A Spectrum of Disputes
Susan Notes: As the researchers note, more and more children are being labeled with autism. But, as they also observe, an expansion of criteria makes it impossible to compare apples to apples when looking at data on long-term trends, because what counts as “autism” is simply quite different today. However it is defined, the burden on schools and families is clear.
By Paul T. Shattuck and Maureen Durkin
TODAY the special "vaccine court" at the United States Court of Federal Claims in Washington will begin hearing Cedillo v. Secretary of Health and Human Services, the first case of about 4,800 similar ones to examine whether childhood vaccinations can cause autism. We have no wish to comment on these legal issues. But having spent years researching the prevalence of autism in American children, we are concerned that publicity surrounding the case will only drag out debate about whether past trends indicate we face an autism "epidemic."
The claims for or against an autism epidemic simply cannot be proved given the evidence available. (Public attention to the issue was set off primarily by a 1999 report by the State of California that found "a 273 percent increase in the number of persons with autism between 1987 and 1998.") In the end, arguing over what the old data mean just detracts from the more pressing issues involved with scientific research and building cost-effective support systems for affected children, adults and their families.
The latest data from the Centers for Disease Control and Prevention indicate that 1 in 150 8-year-old children are on the "autism spectrum." This proportion is alarming if compared directly to estimates of the frequency of autism before the 1990s, which were in the range of 1 per 2,000 to 5,000. But does this really mean we have a growing autism epidemic, or have we just become better at counting autistic traits in the population that have always been there at roughly the same level?
What remains undisputable is that more and more children are being labeled with autism. In 1943, when the child psychiatrist Leo Kanner published the first case studies of autism as a medical condition, it was characterized by very severe impairments in social interaction and language and communicative abilities, combined with the presence of unusual repetitious behaviors.
As our understanding of the symptoms Kanner described has grown, the notion of an autism spectrum, based on gradations of severity, has evolved. Most of the more mildly affected children who are considered to be on the spectrum today would never have qualified for an autism diagnosis using older criteria. This expansion of criteria makes it impossible to compare apples to apples when looking at data on long-term trends, because what counts as "autism" is simply quite different today.
Another fact hinders our ability to accurately measure changes in autism prevalence: while scientists have made improvements in the ability to detect autism earlier, its causes are still not known and there are no biological diagnostic tests. Diagnosis is made by observing behavior to see if it corresponds with formal diagnostic criteria. Disagreements over whether a child has autism are not uncommon, especially among those more mildly affected.
Advances in techniques for monitoring autism as a public health problem have also led to statistical confusion. Whereas older studies used narrow definitions of autism and were generally based on counting the number of patients in a clinic or hospital with diagnoses of autism, modern methods use broader criteria and leave no stone unturned in the effort to find every autistic child in a defined geographic area, including those not previously given a diagnosis. This virtually guarantees that new estimates will be higher than previous ones, even if the underlying prevalence of the condition has not changed.
Also contributing to the rising identification of autism "and many other conditions that are diagnosed in childhood like attention deficit hyperactivity disorder and bipolar disorder" was the growing number of federally financed efforts to find and help preschool children with developmental disabilities in the 1980s and 1990s. More recently, increased federal financing has been specifically aimed at raising awareness of the early signs of autism and the importance of early identification. It's hardly a surprise that looking harder to find children with disabilities has resulted in more diagnoses.
Clearly there are many alternative explanations for changing autism prevalence. This is what makes it such a difficult phenomenon to investigate. Some who feel that autism has reached epidemic proportions cite enrollment trends in special education. In an effort to clarify things, we collected and examined the special education enrollment data from every state and for every classification category between 1984 and 2003. We analyzed the numbers using several different statistical approaches, and we arrived at similar conclusions regardless of which method we used.
According to federal data, the number of children receiving special education services for autism has increased each year since 1991, when Washington's guidelines first required schools to break out counts of children on the spectrum. In the 1991-92 school year, 5,415 students ages 6 to 21 in special education programs were identified as autistic. By 2005-06, this number had grown to 192,643, a 3,500 percent increase.
Though large, this increase does not necessarily provide evidence of an autism epidemic or an actual increase in the prevalence of autism. Given that schools cannot respond to a federal mandate overnight, the baseline count in 1991 was clearly an underestimate of how many children in the nation truly had autism. This renders the resulting percent change estimates over time utterly uninformative about changes in true prevalence. (In fact, if the C.D.C.'s estimate that 1 out of 150 children are on the spectrum is correct, the 2005 count still under-represents how many truly have autism.)
In addition, federal criteria for classifying children into the autism special education category are not equivalent to those used for medical diagnoses, and identification practices vary widely among school districts. Before the 1990s, children with autism were tallied in other categories, though no data exist to tell us exactly which ones.
Urban planners know that "if you build a new road, people will drive on it." Likewise, in special education, if you create a new counting category, people will use it. At the same time, older counting categories may then be used less frequently as our ability to differentiate diagnoses improves. And true enough, our study found that the growth in children classified with autism was accompanied by a corresponding decrease in the number of special education students with other designations.
This is an example of "diagnostic substitution" --s information on new autism classifications has gradually spread, the label "autistic" has been used more commonly for children who previously would have been labeled something else. Our analyses demonstrated that from 1994 to 2003, in 44 of 50 states, the increase in autism was completely offset by a decrease in the prevalence of children considered "cognitively disabled" or "learning disabled."
What does this all mean? First, we should be wary of "epidemic" claims and percentage increases based on administrative data. Second, we should not be surprised if school counts of children with autism continue increasing as they play catch-up to the number who truly have autism.
We want to be very clear: our results do not mean we have nothing to worry about. Scientific and clinical advances have improved our ability to identify autistic children and to differentiate their unique needs from those of children with other types of developmental disabilities. But schools and other social service systems are unable to keep pace with these changes or give the children the help they need.
Research to discover what causes autism, including possible environmental triggers, must be a top priority. However, autism is not purely a medical puzzle -- as we invest in new ways of understanding autism, we have a corresponding responsibility to invest in the capacity of our schools, medical centers and social workers to provide up-to-date treatment for those with the condition and support services for their families. In the end, we should not have to deliver a verdict on whether there is an epidemic to fulfill these obligations.
Paul T. Shattuck is an assistant professor at Washington University's School of Social Work. Maureen Durkin is an epidemiologist at the University of Wisconsin's School of Medicine and Public Health.
Paul T. Shattuck and Maureen Durkin
New York Times
INDEX OF RESEARCH THAT COUNTS