Psychometric properties of the Chinese version of the Autism Spectrum Quotient (AQ)

https://doi.org/10.1016/j.ridd.2012.08.005Get rights and content

Abstract

The Autism Spectrum Quotient (AQ) has been widely used for measuring autistic characteristics in parents of children with autism spectrum disorders (ASD). Nonetheless, its psychometric validity is yet to be justified. This study tested the factor structure of the AQ by means of principal component analysis and confirmatory factor analysis using, for the first time, data from 4192 Taiwanese parents (1208 with ASD children and 2984 with typically developing children). Results yielded a 35-item, 5-dimensional factor solution that had favorable psychometric characteristics (RMSEA = .054; NNFI = .962; CFI = .969) than any of the previously-published AQ factor solutions. Subscales of this new AQ-Chinese model were statistically and semantically coherent, namely: Socialness, Mindreading, Patterns, Attention to Details and Attention Switching. The psychometric properties of the AQ-Chinese did not change between clinic-based and community-based data suggesting good fitting for a continuum of autistic expression. Furthermore, the considerable overlap between the AQ-Chinese and the AQ factor structures derived previously using student samples indicated consistency in the manifestation of the autistic profile across different cultures and age groups. Group differences in the AQ-Chinese scores were in line with previous studies, i.e. males generally scored radically higher than females except in Attention to Details. Interestingly, mothers of ASD children reported lower total AQ scores than community mothers yet no significant group difference for the fathers. Important research and clinical implications pertinent to parents with children with ASD and the utility of the AQ were drawn.

Highlights

► This is the first factor analytic study of the AQ using an ethnic Chinese sample in Taiwan. ► Results yielded a 35-item, 5-dimensional factor solution we named as AQ-Chinese. ► Psychometric properties of the AQ-Chinese outperformed existing factorial models. ► The AQ-Chinese demonstrated good fitting for a continuum of autistic expression. ► Manifestation of the autistic profiles was consistent across different cultures.

Introduction

Autism spectrum disorders (ASD; encompassing Autistic Disorder, Asperger's Syndrome [AS] and Pervasive Developmental Disorder-Not Otherwise Specified [PDD-NOS]) are a group of neurodevelopmental disorders characterized by: (a) particular deficits in social reasoning skills, (b) marked verbal and/or nonverbal communication skills impairments and, (c) stereotyped patterns of behaviors or interests (American Psychiatric Association, 1994). Though not exclusively determined by heredity, there is compelling evidence that ASD is genetically related (e.g. Eapen, 2011, Folstein and Rutter, 1977, Folstein and Piven, 1991, Ronald and Hoekstra, 2011). Studies on relatives of people with ASD have reliably found normative variants of full-blown autistic profile, known as broader autism phenotype (BAP; e.g. Le Couteur et al., 1996, Pickles et al., 2000, Sucksmith et al., 2011). There is now a clear recognition of BAP in parents of children with ASD (e.g. Bernier et al., 2011, Bolte and Poustka, 2006, Dawson et al., 2007, Happe et al., 2001, Pisula, 2002, Piven et al., 1997), drawing research and clinical attention towards the effect of ASD symptomatology on these parents’ well-being (e.g. Ingersoll, Hopwood, Wainer, & Brent Donnellan, 2011).

Although several questionnaires are currently used to measure BAP, to our best knowledge, only limited psychometric data have been published. Among them, the Autism Spectrum Quotient (AQ; Baron-Cohen, Wheelwright, Skinner, Martin, & Clubley, 2001) has been most widely used for ascertaining BAP in parents of children with ASD. The AQ consists of five subscales: Social Skills, Communication, Attention Switching, Attention to Details and Imagination (Baron-Cohen et al., 2001). In an Australian study of 111 parents of children with ASD and 88 parents of typically developing (TD) children, parents of ASD children reported significantly higher autistic traits in the Social Skills and Communication subscales of the AQ (Bishop et al., 2004). These significant findings were supported by two recent studies in Sicily (Ruta, Mazzone, Mazzone, Wheelwright, & Baron-Cohen, 2012) and in U.K. (Wheelwright, Auyeung, Allison, & Baron-Cohen, 2010) but not supported by one conducted in Netherlands (Scheeren & Stauder, 2008). In view of inconsistent findings across different ethnic groups, whether parents of children with ASD in another ethnic group such as Chinese will exhibit higher autistic trait as measured by the AQ than those of TD children therefore, warrants further investigation.

