Development of a Video-Based Instrument for ASD Screening in Infancy

Background:  The most widely used method for ASD screening is parent report but recent studies have demonstrated low agreement with more objective measurements of ASD symptoms. Major sources of error in parent report are comprehension and interpretation errors, such as limited understanding of the constructs or knowledge of developmental milestones. The use of videos has been shown to dramatically increase clarity in other fields, from music instruction to motor vehicle repair; in this study we examined whether video also improves ASD screening.

Objectives:  We developed the Video-Referenced Infant Rating System for Autism (VIRSA), a low-cost, low-burden, web-based ASD screening measure, and examined its psychometric properties and ability to predict autism outcomes.

Methods:  The VIRSA consists of a large library of 20-second video clips of parents interacting with infants, distributed along a continuum from high to low social competence. Parents are shown a pair of videos and asked to choose the one most like their child. Depending upon which video is chosen, the next pair of videos is selected to narrow the search space, and this process is repeated until the ratings converge on a stable score. The VIRSA was administered at 6, 9, 12, and 18 months to parents of 90 infants at familial risk for ASD and 45 infants with no family history of ASD. The infants are followed to 36 months of age to determine diagnostic outcomes. Parents complete the VIRSA from home, immediately before the visit and one week later. Examiners complete the VIRSA immediately after the visit. Finally, parents complete the Infant Toddler Checklist (ITC), a screening instrument that uses written descriptions of behavior, providing a measure of convergent validity.

Results:  Split-half reliability is good to excellent (r=.63 for parents, r=.86 for examiners). One-week test-retest reliability is good (r=.61), with parents selecting the same videos as the week earlier on 71.9% of paired comparisons. Inter-rater reliability is lower (r=.28) which was not unexpected, given the different contexts and expertise across parents and examiners. There is a significant main effect for the VIRSA as a predictor of the ITC, χ²=579.64, p<.001. This suggests acceptable convergent validity, with the VIRSA indexing many of the same behavioral constructs as the ITC. To examine predictive validity we compared VIRSA scores of participants who are typically developing (n=135) to those with an ASD diagnosis (n=16) or for whom developmental concerns have been raised (Atypical group, n=18). As seen in Figure 1, VIRSA scores of the ASD group worsen over time, while those of the Typical and Atypical group are stable or increase with age (main effect of group, χ²= 6.68, p<.05; group x time interaction, χ²=6.41, p<.05. The ASD group differs significantly from the Typical group by 9 months and from the Atypical group by 18 months.

Conclusions: 1) The VIRSA has good psychometric properties. Parents provide consistent ratings over time, demonstrate moderate agreement with expert examiner ratings, and detect changing behaviors as their infants develop ASD. 2) These findings replicate the pattern of declines in development as ASD emerges that have been documented using other measures.