Careful Counting: How Many People Have an ASD?
Connie Anderson, Ph.D.
Date First Published: March 2, 2010
Over the past several years, the media has widely reported on the number of people who have an autism spectrum disorder (ASD). That number is high, and keeps going up, causing both interest and alarm. The latest figure of 1 in 88, released by the U.S. Centers for Disease Control and Prevention (CDC) in March of 2012, 1 again showed an increase, even over figures from just a few years ago.
Where are these numbers coming from, and how were they calculated? How are we defining autism spectrum disorder, and how many people have one according to that definition? Do more and more people truly have an ASD, or are we getting better at identifying the condition? Are we counting everybody, or are some people being missed? We look to the latest studies on autism rates to find out.
Counting and Comparing: Harder Than it Sounds
You would think that counting the number of people with a condition would not be that difficult, but it is actually very tricky.
Researchers focused on this task are trying to calculate what they call "prevalence." According to the U.S. National Library of Medicine online medical dictionary, prevalence means "the percentage of a population that is affected with a particular disease at a given time." For example, there might be interest in the percentage of Latino men in California who were diagnosed with lung cancer in a certain year, or the percentage of school children in Minnesota who had swine flu in the fall of 2009.
When it comes to ASD, or to any condition, the difficult part is how to count. How do you define a "case"? Do you count only people who were diagnosed by a doctor? How about those who believe they have the condition but haven't received a formal diagnosis? (Will they be evaluated, and included, if it turns out they have it?) What about those who used to have what is supposedly a lifetime condition, but have improved so much they no longer qualify as having it? (Were they misdiagnosed in the first place? Or have they recovered?) Do you base your count on medical or school records of those already diagnosed, or do you try to identify cases from scratch using interviews, surveys, or in-clinic assessments? You don't want to include people who don't have the condition, and you don't want to leave out people who do.
Since your rate equals the number of cases divided by the total number of people in a group, a careful count of the noncases is necessary, too. If you went door-to-door, then this is all the people you talked to; if it's a telephone survey, it's all the people you called. But what do you do about people who didn't answer the door or the phone? If you're using census data, what about people who avoided the census due to immigration concerns or political beliefs? It gets pretty complex.
Figuring out whether the percentage of people with a disease or disorder is increasing, decreasing, or staying the same is also complicated. To compare, you have to keep the definition of what a "case" is the same, and count the same way at least two different times. You also have to take into account the precision of your count. (Just like polls reported on TV news programs, there's always "sampling error" -- that is, numbers aren't exact but only accurate "plus or minus some percent.")
Finally, you have to be aware of all the other factors that could influence your count. Suppose there's an area with some kind of chemical contamination in the groundwater, but it's also the area with the best public school autism program in the state. A new count of children with ASDs in the area shows an increase since the last count was taken. Is this because the chemical contamination is causing autism, or because people with children on the autism spectrum are moving to the county to get access to the excellent school program? Could it be a little of both, or a third factor not yet considered? It takes very careful science to address such questions.
Numbers on the Rise
Over the past few years the rate for ASD has kept going up, and often words like explosion 2 or epidemic 3 have been used. Comparing past and current autism cases has been a challenge, however. Until recently, nobody had ever counted people in the same place, the same way, twice. At the same time, definitions kept changing. To count "autism" cases was one thing, but what about counting the conditions falling under the broader autism spectrum? These would include pervasive developmental disorder-not otherwise specified (PDD-NOS), which only became an official diagnosis in 1987, and Asperger syndrome, which became an official diagnosis in 1994. 4 Once you include such conditions, you cannot compare your count to those taken in previous years for autism alone.
In a review of many prevalence studies conducted over the past four decades, researchers did find rates were increasing. The median prevalence rate for autism in studies conducted between 1966 and 1993 was 1 in 2,128, while the median prevalence rate for autism in studies conducted between 1994 and 2004 was 1 in 787. 5 They also found that recent efforts to measure the combined prevalence of autism, Asperger syndrome, and PDD-NOS were converging around a single number. Four separate studies using "improved methods," and which had been "conducted in different regions and countries by different teams," had arrived at a figure of about 1 in 166. 6
Had rates of ASD increased, and were they still increasing? It was difficult to tell for two reasons. First, ASD was still diagnosed based on a person's behavior. It was not based on a blood test, brain scan, or other biological measure. This made deciding who "had" autism challenging. Even experts might not always agree that a person did or did not have an ASD. Second, just counting people who were already diagnosed could mean missing a great many people. Researchers stressed that they needed to find ways to make sure "undiagnosed individuals with significant autism symptoms" were counted, too. 7
Although there is still no biological test for autism, recent studies have tried to overcome some of the problems of the past when counting and comparing ASD cases, including counting people who may have never received a formal diagnosis.
