Understanding Autism Prevalence: Why Diagnosis Rates Have Risen From 1 in 150 to 1 in 31 and What the Research Actually Shows

In April 2025, the Centers for Disease Control and Prevention released its latest autism prevalence data, reporting that an estimated 1 in 31 eight-year-old children in the United States had been identified with autism spectrum disorder (ASD) — up from 1 in 36 in the previous report (based on 2020 data) and a dramatic increase from the CDC’s first estimate of 1 in 150 in 2000. The numbers have triggered sharply different reactions: HHS Secretary Robert F. Kennedy Jr. described an “autism epidemic,” while researchers and autism advocacy organizations pointed to expanded diagnostic criteria, improved screening, and greater awareness as the primary drivers of the increase.

This article examines what the CDC data actually show, why reported autism rates have risen so dramatically over the past 25 years, what researchers know about the causes of autism, and what the evidence says about the factors most commonly cited in public debate.

What the CDC Data Show

The CDC monitors autism prevalence through the Autism and Developmental Disabilities Monitoring (ADDM) Network, a surveillance system that has been collecting data since 2000. The ADDM Network operates across 16 sites in different states, reviewing health and education records of eight-year-old children. Researchers chose age eight because most children with autism have received a formal evaluation and diagnosis by that age.

The 2025 report (based on 2022 surveillance data) found:

Overall prevalence: 1 in 31 (3.2%) of eight-year-old children identified with ASD, up from 1 in 36 (2.7%) in 2020 and 1 in 44 (2.3%) in 2018.

Gender disparity: Boys were 3.4 times more likely than girls to be diagnosed, though this ratio has narrowed from approximately 4.5:1 in earlier reports, suggesting that autism in girls is being identified more effectively than in the past.

Racial and ethnic disparities have narrowed dramatically. Historically, white children were identified at significantly higher rates than children of other races. In the 2022 data, prevalence was highest among American Indian/Alaska Native children (1 in 27), Asian/Pacific Islander children (1 in 26), and Black children (1 in 27), compared to white children (1 in 36). This reversal of historical patterns strongly suggests that improved screening and reduced barriers to diagnosis in previously underserved communities are contributing to the overall prevalence increase.

Age at diagnosis: The median age of first evaluation was 36 months, and the median age of ASD diagnosis was 47 months (just under 4 years old). Only about half of children were diagnosed by age 3, indicating continued room for improvement in early identification.

Why Reported Rates Have Increased

The most important question about autism prevalence data is whether the increase reflects a genuine rise in the number of people who are autistic, or whether it reflects changes in how autism is defined, diagnosed, and identified. The scientific consensus, supported by multiple lines of evidence, is that the majority of the observed increase is explained by non-biological factors:

1. Expanded Diagnostic Criteria

The definition of autism has broadened significantly over the past 40 years. In the DSM-III (1980), autism was defined narrowly as “infantile autism,” requiring severe social withdrawal, language deficits, and restricted behaviors with onset before 30 months. The DSM-III-R (1987) expanded the criteria and renamed it “autistic disorder.” The DSM-IV (1994) added Asperger’s disorder and PDD-NOS (pervasive developmental disorder not otherwise specified) as separate diagnoses on the autism spectrum. The DSM-5 (2013) consolidated all of these into a single “autism spectrum disorder” diagnosis with varying severity levels.

Each expansion of the diagnostic criteria brought previously undiagnosed individuals into the autism category. People who would have been diagnosed with “mental retardation,” “childhood schizophrenia,” “language disorder,” or simply considered “quirky” or “socially awkward” under earlier diagnostic systems are now recognized as being on the autism spectrum. Studies comparing DSM-IV and DSM-5 criteria have found that the DSM-5 captures a broader population, though the exact extent of the difference varies by study.

2. Diagnostic Substitution

Multiple studies have documented “diagnostic substitution” — the phenomenon in which an increase in autism diagnoses is accompanied by a corresponding decrease in diagnoses of other conditions. A widely cited 2006 study by Shattuck published in Pediatrics found that much of the increase in autism prevalence in U.S. special education data during the 1990s was accompanied by a decline in diagnoses of intellectual disability, speech/language impairment, and emotional disturbance. In other words, many children who would previously have received a different label were now being identified as autistic — not because their condition had changed, but because the diagnostic category had expanded to include them.

