Resolution of the Factoral Structure of Quantitative Autistic Symptomatology in 11, 000 Assessments of School-Aged Children and Adults

Background: Understanding the factor structure of autism is critical to the discovery and interpretation of causal mechanisms in autistic syndromes. Newly-identified susceptibility factors involving single gene mutations, co-occurring variations in small groupings of risk genes, or the effects of a multitude of common variations—occurring in specific combinations and jointly influencing risk—are being elucidated every month.  In order to examine specific associations between behavioral variations and their underlying genetic and neural causes, it is important to continue to explore and resolve questions about how traits and symptoms in autistic syndromes co-vary, using data from large, diverse populations that encompass the full range of symptom structures underlying autism spectrum disorders (ASD).

Objectives: To evaluate the factor structure of quantitative autistic traits in a large, diverse sample of children and adults, representing the full range from typical (general population) variation to severe, clinical-level affectation.  Confirmatory factor analysis and assessment of measurement invariance across age, sex, informant, and ASD diagnosis (within autism-affected families) were examined in the largest sample ever assembled for this purpose.

Methods: Data were acquired using the Social Responsiveness Scale-2 (SRS-2) from three distinct samples: 1) a child clinical sample involving children affected by ASD and their unaffected siblings, participating in the Interactive Autism Network (IAN) volunteer registry (N=7,921) and reported-upon by their parents; 2) a child population-based sample (N=1,012) rated by a parent; and 3) an adult population-based sample (N=702) in which at least one report was obtained from a relative or other close acquaintance (n=1573), and in addition most subjects provided a self-report (n=673).  Confirmatory factor analysis and assessment of measurement invariance were implemented on the accumulated data set.

Results: A two-factor structure differentiating social-communicative impairment (SCI) and restricted repetitive behaviors (RRB)--as elaborated in the updated DSM-5 criteria for autism spectrum disorders--exhibited a highly acceptable model fit when confirmatory factor analysis was applied to the data. These factors exhibited measurement invariance across age, sex, and reporter (self vs. other), but exhibited a somewhat lower level of measurement equivalence between clinical and non-clinical populations. The statistical power afforded by this large sample allowed further factoral separation within each of the two principal factors, yielding three SCI sub factors (emotion recognition, social avoidance, and interpersonal relatedness) and two RRB sub factors (insistence on sameness and repetitive mannerisms). Cross-trait correlations between SCI and RRB remained extremely high, i.e. on the order of 0.95 for the general population, 0.94 for unaffected siblings in ASD-affected families, and 0.87 among children affected by ASD.

Conclusions: This study provided strong evidence of separable, but highly correlated, autism traits corresponding to DSM-5 domains. The statistical power afforded by quantitative analysis in this large sample allowed resolution of sub factors (themselves highly inter-correlated) which potentially represent subtle aspects of differentiation between deficits in emotion recognition, social avoidance, and interpersonal relatedness in autism and related neuropsychiatric syndromes. These components of behavioral dysfunction may constitute important targets for intervention, and for association with biological markers, particularly in gene discovery and in the exploration of neural signatures of autism.