News · Brain & Mental Health
Autism diagnoses rose only in women and girls after 2020
Tracking 171,134 patients in a US health system, JAMA Network Open found male autism diagnoses flat or falling while female rates climbed 19% a year.
Based on a peer-reviewed retrospective cohort study in JAMA Network Open
- Researchers tracked newly recorded autism diagnoses from 2016 to 2024 in 171,134 people continuously engaged with a large US health care system, publishing in JAMA Network Open.
- Females were diagnosed a mean 3.4 years later than males: 15.7 years versus 12.3 years.
- Male autism incidence remained stable or declined across the whole period.
- Female incidence rose sharply after 2020 (annual percentage change, 19.0%; 95% CI, 12.1% to 33.4%).
- The split held in children (female APC, 17.4%) and adolescents (female APC, 22.3%), while male pediatric incidence fell slightly.
- In adult females the increase did not reach statistical significance (APC, 8.0%; 95% CI, -0.9% to 20.0%).
- The authors read this as better identification of presentations that were historically missed, not a population-level rise.
- One health system, and only patients with at least one visit every year. That is a selected group, not the general population.
Rising autism numbers are usually reported as a single line going up. A study in JAMA Network Open split that line by sex, and it stopped being one trend and became two, moving in opposite directions.
The authors note that despite the increasing prevalence of autism spectrum disorder (ASD) in the US, underlying temporal trends in newly recorded diagnoses remain poorly understood. Prevalence counts everyone carrying a diagnosis. Incidence counts who is newly receiving one, which is the more informative number and the harder one to get.
How it was counted
The analysis used electronic health record data from a large US health care system, covering 2016 to 2024.
The study population included 171,134 participants stratified as children, adolescents and adults, all of whom maintained continuous engagement with the health care system, defined as having at least 1 clinical encounter annually from 2016 to 2024.
That continuity requirement is deliberate and it cuts both ways. It means a rise cannot be an artifact of people simply joining the system. It also means the population studied is unusually well connected to healthcare, which is not most people.
The delay
Before the trends, a gap. Among the 171,134 participants, females received autism diagnoses a mean 3.4 years later than males - 15.7 years old versus 12.3.
Three and a half years is most of secondary school. It is the period in which social demands escalate sharply and in which an unexplained struggle tends to get labeled as something else.
Two trends, not one
Then the split.
Male autism incidence remained stable or declined 2016 to 2024; female incidence increased significantly post-2020, at roughly 19% a year.
Break it down by age and the same shape recurs. In the pediatric cohort there was decreasing annual incidence among males, alongside increasing incidence for females since 2020, at about 17% a year. In adolescents, female autism incidence increased significantly between 2020 and 2024, at around 22% a year, while adolescent male rates had already declined between 2016 and 2018 and then flattened.
Adults were the exception. The female increase there pointed the same way but was too uncertain to rely on: in adult females, the annual change was positive, with a range that still included no change at all.
Why the asymmetry is the argument
If something were increasing autism itself, it would be strange for it to affect only one sex while the other declined.
That asymmetry is what the authors lean on. These patterns suggested that increasing ASD prevalence reflected improved identification of historically underrecognized presentations rather than a uniform population-level increase.
It also sits neatly alongside separate work in Denmark finding that the average genetic liability of newly diagnosed people has fallen over time. Both point at recognition rather than incidence, from completely different directions.
What the design cannot do
This is one health care system. Diagnostic practice varies enormously between systems, and a single organization adopting new screening guidance could produce much of a trend like this on its own.
The continuous-engagement criterion also selects a particular kind of patient: insured, stable, in regular contact with clinicians. People whose access is intermittent are exactly the people most likely to go undiagnosed, and they are absent here by construction.
The sample is also 63.8% female overall, which reflects who uses healthcare regularly rather than the population, and diagnosis counts depend on coding practice as much as on clinical reality.
And a post-2020 inflection invites an obvious question about the pandemic, which disrupted both healthcare delivery and, for many people, the routines that had been masking difficulties. This study identifies when the change happened, not why it happened then.
What it adds up to
The headline conclusion is a practical one: these findings highlight the need for screening strategies that address sex-specific presentations.
Which is a restrained way of saying that a diagnostic model built on boys spent decades missing girls, that the correction is now visibly underway in the data, and that the women being diagnosed at 15 or 30 were autistic the whole time.
People also ask
Why were women and girls missed for so long?
Because the diagnostic picture was built largely from studies of boys. Autistic girls more often mask their difficulties, copy social scripts, and present with anxiety or an eating disorder rather than the overt patterns clinicians were trained to look for. Special interests that fit gendered expectations, such as animals or fiction, attract less clinical attention than the stereotype. The result is a long history of women reaching adulthood without an explanation for their difficulties.
What does a 19% annual percentage change actually mean?
It is the average year-on-year growth in the rate of new diagnoses across the period examined, compounding rather than adding. Sustained over four years it represents roughly a doubling. The uncertainty range here is wide, from 12.1% to 33.4%, so the precise figure is soft even though the direction is clear.
Does this mean autism is becoming more common?
The authors argue not, and the shape of the data is their reason. If something in the environment were increasing autism itself, you would expect rates to rise in both sexes. Instead male rates were flat or falling while female rates climbed. That asymmetry looks far more like a change in who gets recognized than a change in who is autistic.
Why does being diagnosed 3.4 years later matter?
Because those are years without an explanation or appropriate support, often spent being treated for something else. Late diagnosis is associated with worse mental health outcomes, and the delay tends to land in adolescence, when social demands intensify. It also means school-age support was never available to those individuals at the point it would have helped most.
What should someone do if they think they might be autistic?
This is general information rather than medical advice. Adult autism assessment exists and is increasingly available, though waiting lists in many places are long. A primary care clinician is the usual starting point for a referral. A late diagnosis does not change what someone has always been; what it can change is access to accommodations and a more accurate frame for difficulties previously attributed to something else.
References
- Messias E, Casanova E, Huang R, et al. Autism Spectrum Disorder Incidence by Age and Sex in a US Health Care System, 2016 to 2024. JAMA Network Open (2026).
- Centers for Disease Control and Prevention. Signs and Symptoms of Autism Spectrum Disorder.
- National Institute of Mental Health. Autism Spectrum Disorder.