAS-P full scale and subscales had excellent internal consistency and showed good evidence of convergent, divergent and discriminant validity, with the exception of the Physical injury subscale. Despite its promising psychometric properties, the lack of an adequate factor structure fit is concerning and suggests that future research should look into modifying and re-evaluating the SCAS-P for use in children with ASD.