Validation of the Early Video-Guided Autism Screener (EVAS) in a Clinically-Referred Sample

Background: Although ASD can be detected as early as 14 months (e.g. Landa et al., 2007, 2013), the average age of diagnosis is four years (ADDM, 2014). The American Academy of Pediatrics (AAP) recommends universal screening at ages 18 and 24 months, though screening and referral practices are inconsistent due to cost, lack of staff training, and time needed for additional in-person evaluation. Access to a low-cost and efficient screening tool to promote further assessment can lead to earlier access to diagnosis and treatment. Existing ASD screeners have modest sensitivity and positive predictive value, and lack video examples of specific behaviors. Evidence is emerging that video guided examples of child development behaviors activate parents’ evaluative thinking about their child’s developmental well-being and may improve accuracy of parent report (Lievens et al., 2008; Marrus et al., 2015; Sices et al., 2008).

Objectives: The present study sought to validate the Early Video-guided Autism Screener (EVAS) in a clinically-referred sample. The EVAS is an innovative, accessible, and cost-efficient screener for children ages 18-48 months. The EVAS previously demonstrated strong sensitivity (SN), specificity (SP), positive predictive value (PPV), and negative predictive value (NPV) in a relatively small, but developmentally diverse sample (n=109). The present study will evaluate the previously derived EVAS cut-off score, and compare it with the gold-standard diagnostic tool the Autism Diagnostic Observation Schedule Second Edition (ADOS-2), and clinical best estimate (CBE) diagnoses.

Methods: A cut-off score of 53 was derived in a sample of 109 children with typical development (n=29), suspected developmental delays (n=15), suspected ASD (n=35), and diagnosed ASD (n=30) (Landa et al., 2014). The present sample included 170 children aged 18-48 months (mean age 2.75 years (.75 SD), 84.1% male, 55.9% Caucasian) clinically referred for ASD evaluation due to suspicion of ASD.  All were administered the EVAS online as part of their clinic intake. The ADOS was completed during diagnostic evaluations. Examiners rated CBE following the evaluation. Most (74.1%) received an ASD diagnosis; 7.7% had CBE of ‘at-risk for ASD’.

Results: The EVAS cut-off score for ASD risk of 53 was replicated resulting in moderate SN (84.1%; 95% CI: 78.6-89.6), SP (71.4%; 95% CI: 64.6-78.2), and strong PPV (98.6%; 95% CI: 96.8-100.4). Additionally, EVAS scores correlated moderately with the ADOS Toddler Module algorithm score (n = 54, r = .29, p < .05) and the Module 1 algorithm score (n = 84, r = .24, p < .05). The correlation between the EVAS and ADOS Module 2 algorithm score failed to reach significance (n = 32, r = .32, p = .07).   

Conclusions: Results provide strong evidence of the criterion validity of the EVAS suggesting it could be a valid, efficient, and cost-effective tool for ASD screening in specialty clinic settings. This tool provides a convenient and efficient method for caregivers to consider the degree to which their child exhibits signs of ASD and may improve caregivers' communication about their child's behavior during diagnostic evaluations.