Daniel sent us this one — he wants to know what the data actually says about global autism prevalence. Not the headlines, not the panic, not the talking points. The actual numbers. He's asking six things: the diagnostic history from Kanner and Asperger through the DSM evolutions, whether rates are truly rising or if it's mostly changing definitions, where the real epidemiology ends and conspiracy theories take over, how prevalence compares across countries, who's being diagnosed now that wasn't before, and what researchers actually believe about any true underlying increase. It's a big one.
A good one. Because this is one of those topics where almost everyone has an opinion, and almost nobody has actually looked at the data. I've been reading through the literature on this for years, and the gap between what the research says and what the public conversation sounds like is enormous.
Before we dive in — quick note. Today's script is being generated by DeepSeek V four Pro.
New model in the rotation. Alright, let's get into it. Where do you want to start?
Let's start at the beginning. When did autism actually become a medical diagnosis? Because I think most people assume it's ancient, but it's not.
It's remarkably recent. The term "autism" was coined in nineteen eleven by a Swiss psychiatrist named Paul Eugen Bleuler — same guy who coined "schizophrenia," by the way. But he used it to describe what he thought was the childhood version of schizophrenia. So the word existed, but it wasn't its own thing.
For decades, kids we'd now call autistic were being labeled as having childhood schizophrenia.
And then you get the two big names. Leo Kanner at Johns Hopkins published "Autistic Disturbances of Affective Contact" in nineteen forty-three. He described eleven children with what he called "early infantile autism." His very first case, Case One, was a kid named Donald Triplett. Kanner initially thought this was a form of childhood schizophrenia, but by the early nineteen fifties he was already pushing for a biological understanding.
Asperger shows up right around the same time.
Nineteen forty-four. Hans Asperger published his observations of children who had social and communication difficulties but preserved grammatical language — they could speak, often elaborately, but couldn't really connect. What's interesting is the term "Asperger's syndrome" wasn't actually his. It was coined much later by a British psychiatrist named Lorna Wing, who essentially reclaimed Asperger's nineteen forty-four article and brought it into the modern conversation.
The diagnostic framework we have today is really a product of the last forty years, not some ancient medical tradition.
Autism didn't even enter the DSM until nineteen eighty. Before that, it didn't officially exist as a separate diagnosis. And the criteria were narrow — "infantile autism," onset required before thirty months, very specific presentation. The prevalence at that time was estimated at about one in fourteen hundred.
One in fourteen hundred.
The most recent CDC data, released just this month actually — April twenty twenty-five, looking at twenty twenty-two data from the ADDM Network — puts it at one in thirty-one among eight-year-olds in the United States. That's a two hundred forty-three percent increase since the year two thousand.
Okay, so that's the headline number that freaks everyone out. One in fourteen hundred to one in thirty-one. Something must be causing an epidemic, right?
That's the question. And this is where you have to look at what happened to the diagnostic criteria in between. Because the DSM didn't stay frozen in nineteen eighty. DSM-Three-R in nineteen eighty-seven broadened the criteria significantly — they introduced a polythetic set of sixteen items across three domains, dropped the thirty-month onset requirement, and added something called PDD-NOS, pervasive developmental disorder not otherwise specified. That alone brought a lot more kids under the umbrella.
The definition expands, and suddenly more people fit.
Then DSM-Four in nineteen ninety-four introduced the "autism spectrum" concept with separate categories — autistic disorder, Asperger's disorder, and PDD-NOS. By the early two thousands, prevalence was already up to about one in one hundred fifty. And then DSM-Five in twenty thirteen consolidated everything into a single diagnosis — autism spectrum disorder — with two core domains and severity levels. The umbrella got wider while aiming for more precision.
Each revision of the DSM has made it easier, not harder, to get a diagnosis.
Easier in the sense of broader criteria, yes. But also there's better screening, more awareness, reduced stigma. Parents are more likely to seek evaluation. Pediatricians are more likely to refer. Schools are more likely to identify. The question is how much of that two hundred forty-three percent increase is real versus artifactual.
That's where diagnostic substitution comes in, right?
This is one of the most important and least understood pieces of this whole puzzle. There was a major study out of Penn State that looked at special education classifications between two thousand and twenty ten. They found that nearly sixty-five percent of the threefold increase in autism classifications could be accounted for by a reduction in intellectual disability classifications. Kids who were previously labeled as having intellectual disability were being reclassified as autistic.
It's not that there are more kids with these traits. It's that we're calling the same kids something different.
For a substantial portion, yes. And here's the really striking number — for fifteen-year-olds, this diagnostic substitution explained up to ninety-seven percent of the increase. Ninety-seven percent. That means almost the entire rise in autism classifications among older teens was just relabeling.