A Careful Tally
In 2000, the U.S. Centers for Disease Control and Prevention (CDC) created a program to measure the rate of ASD in the United States: the Autism and Developmental Disabilities Monitoring Network (ADDM). They set up a number of sites around the country, counting all children who were 8 years old in any study year. They planned to identify children who had an ASD or who might have one based on their behaviors. 8 At all sites, they screened medical records. At many sites they also looked at special education evaluations made by community professionals, such as developmental pediatricians, child psychiatrists, or speech pathologists.
The first counts, taken in 2000 and 2002, resulted in an autism rate of 1 in 150, that is, one out of every 150 children had an ASD. 9,10 The next, taken in 2006, yielded an even higher autism rate of 1 in 110. 11
The Latest Count(s)
As of early 2012, a number of studies, including a follow-up to the CDC's "1 in 110" study, have provided a count of autism cases. Although very different, these studies seem to be driving toward a rate of ASD of about 1 in 100, or 1% of the population. Here, we will review how each study was carried out and what it showed.
The CDC repeats its count in 2008: 1 in 88 (U.S.)
In 2008, matching its methods to those it used in 2000, 2002, and 2006, the CDC's Autism and Developmental Disabilities Monitoring Network once again counted 8-year-old children with ASD in selected sites across the U.S. Again, both medical and educational records were used to decide whether a child had an ASD. This time the researchers came up with an overall rate for ASDs of 1 in 88. Eleven of the 2008 sites also had been counted for the 2006 study. A comparison of the numbers for the two years showed an average increase in ASD cases of 23%. A comparison with the earlier 2002 number, based on thirteen sites covered by both the 2002 and 2008 counts, showed an average increase in ASD cases of 78%. 1
Counting based on parent-report by telephone survey: 1 in 91 (U.S.)
One group of researchers used data collected for the 2007 U.S. National Survey of Children's Health (NSCH) to calculate autism prevalence. 12 The NSCH was a telephone survey that gathered information from parents about the health of their children. For this autism study, researchers relied on parent-reported information about 78,037 children aged 3 to 17. In the survey, parents were asked if they had ever been told by a doctor or other health care provider that their child had autism, Asperger disorder, pervasive developmental disorder, or autism spectrum disorder. If a parent said "yes" to the ASD question, they were also asked whether their child currently had the condition, and to rate the severity of any current ASD. If a parent said "yes" to the "ever told by a doctor" question and to the "current ASD" question, the child was counted as having ASD. Based on these parent reports, the researchers found that 1 in 91 children had a current ASD.
One interesting finding was that 38.2% of children who had "ever" received an ASD diagnosis no longer had one. These 453 children were not counted as having ASD for the purposes of the study, but their existence raises the question: Did they move off the autism spectrum because of early intervention? Did they mature or grow out of autism? Were they initially given the diagnosis so they could get services even though the diagnosis was not certain, and it was later changed? Or did they still have the condition but were recategorized by bureaucrats so that services would not have to be provided?
The study could not determine this, but did find that children whose parents had not finished high school had twice the chance of losing their diagnosis as children whose parents had more education. Race also played a role, with black children more likely than white children to lose an ASD diagnosis. Could this indicate that children hadn't really lost their ASD, only their official label and right to services? A parent with less education, or in a racial group suffering more social disadvantage compared with others, might not be able to advocate for the child as effectively. They might be more vulnerable in a situation where a school declared a label and special services withdrawn. Another possibility is that the quality of the diagnosis a disadvantaged child receives is not as high or dependable as one a child with access to better health care receives.
