Factorial Structure of Autistic Traits in a Large Sample of Indian Children

Background: The Autism Quotient – child version (AQ-C) is s 50-item parent-report questionnaire that aims to quantify autistic traits in children between the ages of 4 to 11 years of age in the general population (Auyeung et al., 2009). The AQ-C comprises 5 theoretically derived domains: social skills, communication, attention to detail, attention switching and imagination. The factor structure of the AQ-C has been ascertained in Western countries and cultures, and the extent to which these factors may generalise to non-Western cultures remains unclear.

Objectives: To  derive the factor structure of AQ-C scores from a sample of 4-11 year old school-going children in Kolkata, India, and to compare this to the factor structure reported by Auyeung et al. 2009 (on a British sample). Additionally, to compare the factor structure of AQ-C as a function of respondent language within the Indian sample (English and Bengali).

Methods: Parent report AQ-C was collected from n=2116 4-11 year old children (mean age=6.4, SD= 1.32)  from schools  across all socio-economic statuses (government, private, and special schools) in Kolkata. The AQ-C was administered to parents, who filled in the questionnaire in English (n=1635) or Bengali (n= 481). For each respondent language, an initial exploratory factor analysis with principal component analysis and promax rotation was carried out.  Factors were identified from a scree plot, and fit to a model in confirmatory factor analysis using IBM SPSS AMOS 19.

Results: For the English version of the AQ-C a 4 factor model provided the best fit (GFI = 0.912, RMSEA= 0.048) and explained 27% of the total variance after rotation. The factors corresponded to social skills, attention to detail, mind reading and imagination, resembling the factor analytic structure reported in the original British sample.The Bengali version of the AQ-C, in contrast, was best fit by a 2-factor model with factors of social skills and attention to detail (GFI = 0.902, RMSEA =0.063), explaining 18% of the total variance. In both 3- and 4-factor models for the Bengali AQ-C,  an acceptable GFI (0.88) was obtained after removing items that loaded weakly on the factors.

Conclusions: This provides the first report of the factorial structure of autistic traits in children from a general population sample in India. We observe a similar factor structure for the AQ-C as reported in an Auyeung et al., 2009. Interestingly, the Bengali version of the AQ-C was better fitted by a 2 factor model (similar to the model reported in adults by Hoekstra et al., 2008, and replicated in Valla et al., 2010).  Extending this work with a larger and more socioeconomically diverse sample for the Bengali version can test if the observed differences in factor structure between languages is driven by the lower sample size for the Bengali version of the questionnaire, or by other socio-cultural parameters which may in this current sample be confounded with language.