A Meta-Analysis of Innovative Technology Based Interventions for Autism Spectrum Disorders

Background: The field of innovative technology based interventions for Autism Spectrum Disorders (ASD) is rapidly growing with the last decade showing a steep increase in the number of published studies. However, the field is still perceived to be “emerging”, and the clinical validity of many reported interventions has yet to be acknowledged by the wider ASD research and clinical community. This is due, in part, to the diversity of papers on this topic in terms of their goals and methods as well as to the interdisciplinary nature of the field that involves engineering design, developmental psychology and educational sciences.

Objectives: The goal of the present report was to carry out a systematic evaluation of innovative technology based interventions for ASD which would summarize the state-of-the-art and provide recommendations for improvements in methodologies for future research.

Methods: We used a meta-analytical methodology based on strategies that are recommended in clinical psychology. First, a systematic literature search was conducted via four major online databases with keywords including “computer”, “virtual reality” and “robotics”. The retrieved articles were selected for the meta-analysis if they involved evaluation of a computer-based technology intervention, included a group of participants diagnosed with ASD and assessed training based on a pre-post or randomized controlled trial (RCT) design. Case studies and studies relying on in-system assessments were excluded. In order to assess efficacy, we computed the effect sizes between post-tests of groups receiving the intervention and those of control groups who did not receive the intervention. We also assessed the improvement following the intervention by calculating the within-group effect sizes between pre-tests and post-tests. We employed weighted effect sizes to account for sample size differences between studies. The homogeneity of the resulting effect size estimates was tested. The reliability of the results was evaluated by calculating the number of null studies needed to contradict the findings. We also conducted a moderator variable (age, IQ and duration of treatment) analysis to identify possible sources of heterogeneity between studies.

Results: The systematic search yielded 379 articles, out of which 21 were included in the meta-analysis. The test assessing efficacy in controlled studies was significant and yielded a mean effect size that approached the medium range (Cohen’s d=0.47, Confidence Interval= 0.08 – 0.86). The mean effect size of improvements following treatment was also significant and approached the large range (Cohen’s d=0.79, Confidence Interval=0.50 – 1.09). The effect sizes of controlled studies were found to be heterogeneous. Although the influence of age and IQ moderator variables was not significant, the effects sizes correlated negatively with the duration of treatment.

Conclusions: The present meta-analysis provided evidence in support of the efficacy of innovative technology based interventions, especially for interventions related to desktop computer applications. However, research in the field needs to address the issues that account for the heterogeneity between studies. According to our analyses, sources of heterogeneity could involve treatment duration. This meta-analysis highlights the necessity that future studies adopt more standardized research designs and assess maintenance and generalization of acquired skills.