Exploring the Nature of Quantitative Autistic Traits: A Factor Mixture Modeling Approach

Background:  Recent research suggests that the social and non-social aspects of autism spectrum conditions (ASC) may have distinct causes at genetic, cognitive and neural levels. This is reflected in the new conceptualisation of autism spectrum disorder outlined in DSM-5, comprising two dimensions of social/communication difficulties and restricted and repetitive behaviours and interests. It has been posited that these core features of ASC can be explained by a deficit in empathising alongside intact or superior systemising. A central debate in the development of DSM-5 across fields of mental health has been whether psychopathology is best conceptualised as a continuum of severity or as discrete categories of disorder. In recent years quantitative measures have been developed assessing autistic traits (Autism Spectrum Quotient, AQ), empathising (Empathy Quotient, EQ) and systemising (Systemising Quotient, SQ) on scales that follow a near continuous distribution. First-degree relatives tend to show intermediate levels of autistic traits, scoring in between clinical and general population groups. It is however unclear whether autistic traits, empathising and systemising are dimensional across study populations and, if so, whether meaningful subgroups can still be identified along these dimensions.

Objectives:  To assess whether the latent structure of empathy, systemising and autistic traits is dimensional and whether meaningful subgroups can be identified within this structure.

Methods:  Participants included individuals with an ASC (N=363), parents of a child with ASC (N=439) and general population controls (N=232). Participants completed measures of empathy (EQ), systemising (SQ) and autistic traits (AQ) using an online test platform. Factor mixture models were conducted to assess the most parsimonious number of factors and classes that can be identified. Mixture models allow for assessment of both the number of underlying dimensions as well as the identification of classes or groups of individuals. The factors model the severity of the trait, while the latent class variable allows for the classification of subgroups of individuals. 

Results:  Results highlighted a 2-factor three-class model with one empathising and one systemising factor and three latent classes of individuals. Class 1 was characterised by heightened systemising and low empathy, consisting mostly of individuals with autism and a small proportion of controls. Approximately a third of parents were also represented in the first class. Approximately half the control group and 40% of parents with equivalent empathy and systemising scores were represented in Class 2. Consisting mostly of controls, the third class displayed high empathising and lower systemising scores. 

Conclusions:   Results suggest three classes of individuals based on levels of empathising, systemising and autistic traits. Class 1 confirms that autism is characterised by impaired empathising alongside intact or superior systemising. The finding that a third of parents were best represented by this class highlights the existence of the broader autism phenotype. Moreover, findings support the notion that autistic traits, as well as empathising and systemising, are continuously distributed across all subpopulations, lending support to a dimensional conceptualisation of autism. This has implications for guiding future conceptualisations of diagnostic criteria as well as the development of assessment instruments and more tailored interventions.