Why are Autism diagnoses on the rise? The evidence points primarily to broader diagnostic criteria, not to a proven biological 'epidemic'

The number of people diagnosed with autism spectrum disorder (ASD) has grown dramatically in recent decades, but most recent studies suggest that the increase primarily reflects changes in how the condition is identified and named, rather than a real increase in its biological incidence, according to the comprehensive analysis by Eric W. Dolan in PsyPost , which reviews 14 studies published between 2011 and 2026.

Diagnoses on the rise, but what is actually being measured?

Based on Dolan's review, the figures often cited as evidence of an epidemic measure, in almost all cases, clinical labels, not a stable biological count. The national survey reviewed by Yan and colleagues in the Journal of Autism and Developmental Disorders places that diagnosis at about 4% of U.S. children and adolescents at the end of the 2013–2022 series. Since the data is self-reported by parents and only about half of households responded, it cannot be interpreted as a screening of the entire population.

The international average is not constant either. Zeidan and colleagues, in Autism Research, place the median at around 1%, but the range across studies varies from 1.09 to 436 cases per 10,000 inhabitants. They attribute this variation, a factor of about 400, to differences in how cases are identified and to the country, not to biological differences. Talantseva and others, in Frontiers in Psychiatry, arrive at a slightly lower combined prevalence yet still observe the figure rising over the years. The measurement tool is just as important as the phenomenon itself. When medical records are actively reviewed, the figure reaches 1.22%. When insurance databases are consulted, it drops to 0.35%.

Primary care records highlight the contrast. Russell and colleagues, in the Journal of Child Psychology and Psychiatry, estimate a 787% increase in new diagnoses in England and Northern Ireland between 1998 and 2018, nearly nine times the initial figure. If the sole driver were earlier detection, the average age of diagnosis should decrease. The opposite is true: it has risen from 9.6 to 14.5 years.

Who is being diagnosed now?

The change is not only in volume but also in who is receiving the diagnosis. Russell and colleagues show that autism is no longer concentrated among young children: in British registries, the largest increase in new diagnoses is among adults and women.

The Swedish registry points in the same direction, with a nuance regarding coverage. Fyfe and colleagues, in The BMJ, tracked the records of 2,756,779 children born between 1985 and 2020, and excluded about 26% because neither parent was born in Sweden. Among minors, boys continue to outnumber girls. The ratio reverses later: in 2020–2022, among those aged 15 to 24, there were more diagnoses among females than males, and at older ages the numbers even out. Cumulative parity is not observed across the entire life course. The study projects it to be achieved by age 20 in 2024.

Female "catch-up" or a lowering of the diagnostic threshold?

There is no single interpretation of this change. Fyfe and colleagues view it as a catch-up: girls may have been underdiagnosed, and clinical practice is correcting this bias.

Ranjan and Breunig, in the Journal of Health Economics, test a similar idea in another system. They study Australia's National Disability Insurance Scheme, where public funding is tied to diagnosis, and one of the authors took an unpaid leave of absence from the agency that administers it.

The regional effect, measured by differences in differences, was 0.56 percentage points. Projected to the entire country compared to a scenario without the reform, the scheme would account for 32% of the reported increase in prevalence. The impact was greater among men than among women, was concentrated in metropolitan areas, and did not lower the age of diagnosis. Based on this pattern, the authors conclude that the threshold for recognition has lowered, not that the scheme was merely addressing a long-standing disparity affecting girls.

Funding also changes clinical practice. Since 2018, more children diagnosed with autism have been enrolled in the program than children whose diagnosis was issued by a professional enrolled in Medicare. An official from the program described a model in which private diagnoses are linked to the program in exchange for ensuring the subsequent provision of services.

Broader criteria and diagnostic substitution

The sharp expansion of diagnostic criteria occurred in the 1990s, with the inclusion of Asperger's syndrome in the DSM-IV, the Diagnostic and Statistical Manual of Mental Disorders that defines what constitutes a psychiatric diagnosis. That change accounts for about a quarter of the increase recorded in California.