3. Increased Awareness and Screening

Public awareness of autism has increased dramatically since the 1990s. The American Academy of Pediatrics (AAP) first recommended universal developmental screening at well-child visits in 2006, with specific autism screening recommended at 18 and 24 months. This recommendation ensured that millions of children who might otherwise never have been evaluated for autism were now being screened as part of routine pediatric care.

The closing of racial and ethnic disparities in the 2022 ADDM data provides strong evidence for the role of awareness and access. The fact that autism identification rates in Black, Hispanic, and Asian/Pacific Islander children have risen faster than in white children — to the point where non-white children now have higher identification rates in several ADDM sites — suggests that improved outreach and screening in previously underserved communities is bringing children into the diagnostic system who were always autistic but had not previously been identified.

4. Changes in Educational Policy

The Individuals with Disabilities Education Act (IDEA) added autism as a separate disability category for special education services in 1991. This gave schools a financial and administrative incentive to identify children with autism, because autism-specific diagnoses could trigger access to specialized services and funding. Before 1991, children with autism who needed educational support were classified under other disability categories. After 1991, many of these children were reclassified as autistic, contributing to the observed prevalence increase in educational data.

What Actually Causes Autism

Autism is a neurodevelopmental condition with strong genetic underpinnings and complex environmental contributions. The current scientific understanding includes:

Genetics. Twin studies consistently show that autism has high heritability — estimated at 64–91% in large population-based studies. Hundreds of genes have been associated with autism risk, including both common genetic variants (each contributing a small increase in risk) and rare mutations (which individually confer higher risk). No single “autism gene” exists; rather, the condition appears to result from the interaction of many genetic variants, each affecting brain development in subtle ways.

Advanced parental age. Both advanced maternal and paternal age at conception are associated with increased autism risk, a finding replicated across many studies and populations. The mechanism is thought to involve increased rates of de novo (spontaneous) genetic mutations in the sperm and eggs of older parents.

Prenatal factors. Research has identified several prenatal exposures associated with modestly increased autism risk, including maternal infections during pregnancy, prenatal exposure to certain medications (notably valproic acid, used to treat epilepsy and bipolar disorder), extreme premature birth, low birth weight, and complications during delivery that result in oxygen deprivation. Prenatal exposure to air pollution and certain environmental chemicals (phthalates, organophosphate pesticides) has been associated with autism risk in some epidemiological studies, though the evidence is less consistent than for genetic factors.

What does NOT cause autism. The most exhaustive body of evidence in autism research concerns the question of whether vaccines cause autism. The answer, supported by dozens of large-scale epidemiological studies involving millions of children across multiple countries, is definitively no. A 2019 study published in Annals of Internal Medicine, following over 650,000 Danish children, found no association between the MMR vaccine and autism — consistent with the findings of every other rigorous study conducted since the original (and retracted) 1998 Wakefield paper. The Wakefield paper was retracted by The Lancet in 2010 after an investigation found that the data had been falsified, and Wakefield’s medical license was revoked.

The “Epidemic” Framing: Why It Matters

Whether rising autism prevalence is described as an “epidemic” or as “improved identification” has significant policy implications. The epidemic framing implies that something in the environment is causing more people to become autistic, which directs research funding and public attention toward identifying and eliminating environmental causes. The improved-identification framing implies that autistic people have always existed in similar numbers but are now being recognized, which directs attention toward improving diagnostic access, educational services, and support for autistic individuals and their families.

The truth likely includes elements of both. The majority of autism researchers agree that expanded diagnostic criteria, increased awareness, and improved screening account for most of the observed prevalence increase. However, most researchers also acknowledge that a genuine increase in autism incidence — beyond what diagnostic changes alone would explain — cannot be ruled out based on current evidence. Disentangling diagnostic artifact from true incidence change is methodologically extremely difficult, because there is no objective biological test for autism; diagnosis depends entirely on behavioral observation and clinical judgment, both of which are influenced by the cultural and diagnostic context.

The 2025 CDC report’s own authors noted that the increase in prevalence “may reflect several factors including greater awareness, improved screening and diagnostics — especially in communities that were previously underrepresented.” The Autism Society of America, in its response to the report, emphasized that prevalence data should “drive equity and access — not fear, misinformation, or political rhetoric.”

