Peer-Play Assessments for RCTs: Feasibility of Measuring Generalization from Clinic-Based Socialization Interventions

Background:  Previous research has found social skill group interventions to be an effective evidence-based practice for addressing common social deficits in children with ASD (Reichow & Volkmar, 2010). However, generalization is a concern, and the few studies using informant reports have found minimal evidence for generalization across settings and or people (Bellini et al., 2007). Skill maintenance across time has not been supported in studies evaluating retention using caregiver reports (White et al., 2010; Barry et al., 2003). In contrast, research using school-based social skill observational measures have found reliable, feasible strategies for measuring target social skills (Kasari et al., 2012). To our knowledge, these observation strategies have yet to be used to measure target behaviors in clinical settings.

Objectives:  The purpose of this study was to evaluate the feasibility of a clinic-based assessment to measure generalization from a randomized controlled trial (RCT) of social cognitive skills training groups (Seaver-NETT, Nonverbal Emotion recognition and Theory of mind Training) and facilitated play groups (Soorya et al., 2014).

Methods:  A sample of 38 verbal children with ASD ages 8-11 years old (31 male, 5 female) were enrolled during the last 3 cohorts of the larger RCT. 34 participants were included in baseline analyses and 12 in pre-post evaluations. Analogue peer play sessions conducted at baseline and endpoint included 1-2 participants with ASD and 2-3 unfamiliar typically developing peers. Sessions were approximately 30 minutes long, consisted of limited interaction with adult facilitators, and included a cooperative task (e.g. art collage) and a group game (e.g. Uno). Using ObserverXT 11, blinded raters coded the sessions for frequency of the following behaviors: instrumental bids, relational bids, nonverbal gestures and continuous interactions. Adequate to high interrater reliability was achieved (mean index of accordance = 81.4%, median = 92.21%). Data analyses evaluated the number of discrete behaviors (e.g. instrumental, relational, nonverbal) and the number of continuous interactions within each session.

Results:  Baseline data indicated differential numbers of target social behaviors in children with ASD. Chi-square analyses revealed a significant difference between the number of relational and instrumental behaviors within the sample, with significantly fewer instrumental bids observed relative to relational bids (X2= 23.72, p < 0.008). Analyses of target behaviors at baseline and endpoint indicated an increase in the number of continuous interactions ((X2= 9.00, p < 0.003). Correlational analyses revealed a positive relationship between the number of post-treatment continuous and instrumental behaviors (r = .82, p < .001). Baseline social cognition scores were also positively associated with the number of post-treatment instrumental behaviors (r = .57, p < .020). 

Conclusions:  Data suggest prolonged interactions with unfamiliar peers improved following outpatient social skills groups. These findings support the utility of observational assessment paradigms to measure generalization within clinic-based social skills RCTs. The positive relationship between baseline social cognition scores, endpoint instrumental behaviors, and continuous interactions warrants further study of the role of social cognition in predicting social skills generalization.