Autism in women and girls: the underdiagnosis numbers behind the sex ratio

She was 31 the first time anyone said the word autism out loud. By then she had two children, a stack of anxiety and depression labels, and a reputation for being "sensitive." In high school she had been the quiet girl with perfect grades who spent lunch in the library. In residency I had watched versions of this story long enough to stop treating it as rare: a woman who worked hard to look fine in every room, then fell apart when she got home.

Her childhood records were full of social worry, sensory overload, and rigid routines that nobody connected. She had been told she had generalized anxiety. Later, borderline traits. Nobody asked how much effort it took to hold a conversation or sit through a birthday party without planning an exit.

That path is one of the main reasons the widely quoted male-to-female ratio looks the way it does. When clinics count diagnosed children, boys still outnumber girls by roughly three to one. When researchers screen whole populations carefully, the gap shrinks. The difference between those two numbers is where underdiagnosis lives. This article gathers the citable figures: the current U.S. sex ratio, careful-ascertainment estimates, age at diagnosis by sex, camouflaging research, co-occurring conditions, and misdiagnosis patterns.

Surveillance year	Prevalence per 1,000
2014	16.8
2016	18.5
2018	23.0
2020	27.6
2022	32.2

Source: CDC ADDM Network combined site estimates (Data and Statistics on Autism Spectrum Disorder; 2022 detail in Shaw et al., MMWR 2025).

Estimate	Male-to-female ratio	Source
U.S. ADDM 8-year-olds, 2018	4.2	Maenner et al., MMWR 2021
U.S. ADDM 8-year-olds, 2020	3.8	CDC / Maenner et al. 2023 series
U.S. ADDM 8-year-olds, 2022	3.4	Shaw et al., MMWR 2025
All prevalence studies (pooled)	4.20	Loomes et al., 2017
High-quality studies only	3.32	Loomes et al., 2017
General-population screening	3.25	Loomes et al., 2017
Pre-existing diagnosis samples only	4.56	Loomes et al., 2017

Sources: CDC ADDM Network MMWR reports; Loomes, Hull & Mandy, Journal of the American Academy of Child & Adolescent Psychiatry, 2017.

The sex ratio that is not fixed

For years, textbooks and clinic handouts repeated a simple figure: autism is four times more common in boys than in girls. That number still appears everywhere. It is not entirely wrong as a description of who gets counted in many clinical samples. It is a poor description of who meets autism criteria when researchers look harder.

The CDC's latest ADDM Network report, covering 16 U.S. sites in 2022, found that 32.2 per 1,000 children aged 8 years (about 1 in 31) had been identified with autism spectrum disorder. Prevalence was 49.2 per 1,000 among boys and 14.3 per 1,000 among girls. The male-to-female prevalence ratio was 3.4. That is lower than the 4.2 ratio the same network reported for 2018 and the 3.8 ratio for 2020, per the 2025 MMWR discussion of those consecutive reports.

Here is the nuance the ratio alone misses. The CDC authors note that while the ratio has narrowed, the absolute prevalence gap between boys and girls has widened: from 27.7 per 1,000 in 2018 to 31.7 in 2020 to 34.9 in 2022. More girls are being identified, but more boys are being identified too, and the boy count has risen enough that the absolute difference keeps growing. A shrinking ratio is not the same thing as closing under-identification of girls.

The Loomes et al. meta-analysis remains the cleanest statement of diagnostic bias. Across 54 studies and more than 13.7 million participants, the overall pooled male-to-female odds ratio was 4.20. High-quality studies fell to 3.32. Studies that screened the general population regardless of prior diagnosis sat at 3.25. Studies that only counted children who already had a clinical diagnosis sat at 4.56. Their conclusion: of children meeting criteria for ASD, the true ratio is closer to 3:1 than 4:1, and girls who meet criteria are at disproportionate risk of never receiving a clinical diagnosis.

Biology still matters. No serious reading of the literature says the sex difference is entirely artifact. Sex-linked genetic load and co-occurring intellectual disability still show up in better studies. The useful point is narrower: part of the gap is diagnostic, and the size of that part depends on how hard you look.

When girls and women get diagnosed

Timing is half the clinical story. The CDC ADDM Network reports a median age of earliest known ASD diagnosis of 47 months among 8-year-olds with a diagnostic statement in 2022, just under four years. That is a network-wide figure for identified children and does not by itself prove a large childhood sex difference in median age. The adult clinic data tell a different, and more uncomfortable, story about who is still waiting.

Gesi and colleagues studied 61 adults with autism and no language or intellectual disability referred to a tertiary service in Milan for co-occurring psychiatric care. Women first reached mental health services at a mean age of 21.1 years, versus 11.7 years for men. Mean age at ASD diagnosis was 29.4 years for women and 19.8 years for men. About three-quarters of the whole sample (75.4%) had not received an autism diagnosis at first evaluation; the average lag from first mental-health contact to ASD diagnosis was about eight years.

Those numbers come from one specialty clinic, not a national registry. Treat them as a description of late-diagnosed, cognitively able adults (the group where female under-recognition concentrates) rather than the average for every autistic girl. They still match what many of us see: boys flagged in preschool, women arriving in adult psychiatry after a decade of partial answers.

Grosvenor and colleagues' 2024 analysis of more than 12 million U.S. health-plan members shows the catch-up pattern. In 2022, autism diagnosis rates were about 9.8 per 1,000 males and 3.1 per 1,000 females, still roughly 3 to 1. From 2011 to 2022, relative increases were larger for females: 305% among female children versus 185% among male children, and 315% among adult women versus 215% among adult men. The child male-to-female ratio fell from 4.29:1 to 3.01:1; the adult ratio fell from 3.45:1 to 2.60:1.