That's wild. But you said sixty-five percent — that still leaves thirty-five percent unexplained.
And this is where the debate gets nuanced. A separate California study estimated that about twenty-six percent of the increased autism caseload could be attributed to changing diagnostic practices, substitution, and what they call accretion — kids keeping their diagnosis rather than losing it. But that means roughly seventy-four percent remained unexplained by those factors alone. And a later analysis from Penn State found that in most states, intellectual disability prevalence actually stayed flat while autism rose sharply. So substitution is real but it's not the whole story.
The honest answer is: we don't know exactly how much is real versus diagnostic change, but most of it is diagnostic change.
That's the consensus. The global meta-analysis — this is a huge study published in Frontiers in Psychiatry in twenty twenty-three, looking at eighty-five studies across twenty-nine countries from nineteen ninety-four to twenty nineteen — the authors explicitly say, and I'm quoting, "these discrepancies in estimates over time probably are not caused by an increase in true prevalence but are instead associated with changes and improvements in diagnostic categories, methodology, and quality of research.
That's about as clear as academic language gets.
It really is. And the global pooled prevalence they found was zero point seven two percent — about one in one hundred thirty-nine. was at one point one two percent, Sweden at zero point nine, Taiwan at zero point one one. The variation overwhelmingly reflects differences in detection, not biology.
Let's talk about that global variation, because those numbers are all over the map.
The single strongest finding from the -analysis is that the Human Development Index strongly predicts reported prevalence. The higher the HDI, the higher the reported rate. North America has a pooled prevalence of about one percent, Europe about zero point seven three percent, Asia about zero point four one percent. High-income countries are at zero point seven nine percent versus zero point three two percent for lower-middle-income countries.
Richer countries with better healthcare systems find more autism.
And the within-country variation tells the same story. In the U., the latest ADDM data shows prevalence ranging from one in one hundred three in Laredo, Texas, to one in nineteen in San Diego, California. That's not biology. That's screening access, service availability, awareness campaigns, diagnostic culture.
One in nineteen versus one in one hundred three. In the same country.
And look at the international extremes. Countries like Brunei at about one point six five percent, Singapore and Japan around one point four percent, South Korea around one point three to one point five. Then you've got France at about zero point six nine percent, Bangladesh around zero point six, Taiwan at zero point one one in the -analysis. These differences are almost certainly about detection infrastructure, not genuine biological variation.
Which also means the global prevalence number — one in one thirty-nine — is probably an underestimate, because so many countries barely screen at all.
The -analysis included only one study from Africa and four from South America. The data gap is itself a finding. As diagnostic infrastructure improves worldwide, rates are going to appear to rise further. That doesn't mean more people are becoming autistic. It means more people are being recognized.
Alright, let's get to the elephant in the room. Because you can't talk about autism prevalence without someone bringing up the MMR vaccine and Andrew Wakefield.
We need to be precise about what happened here, because the Wakefield saga is one of the most consequential scientific frauds in modern history. In nineteen ninety-eight, Andrew Wakefield published a paper in The Lancet claiming a link between the MMR vaccine and autism in twelve children. The paper was fully retracted in twenty ten after an investigation by a journalist named Brian Deer, writing for the British Medical Journal, revealed that Wakefield had misrepresented or altered the medical histories of all twelve patients.
Five of the children had documented developmental issues before receiving the MMR vaccine, which directly contradicted the paper's claims. Wakefield was struck off the UK medical register for fraud and unethical conduct. It was later revealed that he had been funded by lawyers who were suing vaccine manufacturers — a conflict of interest he never disclosed.
It wasn't just bad science. It was fabricated.
Fabricated, and with a financial motive. But the damage was done. The paper fueled vaccine skepticism that persists to this day. It dropped MMR vaccination rates, contributed to measles outbreaks. A PLOS ONE study in twenty twenty-one quantified that vaccine skepticism increased following the paper's publication, amplified by negative media coverage. And no large-scale epidemiological study — by the CDC, the American Academy of Pediatrics, the Institute of Medicine, the National Health Service, the Cochrane Collaboration — none of them has ever found a causal link between vaccines and autism.
Yet the conspiracy theory persists because the prevalence numbers keep going up, and people see correlation and assume causation.
The "autism epidemic" framing is politically weaponized from both sides. Vaccine-autism conspiracy theorists use rising numbers to claim there's a cover-up. And then some skeptics use the "it's all diagnostic change" argument to dismiss the very real needs of a growing population that requires services. The data supports neither extreme. The increase is mostly real in terms of identification, but not necessarily in terms of incidence.
