Yeargin-Allsopp, M, C Rice, T Karapurkar, N Doernberg, C Boyle and C Murphy. 2003. Prevalence of autism in a US metropolitan area. Journal of the American Medical Association 289:49-55.

Press coverage:
Atlanta Constitution Journal
New York Times
Reuters Health

Recent surveys of autism in Brick Township, NJ, California and Britain had surprised experts because the numbers were so large, running up to ten times higher than historical rates would indicate. This new report from the US Centers for Disease Control finds that the number of autistic children living in the Atlanta region is also high, 3.4 per thousand, consistent with the more recent surveys. Most of the autistic children were boys.

This rate does not indicate that Atlanta's rate is higher than other regions in the US. Too few current data are available to reach that conclusion. More likely, other regions in the US are experiencing these higher rates also, at least judging from reports of increased demand for special education services for autistic children.

This survey, the largest of its kind ever undertaken in the US, was unable to resolve a key question: how much of the increase is a real change in the frequency of autism vs. artifactual changes that might result from changes in diagnosis definition, surveillance method or parental incentives to report cases? It nonetheless serves as an important, perhaps definitive benchmark against which to compare future research. It also yielded important insights into the demographics of autism, particularly on risks for boys vs. girls and on race.

The CDC has studies underway in other 13 other communities around the US designed to provide better information.

What did they do? Yeargin-Allsopp et al. mounted an ambitious 5-county survey in the Atlant area to identify all children with any form of autism between the ages of 3 and 10. Their work was carried out in two phases.

• First, they systematically reviewed as many relevant medical/educational (at many different medical, clinical and educational institutions) to identify all children who had been screened using methods that would allow them to be classified as autistic.
• Second, a team of experts scoured through these records and independently determined whether the children met the criteria for autism. They included three variants of autism (collectively referred to as autism spectrum disorder): autistic disorder ("classic autism), Asperger disorder, and pervasive developmental disorder-not otherwise specified" (PDD-NOS). Their diagnoses followed the criteria described in the standard "Diagnostic and Statistical Manual of Mental Disorders, 4th Edition." This detail is important because it a somewhat broader set of criteria than used in previous decades.

Note also that this survey was all through analysis of records; no children were seen directly.

What did they find? Of the 289,456 children aged 3 - 10 living in the 5-county metropolitan Atlanta area in 1996, 987 were found to meet the criteria for autism spectrum disorder. This rate is 3.4 per 1000.

No differences were found between children of different races. Most of this children surveyed were either black (38%) or white (58%).

Yeargin-Allsop et al. also found that the rate of autism varied with age (left). They note that it is not surprising to observe lower rates at the youngest ages, because the syndrome either may not have developed or been diagnosed. The decline at ages 9 and 10 raises other questions, perhaps having to do with changes in diagnostic criteria. Narrower criteria were used before the publication (1994) of those used in this study.

What does it mean? This research serves best to provide a rigorous benchmark for the prevalence of autism in a large metropolitan area, Atlanta, within the US. Rates in Atlanta are consistent with other recent studies, and differ from those in previous decades.

Unfortunately, because of changes in diagnostic procedures, in public attention given to autism, and in the nature of services provided to autistic children (altering incentives for reporting), "debate continues about whether the overall prevalence of autism has increased or whether past rates underestimated true prevalence."

Recent work conducted for the California State Legislature presented several analysis supporting the interpretation that the change is real. This work is available publicly, but has not yet been peer-reviewed. That review process may reveal weaknesses in the conclusions not readily apparent.

The authors of the current study conclude that this debate "is difficult to conclude retrospectively.