Autism diagnoses in the U.S. have risen several‑fold since the 1990s, with most evidence indicating that expanded diagnostic criteria, greater screening, and diagnostic substitution explain a large share of the increase, while a smaller but non‑zero contribution from true incidence change remains possible but not well quantified. Major reviews and surveillance bodies generally conclude there is an “autism diagnosis” epidemic rather than a clearly demonstrated epidemic of underlying neurodevelopmental pathology.[1][2][3][4][5][6]

1. Prevalence trends (1990s–2024)

ADDM also documents shifts in demographics over time, with recent cycles showing higher identified prevalence among Black and Hispanic children compared with White children and attenuation of earlier SES gradients, consistent with improved access to diagnosis in previously under‑identified groups.[10][2]

2. Severity distribution over time

3. Diagnostic expansion vs. true incidence

Expanded diagnostic criteria

• The formal diagnostic category broadened substantially between DSM‑III, DSM‑IV, and DSM‑5, with DSM‑IV (1994) and DSM‑IV‑TR incorporating pervasive developmental disorders and Asperger syndrome, and DSM‑5 (2013) consolidating these under ASD while allowing for a spectrum of severity; epidemiological reviews consistently note that these definitional changes can raise measured prevalence even if underlying biology is unchanged.[5][4]
• Commentaries summarizing CDC data note that the agency’s use of both older and newer diagnostic criteria in some reports (e.g., around 2018–2020) can inflate apparent prevalence by capturing people who meet either set of criteria.[5]

Increased screening and awareness

• The ADDM Network reports document a substantial improvement in early identification: in 2022, at many sites, prevalence among 4‑year‑olds was already equal to or higher than among 8‑year‑olds, and identification by 48 months was 40–300% higher for children born in 2018 than for those born in 2014, indicating intensified early screening and referral.[10]
• Over two decades of ADDM monitoring, the demographic profile of diagnosed children shifted from higher‑SES White over‑representation to higher prevalence among Black and Hispanic children and weaker SES gradients, consistent with rising awareness and access among groups that were previously under‑diagnosed rather than a sudden biologic change restricted to these populations.[6][2]

Diagnostic substitution

• A multilevel analysis of U.S. special education data (1984–2003) found autism administrative prevalence rose from 0.6 to 3.1 per 1,000, while mental retardation and learning disability prevalence declined by 2.8 and 8.3 per 1,000, respectively; the authors concluded that changes in classification and diagnostic substitution accounted for a sizeable part of the apparent autism increase and that special‑education data did not support an “autism epidemic.”[3]
• Other work using federal special education categories similarly found that as autism labels rose, intellectual disability labels fell, supporting diagnostic substitution as one mechanism, though some studies have not observed perfect one‑to‑one substitution.[16][17]
• Recent expert commentary for a public health audience emphasizes that corresponding declines in “intellectual disability” and “learning disability” diagnoses suggest diagnostic substitution rather than a large recent surge in underlying incidence.[3][6]

Actual changes in biological incidence

• Global meta‑analytic work shows rising prevalence across settings and methodologies, but the authors highlight large heterogeneity and conclude that much of the increase is likely due to methodological and diagnostic factors; they do not rule out some contribution from true incidence change but cannot quantify it precisely.[18][4]
• Reviews of environmental risk factors (parental age, prematurity, prenatal exposures) document modest relative risks (often odds ratios in the 1.3–2.0 range), implying that even if exposure prevalence has increased, such factors are unlikely to fully explain the several‑fold rise in measured prevalence, pointing again to the central role of diagnostic and ascertainment changes.[19][4]

Consensus across epidemiological reviews: