The Importance of Visually Guided Screening: An Examination of an Autism Screener with and without Video Depiction of ASD Symptomology

Background: Autism Spectrum Disorder may be detected during the second year of life (e.g., Landa et al., 2007; 2014), though average age of diagnosis is after 4 years (ADDM, 2014). At present, screeners for ASD do not possess adequate sensitivity and specificity to detect subtle behavioral manifestations. Emerging research shows that video referencing improves accuracy of caregiver behavioral ratings and may improve caregiver-professional communication about behavior (Sices et al., 2008, Marrus et al., 2015).  

Objectives:   Preliminary findings demonstrated that a new web-based video-guided screener for detecting the risk of ASD in young children (the Early Video-Guided Autism Screener [EVAS]) has moderate predictive power when compared to the current diagnostic gold standard, the Autism Diagnostic Observation Schedule (ADOS) and clinical judgment. The present study aims to examine the added value of video examples on the EVAS.  

Methods: Parents of 103 children aged 12-60 months (mean age 24.57 months (11.51 SD), 52.4% male, 70.9% Caucasian) completed a paper-based question-only version (no videos) of the EVAS screener prior to their child receiving an evaluation that included the ADOS. Receiver operating curves (ROC) analysis utilized a diagnostic classification of ASD vs No-ASD generated through the ADOS and clinical judgment to determine this sample’s cut-off for classification and to examine the effect of removing video guidance (using just the EVAS questions) on performance of the screener by comparing results with existing data previously obtained using the full video-guided tool (EVAS).

Results: Using the questions-only version of the screener, findings paralleled some aspects of previously obtained results involving use of the questions with video-guidance. In both versions of the tool, ROC analyses suggested a cut-off of 53 as the optimal selection for classification (Figure 1). A comparison of the area under the curve (AUC) in the two datasets was not significantly different (p=0.23), though the video-guided sample performed slightly better. Further examination of the criterion validity of the questions-only version relative to the video-guided version suggested lower sensitivity (27.6%, 95% CI: 19.0-36.2 vs. 89.4%, 95% CI: 83.6-95.2) and a moderate, but slightly lower, positive predictive value (80.0%, 95% CI: 72.3-87.7 vs. 92.2%, 95% CI: 87.2-97.2). The specificity performed slightly higher at 97.3% (95% CI: 94.2-100.4 vs. 88.4%; 95% CI: 82.4-94.4). See Table 1.

Conclusions: These findings suggest that while the EVAS questions help informants to differentiate between having ASD-related concerns from typical development, video guidance appears to enhance parents’ ability to differentiate ASD-specific concerns from other general developmental concerns. This supports the budding literature highlighting the need for well validated, video-guided, screeners to enhance parent report of behavior and facilitate early detection of ASD.