What the Data Mean for Families and Policy

Regardless of whether the prevalence increase is driven primarily by better identification or by a genuine rise in incidence, the practical implications for families and policymakers are the same: more children are being diagnosed with autism, and the demand for diagnostic services, early intervention, educational support, and adult services is growing faster than the infrastructure to provide them.

Wait times for diagnostic evaluations average 3–6 months in many parts of the country and can exceed 12 months in underserved areas. Early intervention services — which research shows are most effective when started before age 3 — are often unavailable or underfunded. The Autism CARES Act, reauthorized in 2024 with nearly $2 billion in funding over five years, supports prevalence tracking, early screening, clinical best practices, and community-based interventions, but the scale of funding remains insufficient relative to the growing identified population.

For families navigating the system, the CDC recommends learning developmental milestones through its “Learn the Signs. Act Early.” program, requesting developmental screening at well-child visits, seeking evaluation promptly if concerns arise, and connecting with state early intervention programs (available in all states for children under 3) as early as possible. The evidence consistently shows that earlier identification and earlier access to support services produce better outcomes — which means that the ongoing increase in prevalence, to the extent it reflects better identification of children who need support, is a public health success rather than a crisis.

The California Data in Context

California has been a focal point in the autism prevalence debate because the state maintains one of the most comprehensive developmental disability tracking systems in the country through its Department of Developmental Services (DDS). The DDS data, which count the number of individuals receiving autism-related services, have shown steep increases that are sometimes cited as evidence of an environmental cause.

Between 2012 and 2018, the number of individuals with autism receiving DDS services in California increased by approximately 50%. This increase coincided with several changes that make a straightforward environmental interpretation problematic. California expanded its Medi-Cal (Medicaid) program under the Affordable Care Act in 2014, providing health coverage to millions of previously uninsured residents and improving access to diagnostic evaluations. The state also invested significantly in early intervention services and autism screening programs during this period. And the DSM-5, published in 2013, consolidated several previously separate diagnoses (including Asperger’s syndrome and PDD-NOS) under a single autism spectrum disorder category, expanding the diagnostic net.

A 2019 study by Nevison and Zahorodny published in the Journal of Autism and Developmental Disorders analyzed California DDS data and concluded that a substantial portion of the increase could not be explained by diagnostic changes alone, arguing that a true increase in autism incidence was occurring. However, other researchers have challenged this analysis, noting that the DDS data measure service utilization rather than prevalence (a family must actively seek and receive services to appear in the count), and that changes in service availability, eligibility criteria, and awareness directly affect the numbers independent of any change in actual autism incidence.

The most rigorous attempt to disentangle diagnostic artifact from true incidence change in California was a 2020 study by Keyes et al. published in American Journal of Epidemiology, which used birth cohort analysis to examine whether autism rates increased even within groups born in the same year (who would have been exposed to the same environmental conditions). The study found that most of the increase in autism diagnoses could be attributed to period effects — changes in how autism was diagnosed and identified that affected all birth cohorts simultaneously — rather than cohort effects that would suggest a generation-by-generation increase in actual autism incidence. The authors concluded that “the apparent rise in ASD prevalence is largely attributable to changes in diagnostic practices.”

Environmental Factors: What the Research Shows

While the majority of the observed prevalence increase is explained by diagnostic and methodological changes, the question of whether environmental factors contribute to autism risk is a legitimate and active area of scientific research. It is important to distinguish between two very different questions: (1) whether environmental factors contribute to autism risk at the individual level (the answer is yes, for some factors), and (2) whether environmental changes are driving a population-level increase in autism incidence (the evidence for this is much weaker).

Factors With Established Associations

Prenatal air pollution. Multiple large-scale epidemiological studies have found associations between maternal exposure to traffic-related air pollution during pregnancy and modestly increased autism risk in offspring. A 2019 meta-analysis by Chun et al. in Environment International, combining data from 25 studies, found a statistically significant association between prenatal exposure to particulate matter (PM2.5) and nitrogen dioxide (NO2) and autism risk. The effect sizes were modest (typically 10–30% increased risk for higher exposure levels), and the studies could not fully rule out residual confounding from socioeconomic factors correlated with pollution exposure.

