For a long time, children with autism were not considered worth investing in, which changed when studies began to emerge in the early 1980s which showed that exercise can significantly improve development. There is a consensus that it is important to have a lot of training and structured learning for younger children with autism, but there are divided opinions about which interventions give the best results.
In this blog, I will first report on a recently published meta-analysis of Sandbank et al (see reference below) that focuses on the issue in the title of the blog. A meta-analysis provides numerical values for how effective different efforts are.
So which initiatives are supported in the research according to Sandbank et al? And can we trust their conclusions? Let's start with the first question.
A meta-analysis of interventions for children with autism
In the article, the authors distinguish between seven different interventions for children with autism (0-8 years) and I briefly reproduce here their description of the seven interventions.
Behavioral interventions (1) are the interventions that were first used systematically with children with autism. This type of intervention is usually summarized as ABA (Applied Behavior Analysis) therapy and is the intervention that is most common in the American context. There are rigorous training programs developed at universities and colleges in how to work with behavioral interventions.
Development-related interventions (2) are partly in opposition to the more behavior-oriented interventions and are based on constructivist principles such as a view of the child as active. Vygotsky's theories have also influenced these methods, such as the idea that children develop in interaction with more knowledgeable people.
Naturalistic developmental behavioral interventions (3) are described as a mixture of (1) and (2) while TEACHH (4) is a program that contains carefully planned activities and structured environments, recurring routines and a high degree of visual support. Sensory-oriented interventions (5) are based on the idea that basic sensory functions need to be stimulated in order for children with autism to eventually develop more advanced functions.
Efforts built around relationships with animals (6) usually involve riding, which is assumed to lead to positive effects such as increased sensory stimulation and increased motivation. Efforts with technological aids (7), finally, involve technical aids such as computers, video, computer games and robots, which are expected to stimulate a positive development.
A "Public significance statement" reproduced in the article (p 2) summarizes the overall outcome of the meta-analysis:
“This comprehensive meta-analysis of interventions for young children with autism spectrum disorder (ASD) suggests that naturalistic developmental behavioral interventions and developmental intervention approaches have amassed enough quality evidence to be considered promising for supporting children with ASD in achieving a range of developmental outcomes. Behavioral intervention approaches also show evidence of effectiveness, but methodological rigor remains a pressing concern in this area of research. There is little evidence to support the effectiveness of TEACHH, sensory-based interventions, animal-assisted interventions, and interventions mediated solely through technology at this time.”
The meta-analysis puts the development-related interventions in a positive light, but we can be absolutely convinced that the last word has not been said in this discussion (see below). It is also the case that development-related interventions so far have mainly been shown to have a positive impact on communicative skills, while naturalistic developmental behavioral interventions and behavioral interventions have been investigated for and shown to have a positive effect on a wider range of skills.
Gathering evidence - the importance of choices made
It is easy to perceive evidence as an almost thing-like entity that that can be extracted from research. Many people want answers to questions such as "Is there evidence for the intervention / method X" but unfortunately it is seldom possible to give a simple answer. There are different ways of making research reviews and the choices we make have consequences for the conclusions we will end up with.
One choice that Sandbank and others make, which has enormous consequences for the conclusions drawn, is that they completely opt out of research with so-called single-case designs, that is studies with only a few participants. They believe that such studies are not strict enough to be used in a meta-analysis. Furthermore, it should also be pointed out that studies with qualitative approaches have also been excluded.
Other choices they have made that also have consequences for their conclusions concern how studies are searched for in databases, the weight of different studies, the significance given to what the control group has done, what significance should be given how long an effort lasts and so on.
An additional choice with absolutely decisive consequences leaves Sandbank and others partly to the reader. That choice is about whether the conclusions of the meta-analysis should be based on all included studies (note that studies with single-case design have already been excluded) or only on those that live up to high methodological requirements.
When studies with a lower degree of methodological strictness are included, the behavioral interventions turn out to be significantly better than when only those who live up to high methodological requirements (randomized group designs where participants are randomly distributed among the intervention and control groups) are included. If, on the other hand, we set the methodological requirements for methodolgy very high, all studies will lapse!
Compiling evidence - the importance of the question being asked
What evidence is found has further to a large extent also to do with how the question about evidence is posed. In the analysis referred to here, the question is which interventions are effective for children with autism. This means that the special nature in the form of the medical diagnosis has been taken as a starting point for the overview.