In British Columbia, up to one-third of new diagnoses between 1996 and 2004 resulted from the replacement of previous labels, such as intellectual disability. The shift to a single spectrum in the DSM-5, the 2013 edition that eliminated Asperger's syndrome as a separate category and merged it with other forms of autism, did not drive the increase: several studies associate it with decreases or no change in the number of diagnoses.

However, in Australia, when autism diagnoses rose following new funding, diagnoses of intellectual disability also increased, indicating that this is not a simple substitution but rather a net expansion of the diagnosis.

When the entire population is screened, the prevalence stabilizes

Kim and others, in JAMA Pediatrics, screened 62,081 children entering elementary school in a single South Korean city using the same method, across 12 birth cohorts. Only about 42% of those who tested positive completed the clinical evaluation. Researchers used machine learning models to estimate the rest.

Those who met the criteria accounted for about 2–3%, above the global median of about 1%, because the screening identified children who had never been seen in a clinic. That proportion did not change significantly across cohorts. What did increase was the shift from the undiagnosed group to the already-diagnosed group. The study therefore points to detection, not to a rising biological incidence.

Milder cases, less intellectual disability

Beyond the increase in the number of cases, the composition of the population diagnosed with autism has undergone a profound transformation. In an article published at Psychological Medicine, Uta Frith notes that, according to data from Sweden, the proportion of people diagnosed with autism who also had an intellectual disability fell from more than 55% in 2001 to less than 7% in 2020.

This trend is consistent with what has been observed in other parts of the world. A review by Lyall and others in Annual Review of Public Health notes that, historically, about 70% of autism cases were accompanied by intellectual disability, but by 2012 that figure had dropped to approximately 30%. The authors emphasize that much of the increase recorded in the previous decade was concentrated in milder cases, while the prevalence of autism associated with intellectual disability remained largely unchanged.

A study conducted in South Korea by Kim and colleagues provides a complementary example. According to the PsyPost, among children who had already been diagnosed through traditional medical services, 33% had an intellectual disability; in contrast, among previously unidentified cases detected through population-based screening, only 6% had one. This contrast suggests that, as the search expands beyond the usual clinical system, many cases with milder cognitive impairment are included in the count, which substantially alters the profile of the diagnosed population.

Diagnostic shift and risk of overdiagnosis in adults

The shift toward milder cases has changed the public meaning of the term. Uta Frith argues that the concept of autism has become semantically unstable: neurodiversity reduced stigma, and online communities transformed it into a cultural identity. She describes a feedback loop, whereby more people interpret their everyday difficulties through this framework, and rejects the notion that the historical criterion was limited to male children.

Croce and Fusaro, in Frontiers in Psychiatry, warn of a dual risk in adult clinics. Underdiagnosis in women and minorities is a reality, and recognizing the female phenotype has been a significant advance over the past fifteen years. At the same time, they point to a possible overdiagnosis when the clinical picture is better explained by a personality disorder or complex trauma. Their distinction between a Type I, which has childhood onset and a greater genetic component, and a Type II, which is milder and overlaps with other mental health issues, is not included in the diagnostic manuals. Social camouflage is not specific either: Milner shows that it also appears in people with high autistic traits who have not been diagnosed. An isolated self-report may, therefore, lead to a mislabeling.

Biological risk factors: Real, but with limited population-wide impact

A risk factor can only increase population incidence if exposure becomes more frequent. Wang and Wang estimate the heritability of autism to be between 40% and 90%, while Love and colleagues place it at around 50%.

According to Dolan's review, advanced parental age, maternal obesity and gestational diabetes are associated with increased individual risk and have risen during the same decades in which diagnoses have increased, so they could contribute to a modest biological increase.

Vaccines are not part of this explanation: Lyall and colleagues find no association in reviews covering 67 studies. As for pollutants, Duque-Cartagena and colleagues find modest risks, but they rate the overall certainty of this evidence as low or very low.