Spoken Communication Outcomes for Young Children with Autism: A Meta-Analysis and Meta-Regression

Background:   At least 30% of children with autism spectrum disorders (ASD) are persistently minimally verbal even after intensive early intervention for speech and language (Anderson et al., 2007; Tager-Flusberg & Kasari, 2013). Although, some meta-analyses have found positive effects of behavioral interventions on expressive communication (Reinchow & Wolery, 2008; Virués-Ortega, 2010), it is important to understand the extent to which interventions can meaningfully reduce the incidence of minimally verbal status in the ASD population.  Additionally, precursors to spoken communication have been identified as receptive language ability (Weismer, Lord, & Esler, 2010), verbal imitation skills (Paul, Campbell, Gilbert, & Tsiouri, 2013; Toth, Munson, Melzoff, & Dawson, 2006), and joint attention (Toth, Munson, Melzoff, & Dawson, 2006). Yet, it is unclear how these predictors interact with different treatment types, doses, and delivery methods. 

Objectives:   The following questions are addressed in this study: (a) are communication interventions effective for improving spoken language in children with ASD as compared to usual conditions? (b) Are augmented communication interventions effective for improving spoken language in children with ASD as compared to spoken-only interventions? (c) How does effectiveness vary by age, study quality, and type of intervention, and (d) how does effectiveness vary by receptive language, imitation skills, or joint attention?

Methods:   Comprehensive searches were conducted using 11 electronic databases through 2014.  A total of 1,624 references were reviewed after duplicates were removed. Studies that met the following criteria were included: (a) included a set of participants who were children, younger than 9, with ASD who speak English, (b) a comparison group of either a business as usual treatment or an additional treatment, (c) any behavioral intervention that did not include a pharmacology component, (d) an outcome including a pre-post measure of spoken language, (e) inclusion of a group design component, and (f) were published in English. Any setting for the intervention was included. 

Results:  Approximately 25 studies are included in the final analysis (final coding is on-going). A meta-analysis of the studies will be completed using a standardized mean difference effect size estimate to summarize the effects across studies. A standardized estimate will be used because of the heterogeneous measures that will be included. A mean difference score will be used to estimate the effects between groups. A Hedge’s g will be used because of the small sample sizes that occur in some of the included studies. A meta-regression will be used to determine the predictors of outcomes as outlined in question (d) above. 

Conclusions:   The discussion of the results of this meta-analysis will include recommendations for practice and future research regarding the spoken language intervention strategies and areas of need in the field for children with ASD. Although early intervention has made great progress for children with ASD, this study aims to specify the areas of need to better improve the spoken language outcomes for all children this population and to better inform customized intervention decisions for individual children.