A Meta-Analysis of Peer-Mediated Interventions for Young Children With Autism Spectrum Disorders

Education and Training in Autism and Developmental Disabilities

Zhang, J., & Wheeler, J. J. (2011).

Education and Training in Autism and Developmental Disabilities, 46(1), 62-77.

This meta-analysis investigates the effect of peer-mediated interventions (e.g., interventionist's age or gender) on social interaction skills in young children, 8 years and younger, on the autism spectrum.

Not stated

1977-2006

Single-subject design studies

45

Overall, effect sizes suggest that peer-mediated interventions were highly effective in increasing the frequency of social interactions for children under eight years of age&nbsp; (ES= 1.46), with good overall maintenance (ES= 1.49) and generalization (ES= 1.51). Improved duration of social interactions were noted immediately following treatment (ES= 1.27) and at follow-up (ES= 2.44). Subgroup analysis yielded the following findings:<br /> <ul> <li>Improved outcomes were seen in younger children (36 to 59 months; ES= 1.78) and in boys (ES= 1.53).</li> <li>Greater gains were seen if&nbsp; the mediators were boys (ES= 2.14), siblings (ES= 2.16), or were older (ES= 1.80) or at variable age levels (ES= 1.90).</li> <li>Researchers' combined involvement with peers/siblings, school staff, and parents/families (ES= 2.06) was associated with greater outcomes than with the involvement of any singular group.</li> <li>Improved outcomes were noted in home settings (ES= 2.23) when compared to class (integrated ES= 1.13, segregated ES= 1.86), clinic (ES= -1.56) or other settings (1.78).&nbsp;</li> <li>Interventions with a combined length of 421-480 minutes had the greatest improvements (ES= 4.02).&nbsp;</li> <li>Peer modeling (ES= 3.16) had the most robust impact on outcomes compared to other peer-mediated techniques.&nbsp;</li> <li>The greatest effects were noted on social response (ES= 3.23) versus social initiation (ES= 0.71) and combined social interactions (ES= 1.33).&nbsp;</li> <li>No significant difference in outcomes were noted between diagnoses or with varying intensity (numbers of minutes per week).</li> </ul> These findings should be interpreted with caution due to the potential for publication bias, the use of ruler measures for collecting data from graphs when data was not directly reported, the use of Cohen's d with single-subject design data, and limited sample sizes within each comparison. Further research is indicated.