How many children have “autism”? Is that number increasing? Is there an “epidemic” of autism or have we merely been continually refining it, expanding it and moving the goalposts since it was first described by Leo Kanner in 1943?
I met my first child with autism in 1959, almost 60 years ago. I had the good fortune to learn about autism firsthand from Kanner himself, when he was a visiting professor at the University of Wisconsin Medical School and I was a medical student there.
Then, in 1962, I started a Children’s Unit at Winnebago Mental Health Institute in Wisconsin, on which almost all the children were autistic. That’s also the unit on which I met my first savant.
The question of autism prevalence engaged me even then. In 1970, I carried out the first U.S. study of the epidemiology of infantile autism, published in Archives of General Psychiatry. Actually, autism was then most commonly diagnosed formally as childhood schizophrenia.
At that time, the Wisconsin Department of Health and Human services provided me with a printout listing all patients age 12 and under seen for evaluation or treatment and given a diagnosis of childhood schizophrenia between fiscal 1962 and 1967 in 30 community mental health and child guidance clinics; four state and county mental hospitals; three colonies and training schools; and the children’s treatment center, children’s diagnostic center and university hospitals.
I found 280 unduplicated cases, representing a prevalence of autism of 3.1 cases per 10,000 children ages 3–12 in Wisconsin. Interestingly, as a validation, a 1966 study by Lotter in the county of Middlesex, United Kingdom in 1966 found a prevalence of autism of 4.8 cases per 10,000 children. Admittedly this study has some serious limitations, which make the prevalence figure for autism artificially low. But it stands in stark contrast to the prevalence of one child in 59 with a diagnosis of autism in the 2018 Centers for Disease and Control and Prevention report (which uses 2014 data). That report found 168 children per 10,000 instead of 3.1 per 10,000 in my 1970 study. Which of those two divergent figures is closest to being correct? And how did the figure get that divergent?
THE CDC REPORT
When the report came out, the headlines read along the lines of “Autism cases continue to rise: now 1 in 59 children have autism.” But let’s look at that CDC study more critically. It is based on an active surveillance system established in 2000 that estimates autism spectrum disorder (ASD) among children age 8 years living in 11 states.
Using that system, the prevalence of autism (ASD) rose from 1 in 150 children in 2000–2002, to 1 in 68 children during 2010–2012 and 1 in 59 children in 2014. That means the prevalence of autism more than doubled in the 12-year period between 2000 and 2012 and increased nearly 16 percent just in the two-year period between 2012 and 2014.
That is preposterous. From 1 per 150 children to 1 per 59 children with autism in slightly more than a decade? No wonder headlines speak of an “epidemic.” Are these believable figures, or might it be because we keep diluting the condition and expanding the definition, and in so doing we keep moving the goalposts? I believe that to be the case.
There are problems that cast doubt on that method and those numbers for actual prevalence of ASD. Figures include “educational autism,” which is a diagnosis made by teachers or educational specialists in the classroom and “medical autism,” based on review of available medical records. There are no actual in-person evaluations. Casting more doubt is the fact that the prevalence in one state, Arkansas, was 1.31 percent but more than double that in another, 2.93 percent in New Jersey. The prevalence in Wisconsin rose 31 percent between 2012 and 2014. Is that a believable actual increase in ASD in two years in Wisconsin?
I don’t think so. From my perspective as an observer of “autism” for over 60 years, I do believe there is an actual increase in the number of cases of autistic disorder, but it is not an epidemic. And it has not been an increase of 31 percent in two years here in Wisconsin, for example, or a more than 150 percent increase in the U.S. in the past decade. That is simply not believable. Instead much of that “epidemic” is a dilution of the rigor of the criteria for autism.
That may make interesting headlines, increased awareness, expanded insurance coverage or benefit fund raising, but it is not an accurate assessment of the actual prevalence of autism. Recently several leading heart organizations changed the definition of ‘hypertension” from 140 systolic over 90 diastolic to 130/80. Instantly the number of Americans with high blood pressure jumped 14 percentage points from 32 percent to 46 percent.
CALL THINGS BY THEIR RIGHT NAMES
There are many reasons why the diagnosis of autism needs to be precise. Labeling some children as autistic when they have other learning disorders such as hyperlexia or language delay, for example, or “educational autism,” alarms families unnecessarily and can result in the wrong intervention or educational placement, which happens particularly with children who read early or speak late. Even “blindisms”—repetitive self-comforting behaviors such as rocking in children with visual impairments—can be mistaken for autism. As elsewhere in medicine, the first step in treatment is to make the correct diagnosis.
But my concern goes beyond that. I believe that autism is a group of disorders, rather than a single disorder, just as mental retardation or dementia represent groups of conditions rather than single ones. I also believe that one day, as we sort autism into its component parts, we will be able to identify the subgroups with the same precision as we do phenylketonuria with a diaper or blood test or trisomy 21 or fragile x with a chromosome test.
But the less precise and broader the diagnosis becomes, the less chance we have of finding subgroups among increasingly heterogeneous, diluted diagnostic groups. There is a proliferation now of metabolic, enzyme, imaging and brain wave tests or chromosome determinations that hold promise for detecting autism with the precision of the diaper test, trisomy 21 or fragile X.
To do that we need to call things by their right names both clinically and research-wise, and the present method of assessing prevalence in the U.S. is not doing that well since we keep moving the goalposts.
Others share that view. In a Scientific American essay this month titled “Is It Time to Give Up on a Single Diagnostic Label for Autism?” Simon Baron-Cohen states: “But the main argument against a single diagnostic label is that the inclusion of subtypes will likely lead to greater scientific progress in understanding the precise causes of the heterogeneity, and greater translational progress in understanding what kinds of intervention and support are needed, and for whom.” I would add the prospect of prevention to that list of possibilities.
The search for a reliably consistent measure of autism prevalence continues outside the U.S., as well.
A 2017 article in the Universal Journal of Clinical Medicine shows a 2012 global median autism prevalence figure of 17/10,000 or 1 in 588 for autistic disorder and 62/10,000 or 1 in 161 for all pervasive developmental disorders. That is at some considerable variance from the 1 in 59 figure in the U.S.. One of the important remaining tasks in the search is to bring the American Psychiatric Association’s DSM-V definition of autistic spectrum disorder in line with the definition in the World Health Organization’s International Classification of Diseases–Eleventh Revision (ICD-11). They’re still inconsistent.
If that happens, of course, the goalposts may move again.