As has been found in many other studies, children who had ever been diagnosed with ASD, whether they had "lost" this diagnosis later or not, were more likely than never diagnosed children to have had other issues like attention deficit hyperactivity disorder (ADHD), anxiety, or conduct problems.
Counting diagnosed and suspected cases via the "special needs register" of the Child Health Services: 1 in 86 (U.K.)
Researchers in the United Kingdom set out to calculate the rate of ASD in children living in South Thames. 13 They focused on children born between July 1, 1990 and December 31, 1991. These children were 9 or 10 years old at the time of the study. Like U.S. researchers who used the age of 8 as a good time to count, the British researchers believed most ASD cases would have been identified by this point. They screened all children with a current ASD diagnosis and those they thought might be an undetected case. Any child who had a "statement of special educational needs" in their record was considered a possibly undetected case. A sub-sample of 255 children received in depth clinical assessments that included two gold standard tools for identifying autism: the Autism Diagnostic Observation Schedule (ADOS) 14 and the Autism Diagnostic Interview -- Revised (ADI-R). 15 The researchers calculated an ASD prevalence rate of 1 in 86.
They reported that they had found many undiagnosed cases, and that they would have gotten a figure of 1 in 227 if they counted only the children formally diagnosed with ASD before the study assessed them. "We have shown how population screening of children already recognized to have special education needs specifically for ASD identifies many more cases," they said. 16
It was the children of less educated parents who were least likely to have been diagnosed beforehand. In the U.S. telephone survey study, children of less educated parents were most likely to "lose" their diagnosis; in this British study, children of less educated parents were less likely to have had an ASD diagnosis before the study team gave them one.
A school-based population study: 1 in 64 (U.K.)
To test whether the rate of ASDs found in the South Thames study applied to other parts of the United Kingdom, researchers used different methods to estimate the rate of ASD in Cambridgeshire. 17 Rather than using the special needs register to screen for children considered at risk of having an ASD, they focused on screening the entire primary school population. Their goal was to detect all cases of ASD, diagnosed or not. First, they did count children known to have an ASD according to the special needs register. Next, they asked parents to complete mail-in surveys, which included questions about whether a child had been diagnosed with an ASD, as well as questions that screened for undiagnosed ASD. They then brought many they suspected of having an ASD to their clinic for an assessment that once again included the ADOS and ADI-R.
For the children with ASD in the special needs register who had already received a diagnosis, the researchers got a prevalence rate of 1 in 106. For the surveyed parents who said their child already had an ASD diagnosis, the prevalence rate was 1 in 101. However, the researchers also found that for every three diagnosed cases of ASD in the primary school population, there were two undiagnosed cases. Once these cases were accounted for in the numbers, the prevalence rate became 1 in 64.
"The data converge on the conclusion that, despite differences in methodology, the prevalence of known cases of autism spectrum conditions is slightly above 1%," the researchers stated. 18 "Our results…highlight the reality that there are children with autism spectrum conditions, notably children with high-functioning autism, who remain undetected in primary schools." 19
Adults with ASD identified by door-to-door interview: 1 in 100 (U.K.)
In 2007, when it was time to repeat a national survey on psychiatric conditions in England, Scotland, and Wales, ASD was added as a topic of interest for the first time. A stated goal of the survey was "to estimate the prevalence of ASD among adults, using diagnostic criteria, and to identify the nature and extent of social disadvantage associated with ASD." 20
The survey's first phase involved interviews with over 7,000 eligible participants. Part of this interview involved screening for mental illness, ASD, or substance abuse. To identify possible ASD, the researchers used a 20-item questionnaire called the Autism Quotient (AQ-20). The ADOS was completed with 618 individuals suspected of having ASD by clinically trained interviewers.
The researchers found a rate of ASD in adults living in private households in the U.K. of 1 in 100. This figure, the researchers noted, "means that the data provide no support for there being an increase in prevalence in recent decades." 21
Numbers Converging: 1 in 100?
It is fascinating that recent attempts to measure the prevalence of ASD seem to be getting nearly the same result: about 1 in 100 individuals, or 1% of the population. Many researchers do admit, however, that they still may be missing some higher functioning or disadvantaged individuals. Table 1, which follows, summarizes the findings of recent prevalence studies.
Table 1. Recent Autism Prevalence Studies