Prenatal pesticide exposure. Studies of populations living near agricultural areas with high pesticide use have found associations between prenatal organophosphate exposure and increased autism risk. The CHARGE (Childhood Autism Risks from Genetics and Environment) study, conducted in California’s agricultural Central Valley, found that mothers living within 1.5 kilometers of organophosphate applications during pregnancy had a 60% higher risk of having a child diagnosed with autism. However, the study design (case-control, with retrospective exposure assessment) limits causal inference, and the findings have not been consistently replicated in other populations.

Prenatal medication exposure. Valproic acid (an anticonvulsant used to treat epilepsy and bipolar disorder) is the most established pharmacological risk factor for autism. A 2013 study by Christensen et al. in JAMA found that prenatal valproic acid exposure was associated with a 4.4-fold increase in autism risk. This is one of the strongest and most consistently replicated environmental associations in autism research, and it has informed clinical guidelines recommending that women of childbearing age use alternative medications when possible.

Maternal immune activation. Maternal infections during pregnancy, particularly those accompanied by high fever, have been associated with increased autism risk in some studies. The biological mechanism is thought to involve inflammatory cytokines crossing the placenta and affecting fetal brain development. A 2019 Danish population-based study found that maternal hospitalization for infection during pregnancy was associated with a 12% increase in autism risk. However, the association was modest and did not account for many potential confounding variables.

Factors With Insufficient or Negative Evidence

Vaccines. As noted above, the vaccine-autism hypothesis has been investigated more thoroughly than almost any other question in modern epidemiology. The evidence is unequivocal: vaccines do not cause autism. This conclusion is supported by studies involving millions of children across multiple countries, including Denmark (650,000+ children), Finland (95,000+), the UK (500,000+), Japan (30,000+), and the United States (multiple studies). The original 1998 paper by Andrew Wakefield that suggested a link between the MMR vaccine and autism was retracted by The Lancet in 2010 after an investigation found serious ethical violations and data fabrication. Wakefield’s medical license was revoked by the UK General Medical Council.

Thimerosal. Thimerosal, an ethylmercury-based preservative that was used in some childhood vaccines (but never in the MMR vaccine), was removed from virtually all childhood vaccines in the United States by 2001 as a precautionary measure. If thimerosal had been contributing to autism rates, prevalence should have declined or stabilized after its removal. Instead, autism prevalence continued to rise at the same rate — strong evidence against a causal role for thimerosal.

Glyphosate. Claims that the herbicide glyphosate (Roundup) causes autism have circulated widely online, based largely on a 2014 paper by Swanson et al. that plotted rising glyphosate use alongside rising autism diagnoses and found a statistical correlation. However, correlation between two rising time trends does not indicate causation (during the same period, organic food sales, smartphone ownership, and streaming video subscriptions also rose steeply). No peer-reviewed study has established a biological mechanism by which glyphosate exposure could cause autism, and epidemiological studies of agricultural workers with high glyphosate exposure have not found elevated autism rates in their children.

The Genetics of Autism: What Twin and Family Studies Show

The strongest evidence regarding autism’s causes comes from twin and family studies, which consistently show that genetics play a dominant role. A 2019 meta-analysis by Tick et al. in the Journal of Child Psychology and Psychiatry analyzed data from 6,413 twin pairs across seven studies and estimated autism heritability at 64–91%, with the best estimate around 80%. This means that approximately 80% of the variation in autism risk between individuals is attributable to genetic factors.

The genetic architecture of autism is complex. Genome-wide association studies (GWAS) have identified over 100 common genetic variants associated with autism risk, each contributing a small increase in probability. In addition, rare genetic mutations — including copy number variants (CNVs) and de novo point mutations — account for an estimated 10–20% of autism cases. These rare mutations are more common in children born to older parents, which partly explains the association between advanced parental age and autism risk.

Importantly, the high heritability of autism does not mean that the condition is “purely genetic.” Heritability is a population-level statistic that describes how much of the variation in a trait is attributable to genetic differences in a specific population and environment. If environmental exposures change (for example, if a new widespread toxic exposure emerges), the environmental contribution to autism risk could increase even while heritability remains high. The genetic evidence tells us that autism has always been part of human neurodiversity, but it does not rule out the possibility that environmental changes are modifying risk at the margins.

