Web-Based Tool to Assess Social Cognition in Youth with ASD: Reliability and Criterion Validity

Background:  Social cognitive deficits are a critical feature of autism spectrum disorder (ASD; Beauchamp & Anderson, 2010). However, no integrative, objective, psychometrically-sound tool to directly assess the multiple social-cognitive processes implicated in ASD is available. One such assessment, SELweb^TM, has demonstrated strong psychometric properties in large typically-developing (TD) samples (McKown et al., 2015). SELweb is an easy-to-use, web-based, modular assessment of self-control (SC), theory of mind (ToM), social problem solving (SPS), and emotion recognition (ER). If validated for youth with ASD, SELweb could complement outcome assessment batteries for intervention studies and provide a platform for a standardized social-cognitive assessment across settings.

Objectives:  To evaluate (1) internal consistency of SELweb modules, (2) criterion validity of SELweb compared to existing social-cognitive measures, (3) SELweb performance in ASD youth compared to normative data, and (4) feasibility of using SELweb in this population.

Methods:  Thirty-one youth with ASD (see Table 2 for demographics) completed the ADOS-2 (Lord et al., 2012), SELweb, and four validation tasks: Go/No-Go and Distractibility SC tasks (Zimmerman et al., 2005), an advanced ToM task (Happé, 1994), a SPS task (Kupersmidt et al., 2011), and a facial ER task (Tracy et al., 2009). Parents rated youth’s social skills/problem behaviors (SSIS; Gresham & Elliot, 2008) and ASD severity (SRS-2; Constantino & Gruber, 2012). Cronbach’s alpha was calculated to assess SELweb internal consistency. Pearson correlations examined associations of SELweb with demographic variables and scores on corresponding criterion measures and parent-reports. Independent-samples t-tests compared age-corrected SELweb scores to TD normative data. Qualitative feasibility data was collected during SELweb administration.

Results:  Internal consistencies across SELweb modules were acceptable to excellent (α=.57–.91). SC, ToM, and SPS demonstrated convergent validity with criterion tasks (Table 1; all p<.011). SC correlation with the criterion SPS task evidenced discriminant validity (p=.188), as did ToM and SPS with criterion measures of all other SEL constructs (all p>.252). ER correlated with responses to angry faces on the criterion ER task (r=.398, p=.033). Youth who were older or male scored lower on ER and ToM (Table 2). Problem behaviors were associated with lower ER scores. IQ was positively correlated with SC, ToM, and overall SELweb scores. Youth who scored higher on SC or were younger took longer to complete SELweb. Youth with ASD scored significantly lower than the TD normative sample for SC, SPS, ToM, and overall (all t>2.35, p<.019). With one exception, all children completed SELweb and did so independently. As evidenced via comments and behavioral observations during testing, children were engaged with the assessment format. Only five children required one short break.

Conclusions:  Promising psychometric properties and qualitative feasibility of SELweb support its utility as modular social-cognitive assessment for youth with ASD. SELweb performance corroborates well-established deficits among youth with ASD (SC, ToM, SPS) and converges with findings that ASD-specific differences in facial ER are inconsistent (Harms et al., 2010). Overall, SELweb demonstrates the capacity to acquire potentially treatment-sensitive, individual domain-specific profiles, in addition to overall, social cognitive functioning.