Podcast: How accurate are diagnostic tools for autism spectrum disorder in preschool children?

Senior author, Katrina Williams from the Department of Paediatrics at the University of Melbourne in Australia, tells us what they found in this podcast.

There are more than 20 Cochrane Reviews of interventions that might be used in the care of children with autism spectrum disorder. In July 2018, these were added to with an assessment of the accuracy of tests for this condition. Senior author, Katrina Williams from the Department of Paediatrics at the University of Melbourne in Australia, tells us what they found in this podcast.

"Doing our review, we wanted to answer the question: How accurate are tools for diagnosing autism spectrum disorder, which I’ll refer to as autism, in preschool children? And we’ve reached the conclusion that they are not accurate enough. This is important because we need to diagnose autism correctly so that children with autism can access timely support and education and children who don’t have autism avoid unnecessary investigations and treatments.

We found relevant evidence for three diagnostic tools: the Autism Diagnostic Inventory, which is a carer interview and is known as the ADI-R; the Autism Diagnostic Observation Scale, known as the ADOS, which is based on observing the child while they do structured tasks and the Childhood Autism Rating Scale, or CARS, which combines an interview with un-structured observation. 

The 13 relevant studies were mainly from high income countries and included preschool children with language difficulties, developmental delay, intellectual disability, or a mental health problem, presenting to a clinical service or enrolling in a research study. The studies varied in quality and it’s likely that they appear to be more accurate in making diagnoses in these research studies than they would be in routine practice, which means we need to be cautious about the findings.

The largest amount of evidence was for ADOS, with 12 analyses and a total of more than 1600 children. We found it to be the best for identifying children who have autism, but it was similar to CARS and ADI-R in falsely diagnosing autism. To put this into numbers, the pooled analyses for ADOS gave a summary sensitivity of 0.94, meaning that it identified autism in 94 of every 100 children with it, and a summary specificity of 0.80, meaning that it would diagnose autism in 20 children out of 100 without it. This means that if we used ADOS to examine 100 children, 74 of who truly had autism, it would detect autism correctly in 70 but would also suggest that five of the children without autism actually had it. This makes it especially important to think about two specific settings in which these tools might be used. These are services that assess many children who do not have autism and those looking after children with intellectual disability, because a higher proportion of these children are likely to receive an incorrect diagnosis. 

In thinking about the implications of our findings, it’s important to note that autism tools that are currently considered 'diagnostic' were not designed to make an autism diagnosis and are not sufficient for that. A diagnosis involves consideration of several factors, such as whether behaviours are in keeping with the child’s communication ability and intelligence, and the exclusion of causes that require genetic testing and a detailed understanding of the child’s environment. Also, families need more than diagnoses relating to autism to understand their child’s strengths and challenges and to work with professionals to access the supports and interventions they need. That’s why child health specialists are urging a more dimensional neurodevelopmental approach than the tools we found evidence for when young children are being assessed, and we would encourage such approaches."


Monday, November 19, 2018