What about other environmental factors? Because it's not just vaccines. People worry about pesticides, air pollution, heavy metals, plastics.
This is where you have to distinguish between factors that researchers take seriously and factors that have been elevated into panics. Prenatal air pollution, heavy metals, maternal infections, advanced parental age — these are things that researchers study because they may interact with genetic predisposition. But they don't "cause" autism independently. Genetics account for sixty to ninety percent of the risk. As one researcher at the Karolinska Institute put it, "there is no evidence suggesting that environmental factors play a major role in the marked increase in autism.
The environmental stuff is real in the sense that it might modulate risk at the margins, but it's not driving the prevalence numbers.
And the leading hypothesis for how environment and genetics interact is epigenetics — prenatal stress, pollutants, maternal health conditions can alter DNA methylation and gene expression without changing the underlying DNA sequence. That's a real and important area of research. But it's not "pesticides are causing an autism epidemic." It's much more subtle than that.
Let's talk about who's being diagnosed now that wasn't before, because this is one of the most dramatic shifts.
The biggest story is girls and women. Female autism diagnosis rates have risen three hundred five percent over the past twenty years, compared to one hundred eighty-five percent for males. Among twenty-six to thirty-four-year-olds, the increase is four hundred fifty percent.
Four hundred fifty percent.
Despite that, about eighty percent of autistic girls remain undiagnosed before age eighteen. Females are diagnosed on average eighteen months later than males, and twenty-five percent of female patients receive their diagnosis as adults, compared to twelve percent of males.
What's driving that gap?
Masking, also called camouflaging. Girls and women often hide symptoms by mimicking social behaviors. They learn to make eye contact, to script conversations, to suppress stimming in public. They fly under the radar unless cognitive or behavioral issues become obvious. The diagnostic criteria were historically developed based on male presentations, so we've been systematically missing autistic women for decades.
Which raises the question — what's the true male-to-female ratio? Because we've been told it's three or four to one.
Some researchers suspect it could be closer to two to one or even one to one with better recognition. We don't know yet. But the gap is closing fast, and that tells you something about how much of the "increase" is just finding people who were always there.
Adults generally — a lot of adults walking around undiagnosed?
The JAMA Network Open study from twenty twenty-four found that ASD diagnosis prevalence was highest among five-to-eight-year-olds at about thirty per thousand, and it declined with age. Only about one in forty-five adults in the U.has an autism diagnosis. That suggests a substantial undiagnosed adult population. People who grew up before the DSM expansions, before routine screening, before the modern autism concept even existed.
Then there's the shift in intellectual disability co-occurrence.
This is huge. In the global -analysis, only about thirty-five percent of ASD cases had co-occurring intellectual disability — defined as IQ of seventy or below. That means nearly two-thirds of diagnosed individuals today do not have intellectual disability. That's a dramatic reversal from earlier decades when autism was primarily identified in institutionalized populations with severe impairments. The DSM-Five consolidation of Asperger's and PDD-NOS into ASD, combined with the concept of support levels, has brought many previously unrecognized individuals into the diagnostic net.
The face of autism has changed. It used to be a severely impaired child in an institution. Now it's a much broader range of people.
That broader range is closer to reality. But it also means the prevalence numbers from different eras aren't measuring the same thing. Comparing one in fourteen hundred in nineteen eighty to one in thirty-one today is comparing apples to... well, to an entire fruit basket.
What about racial and ethnic shifts? I saw something in the latest CDC data.
In the twenty twenty-five CDC report, prevalence among Asian and Pacific Islander children, Hispanic children, and Black children now exceeds that among White children. Asian and Pacific Islander at four point four seven percent, Hispanic at four point two four percent. That reflects improved outreach and reduced disparities in screening access, not true differences in occurrence. Historically, minority children were underdiagnosed because they had less access to developmental pediatricians and early intervention services. That gap is closing.
The racial disparity is actually a good news story about access to diagnosis.
In a specific sense, yes. It means we're doing a better job of finding kids who were always being missed.
Alright, let's get to the bottom line. What do researchers actually believe about a true underlying increase? You said most of the rise is artifactual. But is there any real increase at all?
No credible researcher uses the term "autism epidemic" to describe a true biological surge. The Autism Society of America, in their response to the latest CDC report, called the rise "diagnostic progress." That's the consensus framing.
You also said the California study found that about seventy-four percent of the increase was unexplained by diagnostic substitution and accretion.
