Testing the Psychometric Properties of the Spence Children’s Anxiety Scale (SCAS) and the Screen for Child Anxiety Related Emotional Disorders (SCARED) in Autism Spectrum Disorder.

Thursday, May 11, 2017: 5:30 PM-7:00 PM
Golden Gate Ballroom (Marriott Marquis Hotel)
S. Carruthers1, R. Kent2, M. J. Hollocks1 and E. A. Simonoff1, (1)Institute of Psychiatry, Psychology and Neuroscience, King's College London, London, United Kingdom, (2)Institute of Psychiatry, Psychology and Neuroscience, Kings College London, London, United Kingdom of Great Britain and Northern Ireland

There is a need to screen for anxiety in individuals with autism spectrum disorders (ASD) as it is one of the most prevalent co-occurring conditions (Simonoff et al., 2008) and individuals have reported these additional symptoms to be more impairing than their ASD symptoms. Questionnaires are most often used for screening for anxiety but have usually only been validated in typically developing populations. Study of the psychometric properties of such measures for use with ASD populations is limited.


To explore the psychometric properties of two commonly used questionnaires, the SCAS and the SCARED, with an ASD sample, in comparison to DSM-IV anxiety diagnoses according to the Child and Adolescent Psychiatric Assessment (CAPA).


The study included 49 males (aged 10-16 years) with a clinical diagnosis of ASD. The SCAS (Spence, 1998) and SCARED (Birmaher et al., 1999) were completed by parents and children. Trained researchers also conducted the parent-version of the CAPA (Angold and Costello, 2000), a semi-structured psychiatric interview schedule.


Rate of comorbidities

Using the DSM-IV CAPA algorithms, 63% of participants with ASD met criteria for one or more anxiety diagnosis: the most common were generalised anxiety (39%) and panic disorder (27%).

Internal consistency

The internal consistency for the total scores on both measures was high for both parent-report (both; α =0.95) and self-report (SCAS; α =.92, SCARED; α =.94). The internal consistency of the sub-scales was higher in the SCAS (parent α =.65–.84; self-report α =.69-.81) than the SCARED (parent α =.47–.75; self-report α =.46-.66) and in the parent report compared to the self-report.

Parent-child agreement

Parent and child scores were significantly correlated for both the SCAS (r =0.59, p <0.001) and the SCARED (r =0.46, p <0.001).

Comparison of the SCAS and the SCARED

The difference in the power of the two parent-report questionnaire measures in predicting an anxiety disorder was not significant when measured using ROC analyses (X2(1) =2.18, p =0.14) or between the two self-report measures (X2(1) =0.16, p =0.69).

Cut-points with parent data

In the ASD sample, using the recommended cut-off on the SCARED (25) gives 77.4% sensitivity and 50% specificity (67.4% correctly classified). ROC analyses in the ASD sample suggest an alternate cut-off of 28.5 to maximise both the sensitivity (74.2%) and specificity (72.2%), which would correctly classify 73.5% of the sample.

The suggested cut-off for the parent SCAS (24) gives 87.1% sensitivity and 66.7% specificity (79.6% correctly classified). In this ASD sample, ROC analyses suggest a higher cut-off of 27.5 maximises both the sensitivity (77.4%) and specificity (77.8%), which would correctly classify 77.6% of the sample.


There was no significant difference between the predictive power of the SCAS and the SCARED in an ASD sample; both questionnaires had good levels of sensitivity and specificity using cut-points validated in the typically developing population. Analyses indicate higher cut-points in ASD may be more accurate. This may reflect the symptom overlap in ASD and anxiety.