Construct Validity of Eye-Tracking Indices of Active Engagement in School-Aged Children with ASD

Background: Autism Spectrum Disorder (ASD) is associated with a heterogeneous phenotype marked by social interaction difficulties as well as restricted and repetitive behaviors. Evidence-based treatments targeting core social difficulties in ASD are limited to behavioral or educational methods, and show wide variability in effectiveness and post-intervention outcomes. A major obstacle in the development of more effective, individualized, developmentally-appropriate ASD interventions is the absence of validated clinical outcome measures that objectively quantify treatment effects. From educational theory, the construct of ‘active engagement’ describes how children attend to and participate in learning opportunities. Active engagement is the means by which the content of educational interactions is accessed, and is one of the best predictors of successful learning for children with disabilities (Iovannone et al., 2003). Assays for rapidly and reliably quantifying active engagement are thus crucial for improving educational interventions for ASD. Eye-tracking-based measures of social visual engagement—the way children look at and attend to social information in their surrounding environment—are strong candidates for monitoring active engagement, pinpointing precise opportunities for intervention, and assessing response to treatment in ASD.

Objectives: This study aims to assess the validity of eye-tracking-based, quantitative assays of social visual engagement as measures of active engagement to be used in educational interventions aimed at addressing social difficulties in school-aged children with ASD.

Methods: Eye-tracking data were collected from a large and heterogeneous sample of children with ASD (mean age=10.08 years, n=159), as well as from an age-matched comparison sample of typically developing children (TD, mean age=9.50 years, n=43), during free viewing of naturalistic videos of social interaction. Data were quantified in terms of overall engagement (percent total fixation) and engagement with social information (percent fixation to people’s eyes, mouths, or bodies), and assessed for specificity and sensitivity to ASD severity, and test-retest reliability.

Results: Sensitivity and specificity to ASD social impairment severity were determined using regression analyses of ADOS social affect (SA) and restricted and repetitive behavior (RRB) scores (respectively), on eye-tracking indices of active engagement, controlling for age and cognitive level. Results indicate significant associations between ADOS SA and overall engagement (b=-8.82, p=0.003), and ADOS SA and social engagement (b=-12.47, p=0.002), as well as a significant association between ADOS RRB and overall engagement (b=-3.31, p=0.005), but no association between ADOS RRB and social engagement (p=0.152). Preliminary analyses of test-retest reliability for ASD and TD are promising: overall engagement, ICC>0.8, p<0.001; social engagement, ICC>0.6, p<0.001.

Conclusions: Eye-tracking measures of social visual engagement show promise as reliable, specific, and sensitive quantitative biomarkers of active engagement and social abilities in ASD. Higher overall engagement is associated with higher social affect and lower restricted and repetitive behavior, and higher engagement with social information is associated with higher social affect, but is not related to restricted and repetitive behaviors. Future analyses will explore convergent validity of these measures with clinician-administered assessments of child engagement, as well as comparative validation of eye-tracking measures of moment-by-moment perceived stimulus salience for objective, quantitative monitoring of active engagement, treatment response, and social development.