And that's where the nuance lives. Some of that residual could reflect real increases driven by environmental risk factors interacting with genetic susceptibility. Advanced parental age, for example, is rising globally and is a known risk factor. Prenatal exposures to certain medications, maternal infections during pregnancy, very preterm birth — these things have modest but real associations with autism risk. But "modest" is the key word. If there's a true increase, it's likely small and driven by these kinds of factors, not by any single environmental bogeyman.
The honest answer to "is autism really increasing" is: probably a little bit, but mostly we're just better at finding it.
That's where I land. The overwhelming majority of the observed increase is artifactual — broader criteria, better screening, diagnostic substitution, reduced stigma, improved identification in girls and minority populations. A modest true increase is possible but unproven, and if it exists, it's likely driven by changes in known risk factors like parental age and prenatal exposures. Not any single environmental panic.
What I find striking is how this parallels other conditions. ADHD prevalence has gone up. Every generation thinks the next one is sicker, but a lot of it is just that we're looking harder and defining things more broadly.
The diagnostic labels themselves shape the data. When you create a category, people fill it. That doesn't mean the category is invalid — it means we're finally recognizing something that was always there. The question is whether we're overcorrecting. Are we now diagnosing people who don't actually need the label? That's a legitimate debate within the field, but it's a debate about diagnostic thresholds, not about some mysterious epidemic.
One thing I want to circle back to — the global data gap. You mentioned only one study from Africa in the -analysis. That means we have essentially no idea what autism prevalence looks like across an entire continent.
That's not just an academic problem. If you're a parent of an autistic child in Nigeria or Kenya or Ethiopia, you're navigating systems that may not even recognize the diagnosis. The stigma can be enormous. Services may be nonexistent. The global conversation about autism is heavily skewed toward high-income countries with robust diagnostic infrastructure, but the majority of autistic people in the world probably live in places where they'll never receive a diagnosis.
Which means the global prevalence number — one in one thirty-nine — is almost certainly an underestimate, and as those countries develop better screening, the global rate will appear to rise. We'll have this same conversation again in ten years, with even higher numbers.
The conspiracy theorists will have a field day with it. "Look, it's still going up, they must be hiding something." When in reality, it's exactly what you'd expect as diagnostic infrastructure improves worldwide.
If someone's listening and they want to know what to actually do with this information — what's the practical takeaway?
First, if you hear someone cite rising autism numbers as proof of an epidemic or a cover-up, you now know enough to push back with the actual data. The diagnostic criteria have changed dramatically. Substitution explains a huge chunk of the rise. The female diagnosis rate has gone up four hundred fifty percent in some age groups — that's not biology, that's recognition. Second, if you're a parent concerned about your child, the most important thing is early identification and intervention, not chasing environmental panics.
Don't avoid vaccines because of a fraudulent paper from nineteen ninety-eight.
That paper has been retracted for fifteen years. The author lost his medical license. The data was fabricated. And yet it still circulates. It's probably the single most damaging piece of medical fraud in the last century, measured in lives lost to preventable disease. Measles outbreaks in unvaccinated communities are a direct consequence.
What about adults who wonder if they might be on the spectrum? The undiagnosed adult population you mentioned.
If you're an adult and you've always felt like you were working harder than everyone else to navigate social situations, if you've been called rigid or obsessive, if sensory issues have been a lifelong thing — it may be worth exploring an evaluation. But I'd also say: a diagnosis is only useful if it opens doors to understanding or support. For some adults, just reading about the experiences of autistic people and recognizing themselves is enough. For others, a formal diagnosis can be life-changing.
The masking point you made about girls and women is worth sitting with. Eighty percent of autistic girls undiagnosed before eighteen. That means there are millions of women walking around who have spent their entire lives camouflaging, probably exhausted, probably wondering what's wrong with them, and the answer is nothing is wrong — they were just never seen.
The mental health toll of that is enormous. Higher rates of anxiety, depression, eating disorders. Autistic women are often misdiagnosed with borderline personality disorder or bipolar disorder before anyone considers autism. It's a systematic failure of recognition.
Alright, let's land this. What's the one thing you want listeners to take away?
The autism prevalence numbers are real in the sense that there are real people behind them who need support and services. But the increase is overwhelmingly about how we define and detect autism, not about some mysterious epidemic sweeping the globe. The data is clear on that. And the people who tell you otherwise — whether they're selling vaccine conspiracy theories or dismissing the reality of autistic people's needs — are not telling you the truth.
Good place to end. Thanks to our producer Hilbert Flumingtop for keeping this show running, and thanks to Modal for powering the pipeline that makes this whole thing work.
This has been My Weird Prompts. Find us at myweirdprompts dot com, and if you want more episodes like this one, subscribe wherever you get your podcasts.
We'll be back soon.