The Neurodiversity Perspective

The scientific and policy debates about autism prevalence occur alongside a parallel cultural conversation about how autism is understood. The neurodiversity movement, which has gained significant influence since the 2010s, frames autism not as a disease to be cured or an epidemic to be contained, but as a form of natural human neurological variation — a difference in how the brain processes information, rather than a deficit.

Proponents of the neurodiversity framework argue that the “epidemic” framing of rising autism prevalence is harmful because it: (1) pathologizes a neurological difference that has always existed in the human population, (2) directs research funding toward finding causes and cures rather than toward improving services and supports for autistic people, (3) creates stigma and fear around autism that affects how autistic children and adults are treated, and (4) provides fertile ground for pseudoscientific “cures” and treatments that can be harmful.

This perspective does not deny that some autistic individuals require significant support, or that co-occurring conditions (epilepsy, intellectual disability, anxiety, sensory sensitivities) can cause genuine suffering. It argues that these needs are best addressed through accommodation, support services, and societal adaptation rather than through framing autism itself as a problem to be solved.

The Autism Society of America’s response to the 2025 CDC data reflected this perspective, emphasizing that prevalence data should “drive equity and access — not fear, misinformation, or political rhetoric.” The organization advocated for increased investment in early identification, educational support, employment programs, and housing services for autistic individuals, rather than for cause-hunting driven by the assumption that autism is an epidemic requiring containment.

Policy Implications

Regardless of whether one interprets rising prevalence as better identification or a genuine increase, the policy implications are substantial and immediate. The identified autistic population in the United States now includes approximately 5.4 million adults (CDC estimate), many of whom were diagnosed in childhood and have aged into an adult services system that was not designed for them.

Early intervention capacity. Research consistently shows that behavioral interventions (particularly Applied Behavior Analysis, developmental and relationship-based approaches, and speech-language therapy) produce the best outcomes when started before age three. Yet the average age of diagnosis remains nearly four years, and wait times for evaluation can exceed 12 months. Expanding the workforce of psychologists, developmental pediatricians, and speech-language pathologists qualified to diagnose and treat autism is a critical unmet need.

Educational services. Under IDEA, children identified with autism are entitled to a free appropriate public education (FAPE) including individualized education programs (IEPs) and related services. As the identified population grows, school districts face increasing demand for autism-specific services without proportional funding increases. The federal government has never fully funded IDEA at its authorized level (40% of the average per-pupil cost of educating a child with a disability), leaving states and local districts to cover the gap.

Adult services. Perhaps the most pressing challenge is the so-called “services cliff” that autistic individuals face when they age out of the educational system at 21 or 22. Employment rates for autistic adults remain low (approximately 30–40% are employed, compared to 75–80% of the general adult population), and many lack access to the vocational training, supported employment, and community living services that could help them achieve greater independence. The Autism CARES Act (reauthorized in 2024) provides some research and training funding, but does not directly address the service gap.

Insurance coverage. All 50 states now require insurance coverage for autism treatment (a legislative achievement that took nearly two decades to complete), but coverage mandates vary widely in scope, age limits, and dollar caps. Some plans cover only ABA therapy; others cover a broader range of interventions. Out-of-pocket costs for families remain high, particularly for those with high-deductible health plans or those seeking providers outside of narrow insurance networks.

International Comparisons

Autism prevalence estimates vary significantly across countries, but the upward trend is consistent worldwide. The UK’s National Health Service estimates autism prevalence at approximately 1 in 57 among adults and higher among children. South Korea reported a prevalence of 2.6% (1 in 38) in a landmark 2011 community-based study that included direct screening of all children in a school district, rather than relying on existing medical and educational records as the CDC’s ADDM Network does. The South Korean study’s higher rate suggests that administrative data (which the ADDM Network relies on) may undercount autism prevalence compared to direct community screening, implying that even the 1 in 31 figure may be an underestimate.

Japan provides a particularly instructive case. Following the removal of the MMR vaccine from Japan’s immunization schedule in 1993 (due to concerns about the mumps component, not autism), autism prevalence in Yokohama continued to rise at the same rate as before the vaccine was withdrawn. This natural experiment provided powerful evidence against the vaccine-autism hypothesis and demonstrated that the rising prevalence trend was independent of vaccination practices.

