New methodology
When ADDM released its first report in 2007, autism diagnosis was widely inadequate in the United States. Relying on only confirmed ASD diagnoses would significantly underestimate the number of children with the disorder, so the CDC added “active case finding” to the protocol.
Trained clinician reviewers analyzed individual notes from medical and educational records for every 8-year-old in ADDM Network sites, looking for evidence of characteristics and behaviors associated with autism. The process was labor- and time-intensive and took up to 4 years to complete.
In 2018, the CDC began investigating ways to speed the process and came up with the strategy used in the latest report. The new protocol was faster, easier, and less expensive. Although he says cost was never the deciding factor, Dr. Maenner acknowledges that had they stuck with the original protocol, they would have been forced to reduce the number of ADDM Network sites.
Dr. Maenner argues that a comparison of the two protocols shows the new method doesn’t compromise accuracy and may actually capture children who lacked the medical or educational records the previous protocol required for a count. But not everyone agrees.
“I thought the point was to be as accurate and complete as possible in doing the surveillance,” Walter Zahorodny, PhD, associate professor of pediatrics at Rutgers University, New Brunswick, N.J., and principal investigator of the New Jersey ADDM Network site, told this news organization. “In states where there’s a high detail of information in records, like New Jersey, it’s going to underestimate the count.”
Dr. Zahorodny says the latest data prove his point. In 2016, under the old methodology, ASD prevalence was 3.1% in the state. In 2018, under the new protocol, prevalence was 2.84%, a decrease of about 20% that Dr. Zahorodny pins squarely on the elimination of ADDM clinical reviewers.
But New Jersey is the only state that participated in both the 2016 and 2018 surveillance periods to report a decrease in ASD prevalence. The other eight states all found autism rates in their states went up.
Sydney Pettygrove, PhD, associate professor of public health and pediatrics at the University of Arizona, Tucson, and a principal investigator for the ADDM site in Arizona, told this news organization that when she first learned the CDC was rolling out a new methodology, she and other investigators were concerned.
“People were really upset. I was really upset,” she said. “I had formed an opinion based on the earlier data that this would not be a good idea.”
In 2000, when ASD surveillance began in Arizona, nearly 30% of children identified by ADDM clinical reviewers as having autism had no mention of the disorder in their records. Today, that percentage is closer to 5%.
“In 2000 it would have been catastrophic to try to estimate the prevalence of autism with the new protocol,” said Dr. Pettygrove. As it turns out, under the new protocol, prevalence rates in Arizona increased from 16.0 per 1,000 children in 2016 to 24.9 in 2018.
Built-in bias eliminated?
In addition to speeding up the process, the new methodology might have other benefits as well. Under the old ADDM surveillance protocol, children who lacked certain medical or educational records did not meet the ASD case definition and weren’t counted.