An Eye-Tracking Based Diagnostic Screener for Autism Spectrum Disorders in 18- to 42-Month-Old Children

Background: Mean age of diagnosis for an autism spectrum disorder (ASD) in the United States is currently later than five years of age.  This late age marks the loss of a potentially critical opportunity for improving treatment efficacy and associated outcome.  In addition, gold standard diagnostic evaluations usually require multiple tests proctored by an experienced clinician at specialized centers.  These are not often available in either rural or disadvantaged communities, and primary care providers are hesitant to recommend such tests without being certain of initial concerns.  In order to improve the efficacy of early screening, new tools yielding objective, performance-based measures of risk for autism would be highly advantageous.  Past research using eye-tracking has shown evidence of strong between-group differences when comparing children with ASD to their typically-developing (TD) peers, with large effect sizes.  In some cases, the eye-tracking measures have demonstrated predictive utility for measuring individual levels of social-communicative disability. In the current research, we tested the utility of eye-tracking as a screener for ASD in 18- to 42-month-old children.

Objectives: The objective of this research is to test the extent to which patterns of dynamic visual scanning in 18- to 42-month-old children, measured by eye-tracking, can serve as a screening tool, with sensitivity and specificity values above the accepted range for first-level screeners (>80% per Meisels, 1989).  We also tested the procedure’s ability to quantify levels of social and communicative disability in individual children.   

Methods: Eye-tracking data were collected from a large cohort of children, N = 170, 18 to 42 months of age, while they watched naturalistic video scenes of peer social interaction.  Standardized clinical assessment measures (ADOS, ADI, cognitive and language testing) confirmed diagnostic status for ASD and TD children within the cohort.  The first 50 consecutive referrals in the ASD group (ASD-1) were used for comparison against the TD group; these children provided a training set with which to develop a model of expected differences between ASD and TD visual scanning.  We then tested the remaining ASD children (ASD-2) as an external validation sample. Receiver operating characteristic (ROC) curves were created to analyze sensitivity and specificity.

Results: Preliminary results indicate robust between-group differences in visual scanning between TD and ASD-1 groups. Using this model, the remaining ASD-2 children were classified with sensitivity of 85% and specificity of 77.9%.

Conclusions: This research demonstrates a first step towards an eye-tracking based diagnostic screener intended for children between 18 and 42 months. Further research will test the extent to which such a screener can be optimized and practically deployed.