Nordic countries, which maintain comprehensive population registries that track diagnoses from birth through adulthood, have been able to study autism prevalence trends over longer periods than the U.S. ADDM Network. Swedish registry data analyzed by Lundstrom et al. (2015) found that the increase in autism diagnoses was accompanied by a proportional decrease in diagnoses of intellectual disability without autism — consistent with diagnostic substitution rather than a true incidence increase. Danish registry data have shown similar patterns, with the overall prevalence of developmental disabilities remaining relatively stable even as the proportion diagnosed specifically with autism has increased.

These international comparisons reinforce the conclusion that the global increase in autism prevalence is driven primarily by changes in diagnostic practices and awareness rather than by environmental factors specific to any one country. If an environmental cause were responsible, prevalence changes would be expected to correlate with specific national exposures rather than increase uniformly across countries with very different environmental profiles.

The Funding and Services Challenge

The practical consequences of rising autism identification are most visible in the gap between the growing identified population and the services available to support them. The Autism CARES Act, reauthorized in 2024 with $1.85 billion in funding over five years, represents the largest federal commitment to autism-specific programs. The legislation funds research through the NIH (including the ADDM Network surveillance system), training programs for healthcare providers, and community-based services through HRSA. Since 2007, over $5.2 billion in autism research and training programs have been funded through this legislation.

However, the per-capita funding relative to the identified population has actually decreased as prevalence has risen. When the original Combating Autism Act was passed in 2006, the estimated autism population was approximately 1.5 million. Today, with prevalence at 1 in 31 among children and an estimated 5.4 million autistic adults, the same level of real-dollar funding is spread across a much larger population. The result is that individual access to services has not kept pace with the expanding identified population.

State-level services vary enormously. Some states (notably California, Massachusetts, and New Jersey) have invested significantly in autism services, including comprehensive early intervention programs, school-based support, and adult services. Others provide minimal services beyond what federal law requires. The National Council on Disability has called the state-by-state variation in autism services “the most significant equity issue facing the autism community,” noting that the quality and availability of services a child receives depends more on where they live than on their individual needs.

The economic costs of inadequate services are substantial. A 2024 study published in JAMA Pediatrics estimated the lifetime cost of supporting an individual with autism at $1.4–$2.4 million (depending on the level of intellectual disability), with the majority of costs attributable to adult services, lost productivity, and residential care. Early intervention, while expensive in the short term ($20,000–$50,000 per child per year for intensive behavioral therapy), has been shown to reduce long-term costs by improving functional outcomes and reducing the need for lifelong support services.

References

1. Shaw KA, et al. Prevalence and Early Identification of Autism Spectrum Disorder Among Children Aged 4 and 8 Years — ADDM Network, 16 Sites, United States, 2022. MMWR Surveill Summ. 2025;74(No. SS-2):1–22.
2. CDC. Autism and Developmental Disabilities Monitoring (ADDM) Network. cdc.gov
3. Hviid A, et al. Measles, Mumps, Rubella Vaccination and Autism: A Nationwide Cohort Study. Ann Intern Med. 2019;170(8):513–520.
4. Modabbernia A, Velthorst E, Reichenberg A. Environmental risk factors for autism: an evidence-based review of systematic reviews and meta-analyses. Molecular Autism. 2017;8:13.
5. Shattuck PT. The Contribution of Diagnostic Substitution to the Growing Administrative Prevalence of Autism in US Special Education. Pediatrics. 2006;117(4):1028–1037.
6. Autism Society of America. Response to CDC Autism Prevalence Report. April 15, 2025. autismsociety.org
7. Autism Speaks. Autism Prevalence Rises to 1 in 31 Children. April 2025. autismspeaks.org
8. Tick B, et al. Heritability of Autism Spectrum Disorders: A Meta-Analysis of Twin Studies. J Child Psychol Psychiatry. 2016;57(5):585–595.
9. The Lancet. Retraction: Ileal-lymphoid-nodular hyperplasia, non-specific colitis, and pervasive developmental disorder in children. Lancet. 2010;375(9713):445.
10. Scientific American. The Real Reasons Autism Rates Are Up in the U.S. February 2024. scientificamerican.com

Last updated: September 26, 2026