Development and Validation of a Short Version Eye‐Tracking Paradigm for the Screening and Diagnosis of Autism Spectrum Disorder in Qatar

ABSTRACT Objective behavioral assessments for autism spectrum disorder (ASD) are often time‐intensive and require substantial clinical expertise. Eye‐tracking–based paradigms offer quantifiable measures of social attention that can complement traditional tools. The current study builds on our previously validated Arabic‐language Autism Index (AI) by developing and validating a 4 min short version designed to improve feasibility in clinical and community settings while maintaining diagnostic accuracy. A total of 236 participants (127 with ASD, 109 non‐autistic controls including those with developmental delays (DD)) aged 1–16 years were assessed using an eye‐tracking paradigm consisting of 19 short dynamic videos depicting social and non‐social scenes. The AI was computed as the ratio of dwell time toward social versus non‐social stimuli. Diagnostic classification was established using ADOS‐2 and SCQ. Reliability and validity were assessed using Cronbach's α , Pearson's r , and ROC analyses, including age‐stratified performance and comparison with the original 10 min version. Feasibility was assessed by the proportion of valid stimuli. The short‐version AI demonstrated excellent internal consistency ( α = 0.91) and test–retest reliability ( r = 0.83). Diagnostic accuracy was high (AUC = 0.878, SE = 0.023), with age‐stratified AUCs ranging from 0.846 to 0.939. AI scores correlated strongly with ADOS‐2 severity ( r = 0.54, p < 0.001) and SCQ total scores ( r = 0.43, p < 0.001). Compared with the 10 min original version (AUC = 0.73), the short paradigm achieved higher accuracy and feasibility (valid stimuli: 89% vs. 80%). The current eye‐tracking paradigm demonstrates promising diagnostic performance while substantially reducing assessment time and cognitive demand. The findings provide initial evidence supporting its potential as a scalable and cross‐cultural tool for ASD screening and diagnosis, with further validation in independent and clinical cohorts supporting its translation into routine clinical practice.