Autism spectrum disorder (ASD) is a neurodevelopmental condition that emerges in early childhood and is characterised by persistent deficits in social communication alongside restricted, repetitive interests and behaviours. The disorder frequently co-occurs with other psychiatric conditions, persists throughout the life course, and imposes substantial socioeconomic burdens on education, employment and family functioning (1).

Global estimates generated with hierarchical Bayesian models show that from 1990 to 2019 the absolute number of people living with ASD increased by more than 50%, whereas the global age-standardised prevalence rate (ASPR) remained essentially unchanged—a pattern of rising case counts despite stable rates (2). This “latent expansion” suggests that improved diagnostic reach has not been matched by commensurate service provision and that rate-based surveillance alone may underestimate demand.

According to GBD 2021, an estimated 61.8 million people (approximately 1 in 127 globally) were living with ASD in 2021. Adolescents and young adults (AYA; 15–39 years) account for about 40% of this total, and ASD ranks among the leading causes of disability in this age group. While high-SDI regions currently have the highest ASPR and the age-standardised disability-adjusted life-year (DALY) rate (ASDR), middle-SDI regions exhibit the fastest growth in cases. In contrast, low-SDI settings likely underestimate the ASD burden because of underdiagnosis and limited resources (3).

Electronic health record data from twelve large U.S. health systems further reveal a 450% surge in diagnosed ASD among adults aged 25–34 years between 2011 and 2019—the steepest increase observed across all age strata (4). Longitudinal cohort studies likewise point to a second peak of functional loss around age 30, forming an inverted U-shaped trajectory across the life course (5). Together, these findings underscore a dual vulnerability during the AYA period, in which major social role transitions coincide with gaps in service provision. However, the spatiotemporal burden in this subgroup remains poorly quantified. Previous global assessments have focused primarily on children aged 0–14 years, leaving the epidemiological profile and service needs of AYA populations systematically underestimated (6).

To address these evidence deficits, we used fully updated GBD 2021 data and methods to quantify ASD prevalence and DALYs among 15–39-year-olds in 204 countries and territories from 1990 to 2021, assessed trends using the Estimated Annual Percentage Change (EAPC), and evaluated cross-country absolute inequality (Slope Index of Inequality, SII) and relative inequality (Concentration Index, CIX) alongside stochastic frontier analysis (SFA) to compute gaps between observed and frontier-attainable ASDR. By identifying high-risk populations and unexploited efficiency gains, our study aims to inform evidence-based screening, intervention and social protection strategies grounded in lifelong support and neurodiversity inclusion, thereby advancing the Lancet Commission on Autism’s call for disability-inclusive societies by 2030 (1).

ASD has become one of the most closely watched neurodevelopmental conditions worldwide. Although incidence has reportedly plateaued – or risen only modestly – in several high-income countries in recent years (15), geographical disparities remain pronounced (2). For AYA, the demands of education, employment and social integration converge with a high prevalence of psychiatric comorbidity (16–18). By providing a comprehensive description of the AYA ASD burden across 204 countries and territories from 1990 to 2021, our study provides evidence to inform needed to refine clinical management, epidemiological surveillance and public health strategy.

Over the past 32 years, the absolute number of AYA ASD cases and DALYs increased by approximately 38%, whereas the ASPR and ASDR fluctuated by less than 0.10% per year – a typical pattern of rising absolute burden despite a stable underlying risk. In 2021, the male ASPR and ASDR were approximately 2.10-fold higher than those observed in females, mirroring the latest systematic review of clinical diagnoses. This disparity is likely multifactorial, arising from a complex interplay of biological susceptibility and socio-diagnostic factors. On a biological level, emerging experimental evidence suggests that prenatal exposure to endocrine disruptors may heighten male neural vulnerability by suppressing aromatase activity, a pathway that may be partially reversible (19). Conversely, the phenomenon of ‘social camouflaging’—where females with ASD more frequently mask their social communication difficulties—often leads to underdiagnosis or delayed diagnosis, thereby artificially inflating the male-to-female ratio in prevalence estimates (20). Furthermore, historical diagnostic criteria have been predominantly built upon male presentations of ASD, potentially lacking sensitivity to the female phenotype. This underscores the critical need for sex-specific screening tools and support strategies to ensure timely and accurate diagnosis for all individuals.

SDI-stratified analyses revealed a clear gradient in 2021 – high SDI > high-middle SDI > middle SDI ≈ low-middle SDI > low SDI. High SDI settings recorded the highest ASPR (1090.72 per 1,00,000) and ASDR (205.00 per 1,00,000) yet showed virtually no decline over 32 years. Conversely, low SDI regions exhibited the lowest corresponding rates (845.15 and 158.57 per 1,00,000, respectively) but experienced a surge in absolute case counts, driven by large birth cohorts and expanding diagnostic infrastructure. Achieving precision across “sex, place and wealth” therefore demands tailored priorities: high-SDI countries should pivot from expanding detection to enhancing post-school education and employment support; middle- and low-SDI regions should embed community- and school-based screening and evidence-based parent training within primary care systems; and all jurisdictions should adopt a neurodiversity-affirming framework that places quality of life at the centre (21).

