Crossposted from NeuroLogica Blog

Over the last 20 years the prevalence of autism (now part of autism spectrum disorder, ASD) has been increasing. The medical community is largely agreed that this increase is mostly due to expanding the diagnostic category and greater efforts at surveillance. There remains some controversy over whether or not these factors explain all of the measured increase, or if there is a small real increase hidden in there as well. But largely – we are finding more children with ASD because we are casting a wider net with smaller holes.

If this is true, then we do not yet know what the true prevalence of ASD is. There must be a pool of undiagnosed children out there. Eventually the measured prevalence will hit the ceiling of the true prevalence (unless, of course, we expand the definition further) – but where is the ceiling?

That is the question researchers recently set out to answer, and they did so with a comprehensive 5 year study conducted in South Korea. The results surprised even them:

Results: The prevalence of ASDs was estimated to be 2.64% (95% CI=1.91–3.37), with 1.89% (95% CI=1.43–2.36) in the general-population sample and 0.75% (95% CI=0.58–0.93) in the high-probability group. ASD characteristics differed between the two groups: the male-to-female ratios were 2.5:1 and 5.1:1 in the general population sample and high-probability group, respectively, and the ratios of autistic disorders to other ASD subtypes were 1:2.6 and 2.6:1, respectively; 12% in the general-population sample had superior IQs, compared with 7% in the high-probability group; and 16% in the general-population sample had intellectual disability, compared with 59% in the high-probability group.

The previous estimate of autism prevalence was 1% of the population, or about one child in 100. This study found a prevalence of 2.64%, or about one child in 38 – more than twice the previous estimate. They came upon their higher measurement by taking a thorough survey of the general population. Previous studies have looked at high probability groups – children receiving special services or who have already been diagnosed. This study went into the general population and did a thorough survey for undiagnosed cases. Therefore there is a vast untapped pool of potential ASD diagnoses out there.

The results above also indicate that children with undiagnosed ASD in the general population had less intellectual disability than those in the recognized high probability group. They were also less likely to be male and less likely to have classic autism rather than a more subtle variant than the high probability group – which is not surprising. In other words, the undiagnosed children in the general population met the diagnostic features to be considered on the spectrum, but were largely functioning well in mainstream classrooms. In some cases parents were in denial about their child’s condition, in other cases the parents simply had no idea. In South Korea there is apparently still some stigma attached to the diagnosis.

While the authors conclude that their results indicate the need for still better detection of ASD, many of the undiagnosed children would likely not require or even benefit from special services. Although some would, and of course it would be desirable to capture all of those children.

While 2.6% is a high number for any such disorder, it is not out of line with other common mental disorders such as anxiety, depression, or ADHD. Of course these questions always bring up the very relevant issue of where to draw the line between “normal” and “disordered.” As I discussed recently, categorizing brain function is tricky business. Any identifiable psychological or neurological trait seems to vary at least along the classic bell-curve. You can therefore take any trait and declare two standard deviations to either side as the cut-off for “normal” (a standard practice in much of medicine) and declare those at the fringes to have one or another disorder. That would result in 5% of the population being abnormal.

But it takes more than being at the tails of the bell curve to be considered as having a disorder. The definition also requires that the identified traits are associated with (and plausibly cause) some dysfunction or negative outcome. In the case of ASD the disorder is a lack of social ability (not just learned skills, but the raw neurological hardwiring that underlies our ability to socialize). Interestingly, the current measured rate of 2.64% is almost exactly two standard deviations to the left of the curve (the other 2.5%, making a total of 5%, is the cutoff to the right of the curve – those with high social ability, which is generally not considered a disorder).

Since many of the children captured in the current study seemed to be doing fine, it is possible that the current definition of ASD is simply capturing the left two standard deviations of human variability along the bell curve of social ability. Perhaps the definition is therefore too broad, and needs to be tied more closely with some measure of disability. That is a subject for future research.

The bell curve hypothesis also can be used to support those in the “neurodiversity” community. They argue that ASD is just what I described – normal human neurological variation. I agree with this view to some degree, and I think the data above support that. However – when you get out far enough to the left side of the bell curve you do get to the point where dysfunction is undeniable. At some point it is useful to consider a neurological phenomenon to be a disorder. Children with low social ability (even if they make up for it in other ways) tend to have difficulty in school, with making friends, and later in life functioning in the work environment. No matter what you choose to call it, it is useful to identify children who can benefit from programs to help them compensate for their lack of social ability.

Also – we cannot assume that ASD is simply everyone more than two standard deviations to the left of the bell curve. It is possible that the actual curve is not a pristine bell-shape but is bi-modal, representing one hump of normal human variation, and then another hump at the low end of social ability that represents a theoretically definable separate population. This second hump might represent those with one of a group of genetic variants that leads to what we recognize as autism. This is almost certainly true, as children with autism have a higher incidence of intellectual impairment and seizures, suggesting a neurological disorder and not just normal variation. There is also increasing evidence of genetics links to autism.

Further – the low end of the bell curve of normal variation would blend imperceptibly into the second hump of autism disorder. At present the diagnosis of ASD is based entirely on clinical features, making it difficult to separate out different underlying causes. (As I stated above, we can make subtype distinctions based upon associated neurological conditions, but this is still a clinical inference rather than a distinction based upon known cause.)

As neuroscience advances, h0wever, it may become possible to tease out the current mixed bag of clinical ASD by identifying specific underlying genetic or neurological conditions. We may undo the lumping of all these children into one ASD spectrum by identifying subtypes by either their genetic profile, or perhaps their neurological function as examined by functional MRI scanning or a similar functional scan. We are already making significant progress in this area, but this is still an area ripe for further research.


This study adds an interesting data point to the whole picture of ASD. If correct, then the theoretically upper limit of ASD prevalence is about 2.6% of the population, more than twice the previous estimate. It also indicates that when you undergo a program of thorough searching, you will find more diagnoses. No one can reasonably think that the true prevalence of ASD suddenly doubled.

While it doesn’t prove that the steady increase in ASD diagnoses over the last 20 years was due to increased surveillance, it does support that hypothesis by showing the potential of just looking harder. Those children with ASD were always there, they were simply not identified.

Posted by Steven Novella

Founder and currently Executive Editor of Science-Based Medicine Steven Novella, MD is an academic clinical neurologist at the Yale University School of Medicine. He is also the president and co-founder of the New England Skeptical Society, the host and producer of the popular weekly science podcast, The Skeptics’ Guide to the Universe, and the author of the NeuroLogicaBlog, a daily blog that covers news and issues in neuroscience, but also general science, scientific skepticism, philosophy of science, critical thinking, and the intersection of science with the media and society. Dr. Novella also contributes every Sunday to The Rogues Gallery, the official blog of the SGU.