In fact, the AQ has been administered efficiently to culturally diverse samples over the years demonstrating stable internal consistency on Cronbach's alpha test for population. For instance, cross-cultural comparisons by replicating Baron-Cohen et al. (2001)’s original U.K. study had been conducted in Japan (Wakabayashi et al., 2007a, Wakabayashi et al., 2007b, Wakabayashi et al., 2006a, Wakabayashi et al., 2006b, Wakabayashi et al., 2004), Austria (Voracek & Dressler, 2006), Italy (Ruta et al., 2012), Netherlands (Hoekstra, Bartels, Cath, & Boomsma, 2008), Scotland (Stewart & Austin, 2009) and Canada (Lepage, Lortie, Taschereau-Dumouchel, & Théoret, 2009). Each of these studies validated the usefulness of the AQ in quantifying autistic characteristics in their local general population. A dilemma at hand, nonetheless, is that whilst the AQ shows promise and potential, its use in research and clinical endeavors has outstripped the validation evidence for its psychometric coherence. Indeed, fully convincing evidence of the psychometric validity of the AQ has yet to be demonstrated whether specifically in the context of Taiwanese sample or other ethnic samples. This is particularly true of its factor structure.

Originally, five dimensions were postulated to be discrete subscales within the AQ on purely conceptual ground. Yet no factor analysis was reported to clarify whether these subscales reflected the best reduction of the item set, or whether a different structure exists. Furthermore, the test–retest reliability and internal consistency coefficients (Cronbach's alpha, α, ranging from .63 to .77) on the five domains in the original study (Baron-Cohen et al., 2001) were suboptimal. In response to these shortcomings, several factor analytical studies have been conducted over the past decade to verify the factor structure and reliabilities of the AQ (e.g. Austin, 2005, Hoekstra et al., 2008, Kloosterman et al., 2011). These studies revealed 2–5-factor models of the AQ, suggesting that the AQ is multifactorial and encompasses at least one factor relevant to social behaviors and another cognitive pattern. More specifically, two consistently verified constructs were ‘Social Skills/Sociability’ and ‘Pattern/Attention to Details’ (Austin, 2005, Hoekstra et al., 2008). Internal consistencies of the subscales in each model were varied (α ranging from .40 to .86).

All in all, several limitations regarding the structural validation of the AQ remain. First, none of the existing factorial models has achieved adequate goodness of fit indices as recommended for confirmatory factor analysis (CFA; see Browne and Cudeck, 1993, Hu and Bentler, 1999) suggesting the factor solution of the AQ requires further exploration. Second, internal consistencies of the subscales in each model were incongruent, with some as poor as α = .40. Third, none of the factorial studies included Asian samples, albeit its radical role in ASD research in this region. Given high AQ scores have been shown to be positively correlated with neuroticism whilst negatively associated with extraversion and conscientiousness (Austin, 2005, Wakabayashi et al., 2006a, Wakabayashi et al., 2006b), and that these personality traits are influenced by culture (Stevens, Kwan, & Graybill, 1993), we cannot disregard latent cultural effect by assuming the existing factor solutions for the AQ fit ethnic Chinese sample adequately. Fourth, the existing models were predominantly built on data from university students who are relatively young among the adult population and prone to higher academic caliber. Parents of children with ASD not only are comparatively older than university students, they probably would exhibit more autistic characteristics compared to the general population, as asserted by the theory of BAP. Collectively, their profile may present with unique ASD pertinent constructs as depicted by the items in the AQ hence require a further factorial analysis. In brief, the factorial validity of the AQ for Chinese parents warrants exploration especially if the AQ is increasingly relied on for genetic and endophenotype studies in ASD in ethnic Chinese population.

In light of the aforementioned gaps in the AQ literature, the present study aims to: 1) build on previous research by assessing the factor structure and internal consistency of existing factor solutions for the Chinese version (Traditional Mandarin) of the AQ using data from Taiwanese population; 2) develop and validate an alternative factor model (if indeed distinct from the existing models) that will fit data from a sample of parents of children with and without ASD. Such a group is likely to include individuals with BAP and to encompass the autism spectrum more entirely; and 3) test if the AQ shows sensitivity in discriminating parents of children with ASD from their counterpart, and to check if any particular domains of the AQ are more potent in portraying BAP.

Section snippets

Participants

The sample consisted of two groups: 1) clinic-based group: parents of 604 children with a clinical diagnosis of DSM-IV autistic disorder or Asperger's disorder (N = 1208; mean age of fathers, 43.05 ± 6.34 years; mean age of mothers, 40.13 ± 5.81 years), and 2) community-based group: parents of 1492 TD children (N = 2984; mean age of fathers, 43.59 ± 5.58 years; mean age of mothers, 40.53 ± 5.10 years). The clinic-based sample was recruited through the National Taiwan University Hospital in Taipei (n = 367,

Testing of Hypothesis 1: Goodness of Fit for existing factor models

Table 1 presents the CFA results of model fit for the five competing models. All of the models, as hypothesized, did not meet the recommended criteria for good fit, i.e. RMSEA < .06 (Hu & Bentler, 1995), NNFI > .90 (Bentler, 1990) or CFI > .95 (Browne & Cudeck, 1993). Internal consistencies as measured with Cronbach's alpha for each of the subscales in all the five models were mostly poor, ranging from .28 to .71.