We can also think of other ways to ask the question: What interventions are effective for children with autism who go to a regular group of children? What working methods are generally fruitful for developing communication and social interaction for all children? Which working methods are fruitful for the children who have problems with communication and social interaction? We can also more closely, so to speak, nail down the specialty: What interventions are effective for children with autism who also have an intellectual disability? Each question requires different types of evidence to be answered.
It is important to note that some research reviews only include research conducted in integrated environments (see examples of such reviews below).
Other choices, other conclusions - an example
At the time of writing, I get hold of a fresh research overview on evidence for interventions in autism by Steinbrenner and others (see link below) who use a completely different methodology than Sandbank and others. They also cover a longer development period (0-22 years). I argued above that the choices made when conducting research reviews are of great importance.
Since Steinbrenner and others make different choices than Sandhurst and others, it is illustrative to compare the consequences this has. The differences I want to highlight here are that Steinbrenner and others choose a) to approve of studies with single-case design b) not to calculate effect sizes and c) to identify and name interventions in a different way.
The choice to include studies with single-case studies has the consequence that the number of studies increases enormously, as this is historically the most common study design in the field. It is also used, as I understand it, almost exclusively by researchers with behavioral analytic starting points, and many of these studies also provide support for such initiatives.
The choice not to analyze effect sizes makes it difficult to compare the effect of the interventions, so to speak. To simplify a little bit, we can say that Sandbank and others analyze, just like John Hattie, who jumps the highest, while Steinbrenner and others examine who gets above a certain height.
Steinbrenner and others thus find a very large number of interventions that come over this bar (28) or which they believe have the potential to do so in the future (12). As this research expands rapidly, we can expect even more efforts that, according to this decision logic, have evidence.
The choice of how interventions are defined and named is also of great importance. The number of interventions identified and the terms used are different from those in Sandbank's and others' analysis.
Steinbrenner and others distinguish interventions involving behavioral analysis techniques (to the number clearly dominant involving e.g. modeling, singe-trial learning, behavioral momentum intervention) over more psychologically oriented interventions (for example social skills training, self-management) to more pedagogically oriented working methods (music-mediated interventions, direct teaching). Translating these interventions into the seven interventions that Sandbank and others identify is certainly not an easy task.
It is also worth noting that many of the interventions mentioned in the two reviews probably work for all children. However, the interventions are discussed more or less decoupled from this question and from the question of how the pedagogical environment in the preschool as a whole is shaped. Of course, the lack of mapping and description of the environment also makes it more difficult to see how the mainly American research becomes relevant for e.g. Swedish preschool environments.
A final word
My ambition in this blog has been quite modest, namely to present a research overview and at the same time illustrate and discuss choices made when evidence is compiled and the consequences such choices can have.
From an inclusion perspective, it is of course important that the activities in the preschool maintain a high quality with a learning environment that is adapted based on the fact that children have different prerequisites for participating. It is therefore very problematic if the discussion about interventions for children with autism in preschool is conducted alongside discussions about how the environment in general should be designed and the interventions that benefit all children.
No matter how well the general environment is designed, however, we can expect that some children still need targeted support, which must also sometimes be given outside what happens in the regular child / student group. Many children with autism pose great challenges for educators in preschool and it is of course good to know the research that tries to map interventions that help children reach the goals of preschool education. Ultimately, it is always important to map the situation of the individual child. It is of course important to map both strengths and difficulties and to try to understand the child's behavior based on the complex context in which the child is involved.
Sandbank M, Bottema-Beutel K, Crowley S, Cassidy M, Dunham K, Feldman JI, Crank J, Albarran SA, Raj S, Mahbub P och, Woynaroski, TG. (2020) Project AIM: Autism intervention meta-analysis for studies of young children. Psychological Bulletin, 146 (1), 1-29.
Examples of overview articles where integration / inclusion is taken as a starting point for the overview
Gunning, C., Breatnach, O., Holloway, J. McTiernan, A. and Malone, B. (2019) A systematic review of peer-mediated interventions for preschool children with autism spectrum disorder in inclusive settings. Review Journal of Autism and Developmental Disorders (6), 40–62.
Meindl, J., Delgado, D. and Casey, L. (2020) Increasing engagement in students with autism in inclusion classrooms. Children and Youth Services Review.
Link to blog about the importance of the control group:
Link to systematic research overview by Steinbrenner et al:
(in the review there is also a link to a material that deals with how the evidence-based efforts can be used in practice)