Short (10-item) Versions of the Autism Spectrum Quotient (AQ) as ‘Red Flags' In Identifying Children, Adolescents, and Adults with Autism Spectrum Conditions

Background:  Diagnosis of autism spectrum conditions (ASC) is often delayed, especially when impairments are less severe. Concerns may be raised by parents at 18 months, yet some individuals reach adulthood before their difficulties are recognized as warranting a clinical diagnosis. Often, secondary psychiatric conditions such as anxiety and depression may have developed. According to the National Audit Office report in the UK (2009), 80% of primary care professionals (GPs) believed that they required additional guidance to identify individuals with suspected undiagnosed ASC. Our previous studies reported four parent- or self-report questionnaires aimed at quantifying autistic traits in toddlers, children, adolescents and adults. These are the Autism Spectrum Quotient (AQ), adult, adolescent and child versions (Baron-Cohen et al., 2001, Baron-Cohen et al., 2006, Auyeung et al., 2009), and the Quantitative Checklist for Autism in Toddlers (Q-CHAT, Allison et al., 2008). These questionnaires all show highly significant group differences in scores between individuals with a diagnosis of ASC and controls. They have been useful as phenotyping measures in research, but are too long (50 items) for quick use to identify ‘red flags’ in front line clinical or educational settings. Thus, there is a need to test if shorter (10 item) versions of these measures discriminate ASC from controls with high sensitivity, specificity, predictive validity, and reliability.

Objectives:  To produce short versions of the child, adolescent, and adult AQ and the Q-CHAT to aid professionals in primary health care, social care, and education settings in their decision making about whether to make a referral for a specialist assessment for ASC.

Methods:  Over 3000 control individuals (or their parents), and over 1000 cases (or their parents) completed one of the four questionnaires according to age. Self-report was accepted for individuals over 16 years old. Parent-report was required for those under age 16. Administration was via online or postal questionnaire. The proportion of participants who scored ASC positive on each item was compared on each questionnaire across cases and controls. The ten most discriminating items from each questionnaire were chosen. Receiver Operating Characteristic (ROC) curves were examined for the short versions.

Results:  Adults, adolescents, and children with ASC scored significantly higher (p < 0.001) than controls on each version of the 10 item AQ (AQ-10). The area under the ROC curve on all versions of the AQ-10 was > 0.96, representing excellent sensitivity and specificity. In the control groups, males scored significantly higher than females. Cut-points for referral for a multi-disciplinary assessment for ASC will be suggested for each tool. Q-CHAT data analysis is ongoing.

Conclusions:  This study demonstrates that short (10 item) versions of the AQ discriminate well between individuals with ASC and controls. Further, the 10-item versions of the AQ perform as well as the longer (50-item) versions. This study involved large samples and therefore results are likely to be robust. A limitation is that questionnaires were completed through online or postal testing. Evaluation of these ‘red flag’ guides will be conducted in primary care, social care and educational settings.