Principal component analysis

Given that none of the existing models fitted the data fully convincingly, the sample

Discussions

The present study examined the factor-analytic structure of the AQ and tested its discriminant validity between parents of children with ASD and those with TD children. None of the existing factorial models fitted our Taiwanese data adequately. As a result, a 35-item, 5-dimensional factor solution that had favorable psychometric characteristics than any of the previously-published AQ factor solutions was developed. This new AQ-Chinese model consisted of five tightly statistically and

Conclusions

Our findings confirm that the 35-item AQ-Chinese has good test and retest reliability, comprises of items corresponding to the diagnostic criteria of ASD, can be therapeutically informative, is robust to cultural differences and last but importantly, is quick, easy and economical to use in clinical and community settings. There is, nonetheless, a need for improvement in its psychometric properties, especially in its factor structure and discriminative validity. The factorial analysis in the

Conflicts of interest

All the authors reported no biomedical financial interests or potential conflicts of interest related to this work.

Acknowledgements

This work is supported by grants from National Taiwan University Hospital (NTUH100-S1525) and National Science Council (NSC96-3112-B-002-033; NSC97-3112-B-002-009; NSC98-3112-B-002-004; NSC99-3112-B-002-036), Taiwan. In addition, the first author was supported by the Endeavour Prime Minister Australian Asia Award during the course of this project. We would like to express sincere gratitude to Miss Li-Ting Liu and Miss Chi-Mei Lee for their guidance in the LISREL analyses. We are indebted to the

References (61)

  • A. Wakabayashi et al.

    Are autistic traits an independent personality dimension? A study of the Autism-Spectrum Quotient (AQ) and the NEO-PI-R

    Personality and Individual Differences

    (2006)
  • H. Akaike

    Factor analysis and AIC

    Psychometrika

    (1987)
  • American Psychiatric Association

    Diagnostic and statistical manual of mental disorders

    (1994)
  • W.A. Arrindell et al.

    An empirical test of the utility of the observations-to-variables ratio in factor and components analysis

    Applied Psychological Measurement

    (1985)
  • T. Attwood

    The complete guide to Asperger's syndrome

    (2007)
  • S. Baron-Cohen et al.

    ‘Obsessions’ in children with autism or Asperger syndrome. Content analysis in terms of core domains of cognition

    British Journal of Psychiatry

    (1999)
  • S. Baron-Cohen et al.

    The autism-spectrum quotient (AQ): Evidence from Asperger syndrome/high-functioning autism, males and females, scientists and mathematicians

    Journal of Autism and Developmental Disorders

    (2001)
  • P.M. Bentler

    Comparative fit indexes in structural models

    Psychological Bulletin

    (1990)
  • R. Bernier et al.

    Evidence for broader autism phenotype characteristics in parents from multiple-incidence autism families

    Autism Research

    (2011)
  • D.V. Bishop et al.

    Using self-report to identify the broad phenotype in parents of children with autistic spectrum disorders: A study using the Autism-Spectrum Quotient

    Journal of Child Psychology and Psychiatry

    (2004)
  • S. Bolte et al.

    The broader cognitive phenotype of autism in parents: How specific is the tendency for local processing and executive dysfunction?

    Journal of Child Psychology and Psychiatry

    (2006)
  • M.W. Browne et al.

    Alternative ways of assessing model fit

  • B. Byrne

    Structural equation modeling with LISREL, PRELIS and SIMPLIS: Basic conccepts, applications and programming

    (1998)
  • R.B. Cattell

    The scree test for the number of factors

    Multivariate Behavioral Research

    (1966)
  • J. Cohen

    Statistical power analysis for the behavioral sciences

    (1988)
  • A.B. Costello et al.

    Best practices in exploratory factor analysis: Four recommendations for getting the most from your analysis

    Practical Assessment, Research and Evaluation

    (2005)
  • G. Dawson et al.

    Quantitative assessment of autism symptom-related traits in probands and parents: Broader Phenotype Autism Symptom Scale

    Journal of Autism and Developmental Disorders

    (2007)
  • V. Eapen

    Genetic basis of autism: Is there a way forward?

    Current Opinion in Psychiatry

    (2011)
  • S. Folstein et al.

    Infantile autism: A genetic study of 21 twin pairs

    Journal of Child Psychology and Psychiatry

    (1977)
  • S.E. Folstein et al.

    Etiology of autism: Genetic influences

    Pediatrics

    (1991)
  • Cited by (0)

    View full text