Autism diagnoses have climbed for two decades. Most of the story is how we define it.

The numbers are real

Every two years, the CDC's Autism and Developmental Disabilities Monitoring (ADDM) Network goes through the health and school records of 8-year-olds in communities across the US and counts how many have autism. In 2000 it was 1 in 150. In the most recent count, for 2022, it was 1 in 31.

Autism among US 8-year-olds, 2000–2022. The line is the combined rate; the shaded band is the gap between the lowest and highest reporting sites in each year. In 2022 the rate in the highest-reporting site was more than five times the lowest.Source: CDC Autism and Developmental Disabilities Monitoring (ADDM) Network.

That is a fivefold rise in barely two decades, and it has climbed in almost every report since the network began. Nobody serious disputes that diagnoses have increased. The real question is whether autism itself has increased by anything like that much, or whether we are counting differently.

To answer that, you have to know what "autism" meant at each point along the way.

What counted as autism, and when

The DSM, the Diagnostic and Statistical Manual of Mental Disorders, is the manual US clinicians use to diagnose autism. Its definition has changed more for autism than for almost any other condition.

1952 – 1979 · DSM-I and DSM-IIThere is no autism diagnosis at all. Children we would now call autistic were typically labelled with "childhood schizophrenia" or intellectual disability, or never labelled.
1980 · DSM-III"Infantile autism" becomes its own diagnosis for the first time. The bar is high: onset before 30 months and a pervasive lack of responsiveness to other people. This describes severely affected children.
1987 · DSM-III-RRenamed "autistic disorder" and loosened: no strict age-of-onset cut-off, a menu of criteria to choose from, and a new catch-all, PDD-NOS, for children who fall short of full criteria.
1991 · US special educationAutism becomes its own category in federal special-education law. Schools start identifying and counting it, with services attached to the label.
1994 · DSM-IVAsperger's disorder is added, covering people with autistic traits but no language delay or intellectual disability. Autism is now a family of diagnoses rather than one rare condition.
2007 · Universal screeningThe American Academy of Pediatrics recommends screening every child for autism at 18 and 24 months, not only those who already worry their parents.
2013 · DSM-5Everything merges into one autism spectrum disorder, graded by support needs. For the first time, autism and ADHD can be diagnosed in the same child; the previous rules didn't allow both.

Put the prevalence numbers on top of those definitions and the pattern is hard to miss.

Every widening of the definition, against the numbers counted. Children per 1,000. Shaded bands are the editions of the DSM (the thin line inside the second band marks the 1987 revision), the manual used to diagnose autism in the US. Vertical scale is logarithmic: each gridline step is a multiple, so a straight climb means steady proportional growth. US surveillance did not begin until 2000; earlier studies used narrower definitions and are not directly comparable.Source: Lotter 1966; CDC ADDM Network 2000–2022; American Psychiatric Association DSM editions.

In the mid-1960s, one of the first population studies, in Middlesex, England, found classic autism in roughly 1 in 2,200 children. That study was counting something much narrower than what we count today: mostly children who were non-verbal, profoundly withdrawn and often had severe intellectual disability. A child with what we'd now call Asperger's-type autism wasn't missed by that study. Nobody was looking for that child, because the concept didn't exist yet.

How much of the rise is the definition?

You can't rerun history with the old criteria. But researchers have found clever ways to estimate how much of the increase comes from changes in diagnosis and counting rather than from more autistic children.

Denmark: about 60%. Denmark has whole-population health registers. In 1994 it adopted the broader international criteria, and in 1995 it began counting outpatient diagnoses as well as hospital ones. A study of every child born there from 1980 to 1991 found that those two changes alone accounted for about 60% of the rise in diagnoses.
California: diagnostic change on its own accounted for about a quarter. Researchers reviewing California's developmental-services records found children who'd been diagnosed with intellectual disability and were later reclassified as autistic. That shift by itself explained about 26% of the state's increase, and they noted that this was only one of several kinds of diagnostic change.
US schools: one label replaced another. After autism became a special-education category, the number of children classified as autistic rose, while the numbers classified with intellectual disability and learning disabilities fell. Many children weren't new. They had been relabelled.
Sweden: the traits stayed flat while diagnoses rose. This is the most striking study. Researchers interviewed the parents of more than 19,000 Swedish twins born over a 10-year period about autistic traits, the actual behaviours, without regard to diagnosis. The proportion of children with those traits did not change over the decade. Over the same years, registered autism diagnoses rose steeply. The children were the same; the diagnosing changed.

Clues hiding in the CDC's own numbers

If something in the environment were driving a surge, you'd expect it to show up fairly evenly. Instead, the CDC data shows the fingerprints of who is looking, and how hard.

It depends heavily on where you live. In 2022, the reporting sites ranged from 9.7 per 1,000 in Laredo, Texas, to 53.1 per 1,000 in California, a more than fivefold difference within one country in a single year. That's the shaded band in the first chart. Differences that large track access to evaluations and services far better than they track any known exposure.
The groups have flipped. Until about 2016, autism was identified most often in white children from wealthier neighbourhoods, the families with the easiest access to specialists. Since 2020 the rates in Black, Hispanic and Asian or Pacific Islander children have been higher than in white children. A biological epidemic wouldn't change which communities it favours within a few years. Better outreach to under-served communities would.
Diagnosis is getting earlier. Children born in 2018 were 1.7 times as likely to be identified by their fourth birthday as children born in 2014. That reflects screening, not biology.

Where are all the autistic adults?

If autism had become five times as common, you'd expect today's adults, who grew up under narrower definitions, to have far less of it. A study in England tested that directly: researchers went door to door, assessing a random sample of adults. They found autism in about 1 in 100 adults, close to the rate being reported in children at the time, and almost none of them had ever been diagnosed. The autistic adults were there all along. They just never got the label.

So is there any real increase?

Possibly, and honesty requires saying so. A few things probably add a genuine, smaller increase:

Parents are older on average, and older parental age is linked to a modestly higher chance of autism.
More very premature babies now survive, and they are at higher risk.

These are real effects, but they are modest. Neither comes close to explaining a fivefold jump, and none of them is a hidden toxin or a vaccine. Large studies have found no link between vaccines and autism.

That's why the verdict here is Misleading, not Myth. The rise in diagnoses is real. The claim that it proves an epidemic with a new hidden cause doesn't follow from it.

Why this matters

When a scary number is presented without its history, it gets used to sell fear and, often, products: supplements, "detox" regimens, unproven therapies. It also does something quieter. It frames the children and adults who finally received an accurate diagnosis as victims of a catastrophe, when many of them are people our medical system used to miss.

More diagnoses mostly means more children getting seen, and getting support earlier, when it helps most.

The bottom line

Autism diagnoses in the US have risen from 1 in 150 to 1 in 31 since 2000. Over the same period, and before it, the definition of autism grew from a rare, severe childhood condition into a broad spectrum, schools began counting it, and pediatricians began screening every toddler for it. Studies from Denmark, California, Sweden and England point the same way: most of the rise reflects how we diagnose, not a sudden change in children.

If you're worried about your child's development, don't wait for the debate to settle. Ask your pediatrician about a developmental screen. Early support makes a real difference.

Sources

← All articles

Educational content, not medical advice, and no doctor–patient relationship. Talk to your own clinician about your care. Views are my own and not those of any employer or institution.