By 2021, India (4.53 million cases), China (3.12 million) and the United States (1.27 million) together accounted for more than one-third of global prevalent cases, whereas several high-income, small-population nations – Japan, the Republic of Korea and Singapore – topped the world in age-standardised rates (ASPR > 1,600 per 1,00,000; ASDR > 300 per 1,00,000). More than 90% of countries posted slight increases in these rates, but the pace of change diverged sharply: the Maldives, Equatorial Guinea and Qatar recorded the largest positive EAPCs, whereas the Cook Islands, Kuwait and the Syrian Arab Republic showed declines. The coexistence of “high rate/low population” and “low rate/high population” profiles means that resource allocation must balance per-capita service strain against absolute caseload (22). High-income nations, despite rapid case detection, have not expanded post-school support at a pace sufficient to curb ASDR; middle- and low-income settings, still building diagnostic systems, display “low rates but swelling absolute numbers” (1, 23). Policy responses should therefore differ: high-rate, small-population countries need stronger vocational and social inclusion programmes; populous nations should mainstream community screening and digital parent training; and fast-growing small states should form regional alliances to scale up low-cost early intervention models (24, 25).

Between 1990 and 2021, the peak functional loss attributable to ASD shifted from the 15–24-year age group to the 30–39-year group, echoing meta-analytic findings of increasing overall prevalence amid flat age-standardised rates (26). United States insurance and electronic medical record data show a dramatic rise in diagnoses among 26–34-year-olds between 2011 and 2022 (27); European and North American sentinel sites likewise document a late-youth surge in reporting without a parallel expansion of services, producing a “diagnosis first, support lagging” gap (28). Extending routine screening to include individuals at ages 25 and 39 and building lifelong stepped-care packages – covering vocational training, mental health care and comorbidity management – will be essential to offset the DALY plateau after age 30.

Previous global studies have focused on children or all-age populations, leaving the AYA “second peak” under-examined. Although Zeidan et al. estimated childhood ASD prevalence at approximately 1% and Salari et al. confirmed a rising overall burden, neither study disaggregated outcomes for the 15–39-year age range (29, 30). Leveraging GBD 2021, our analysis is the first to map AYA ASPR, ASDR and EAPC across 204 countries, effectively creating a cross-national map of AYA rates and trends to guide policy (i.e., highlighting the mismatch between “high rate/low population” and “low rate/high population” profiles across settings).

Using both SII and CIX, we found that adolescent- and young-adult ASD DALYs have remained disproportionately concentrated in high-SDI settings since 1990: the absolute gap rose only from 20.30 to 22.53 per 1,00,000 and overlapping UIs indicate no true narrowing, while the CIX stayed at approximately 0.04, confirming a persistent tilt of diagnostic and support systems toward affluent nations—echoing earlier GBD and systematic review evidence across all ages and in children (31). Frontier modelling reinforced this picture: although the “best-attainable” curve shifted slightly downward with increasing SDI, several high-income countries (e.g., Japan, Republic of Korea, Singapore, Australia, Canada, United States) still lay well above the frontier in 2021, whereas low-SDI economies such as Bangladesh and Somalia sat on or below it, a pattern that likely reflects under-ascertainment rather than genuinely lower burden (32, 33). This high-rate/high-resource mismatch implies that economic growth alone will not reduce ASD disability; adult screening, employment support and community interventions must be embedded within universal health coverage, while reinforced primary-level surveillance and parent training in low- and middle-SDI countries are needed to uncover hidden cases and achieve true global convergence. We adopted a half-normal composite-error specification to enforce non-negativity of inefficiency; in sensitivity analyses, exponential and truncated-normal alternatives yielded unchanged qualitative inferences.

Because GBD 2021 does not attribute ASD to specific risk factors, we synthesised high-quality evidence published in the past 5 years. For instance, gestational diabetes is associated with an approximately 40% higher risk of ASD in offspring (34); high PM₂.₅/NO₂ exposure is linked to ~31% higher risk (35); adherence to a Mediterranean diet is associated with about a 23% lower risk (36); and research on the BPA–aromatase–ASD pathway (with partial reversibility) suggests new targets for mitigating chemical exposures (19). However, important limitations of these studies include the concentration of cohorts in high-income countries, heterogeneous exposure windows and doses, and inconsistent case ascertainment. In addition, our outcome is ASD-attributable DALYs: psychiatric comorbidities common in AYA (e.g., anxiety and depression) are assigned to those disorders under GBD attribution rules and thus are not included in ASD DALYs. Consequently, our ASD estimates likely understate real-world burden in adolescents and young adults, consistent with recent evidence documenting high comorbidity rates in autistic populations (37). Precise risk attribution will require establishing longitudinal consortia in the global South that integrate environmental and biological data, enriched with multi-omics markers to support multi-layered prevention.

Between 1990 and 2021, the global absolute burden of ASD among adolescents and young adults increased by approximately 40%, whereas age-standardised rates showed little change, indicating that this rise reflects demographic growth and improved detection rather than increased individual risk. Notably, frontier-adherent or near-frontier ASD rates observed in some low-SDI settings should be interpreted primarily as signals of under-ascertainment and diagnostic/surveillance gaps rather than truly low underlying burden (23). The persistent male predominance (rate ratio ≈2.1) and pronounced rise in disability in the 30–39 age group highlight ongoing and emerging challenges across the lifespan. Given these disparities and significant cross-national inequalities (SII 22.53; CIX 0.04), context-sensitive strategies are needed. High-SDI countries could integrate adult screening, vocational support, and community-based care into health coverage; low- and middle-SDI regions may prioritize enhanced surveillance, school-community screening, and parent-training to improve identification. Adopting neurodiversity-affirming approaches that emphasize lifelong support and social integration is widely encouraged. Reducing global disparities in ASD care will require more efficient resource use in high-capacity settings and sustained investment in foundational capacity in underserved areas.