Research has provided evidence of the effectiveness of video modeling (VM) interventions on social skills for individuals with autism spectrum disorders (ASD; Shukla-Mehta, Miller, & Callahan, 2010). Researchers have reached agreement on using meta-analysis in synthesizing single-case designs studies as it can result in more objective evaluation of multiple studies (Van den Noortgate & Onghena, 2003). Statistical procedures have been developed to provide more rigorous evaluations of the effectiveness of the intervention for single-case studies than traditional methods such as visual analysis.
Objectives:
The purpose of the study was to examine (1) the effectiveness of VM on social and communication skills of young children with ASD, (2) the relative effectiveness of VM in comparison to VM plus additional strategies, (3) the effectiveness of different models used in VM (self vs. others), and (4) the effects of potential moderators (child age, gender, or setting) on the effectiveness of VM on the outcomes. Four single-case research metrics were computed: the percentage of non-overlapping data (PND), the percentage of data points exceeding the median (PEM), the pairwise data overlap squared (PDO2), and the robust improvement rate difference (IRD).
Methods:
Inclusion criteria were that studies must (a) be published in English language, peer-reviewed journals between 1985 and 2011, (b) include at least one participant with ASD aged from 2-8 years, (c) utilize a single-case design that demonstrated experimental control, (d) have a baseline with at least three data points, (e) include a graphic display of child outcomes, (f) use VM only or VM with additional components, and (g) utilize outcome measures that targeted on social and communication skills as the primary dependent variables.
Results:
Twenty-six studies with 59 effect sizes were included for the meta-analysis. We adopted the criteria set by Scruggs and Mastropieri (1998) to categorize effects using the PND: a PND more than 0.90 is considered as very effective intervention, 0.70 to 0.89 as effective, and less than 0.70 as questionable or ineffective. We used these criteria to evaluate all the PND, PEM and PDO2 methods. We used the Park et al. (2009) criteria to categorize effects using the IRD: an IRD more than .50 is considered an effective intervention and less than .50 ineffective. The mean values of PND (.73), PEM (.82) and PDO2 (.85) obtained for 59 participants across studies all indicated that VM was considered an effective intervention for improving social communication skills of young children with ASD. The mean IRD values of .72 suggested a 72% improvement rate from baseline to intervention phrases. Child gender, age, and intervention settings, VM types (VM only vs. VM plus additional strategies), and model types (other vs. self) were not related to the outcomes of the study.
Conclusions:
All four metrics calculations indicate that VM intervention is effective to increase social and communication skills of children with ASD. However, findings should be interpreted with caution because of the limitations of using percentage of nonoverlapping data.