For families tracking developmental concerns alongside menstrual and reproductive health (another place sensory and anxiety symptoms get misfiled), our period calculator and cycle length tracker can organize when symptoms spike. They are not autism screens. They are a way to stop losing the pattern in memory.

Measure	Women	Men
Mean age at first referral	21.1 years	11.7 years
Mean age at ASD diagnosis	29.4 years	19.8 years
Misdiagnosed at first evaluation	45.5%	17.9%
ASD diagnosis missed at first evaluation	54.5%	52.6%

Source: Gesi et al., Brain Sciences, 2021 (n = 22 women, 39 men; tertiary adult ASD comorbidity clinic).

Camouflaging and the hidden presentation

Camouflaging is the set of strategies autistic people use to hide autistic traits or force a more neurotypical social performance: scripting conversations, forcing eye contact, copying gestures, suppressing stimming, rehearsing facial expressions. It is work, not a personality style, and the research on who does more of it keeps pointing toward autistic women.

Hull and colleagues developed the Camouflaging Autistic Traits Questionnaire (CAT-Q), a 25-item self-report measure with three factors (Compensation, Masking, Assimilation), published in the Journal of Autism and Developmental Disorders in 2019. Autistic adults scored substantially higher than non-autistic adults. In a 2020 gender analysis of 306 autistic and 472 non-autistic adults, Hull and colleagues found that autistic women scored higher on total camouflaging than autistic men (partial η² = 0.08), with higher Masking and Assimilation subscale scores; there was no gender difference among non-autistic adults.

Those findings are research results, not diagnostic cutoffs. The CAT-Q is not an autism test. High camouflaging can still suppress what an observer sees in a short clinic visit. If instruments and clinician expectations were built largely on male childhood presentations, a girl who has learned to smile on cue can score below threshold while burning out at home. In Hull's validation work, higher CAT-Q scores correlated with higher anxiety and depression among autistic participants. Camouflaging may keep a girl enrolled; it does not keep her well.

Self-report camouflaging is imperfect, and some autistic men camouflage heavily. The clinical takeaway is still solid: if you only look for the child who cannot hide, you will miss the child (and later the woman) who can.

Co-occurring conditions

Autism rarely arrives alone in the medical record, and for women the co-occurring labels often arrive first.

Lai and colleagues' 2019 systematic review and meta-analysis in The Lancet Psychiatry pooled prevalence estimates across dozens of studies. Among autistic people overall, estimated co-occurrence was 28% for ADHD, 20% for anxiety disorders, 13% for sleep-wake disorders, 12% for disruptive/impulse-control/conduct disorders, 11% for depressive disorders, 9% for OCD, 5% for bipolar disorders, and 4% for schizophrenia spectrum disorders. Clinical samples ran higher than population or registry samples. Heterogeneity was large; age, sex, intellectual functioning, and country all mattered.

Those are not "female only" numbers. They are base rates any women's-health clinician should carry. They also map onto conditions already overrepresented in women (anxiety, depression, sleep disruption), which is how autism gets filed under a mood diagnosis and never reopened. Our coverage of women's anxiety and depression statistics and sleep disorders in women describes that broader burden.

In the Gesi adult clinic sample, 49.2% of patients had one co-occurring diagnosis and 39.3% had two. Mood disorders (31.1%), trauma- and stressor-related disorders (26.2%), and anxiety disorders (16.4%) were common. Eating disorders appeared in 6.6% of that small sample. That rate is too imprecise to overgeneralize, but it fits literature linking restrictive eating and autism in girls. See our eating disorder statistics page for the wider women's health picture.

Among 8-year-olds with autism in the 2022 ADDM Network, 39.6% of those with cognitive data met criteria for intellectual disability (IQ ≤70), in similar proportions of girls and boys (40.4% vs. 39.5%). The sex ratio is often narrower when intellectual disability is present and wider among cognitively able autistic people. That is exactly where camouflaging and missed diagnoses concentrate.

Misdiagnosis patterns

Missed diagnosis and wrong diagnosis are different problems. Both are common in autistic women without intellectual disability.

In Gesi et al., 45.5% of women had been misdiagnosed at first mental-health evaluation, compared with 17.9% of men. Among the 10 women with a prior misdiagnosis, eight had received a personality-disorder label, one an anxiety diagnosis, and one a psychotic-spectrum diagnosis. Men who were misdiagnosed were more often labeled ADHD or with psychotic or behavioral problems. The sample is small; the pattern is familiar to anyone who has sat through adult ASD evaluations for women.

Why personality disorders, especially borderline personality disorder? Overlapping surface features: emotional intensity after social rupture, self-injury in some autistic people, relationship instability that is actually social-communication mismatch, identity questions after years of masking. Sensory overload can look like affective lability if nobody asks about fluorescent lights or clothing tags. Gesi found that women with a history of misdiagnosis scored higher on sensory hyper/hyporeactivity than misdiagnosed men.

Other common wrong or incomplete stops include social anxiety, recurrent depression, and eating-disorder pathways that never reassess social development. Co-occurrence is real; the failure mode is stopping at the first label that partially fits. Three-quarters of Gesi's sample overall had autism missed at first evaluation, with similar "missed" rates by sex, but women were significantly more likely to leave with an actively wrong diagnosis.

I will not invent a tidy national "80% of autistic women are misdiagnosed" figure. That number circulates online without a primary source I can defend on deadline. What I can defend is narrower: in one carefully described adult clinic sample, nearly half of autistic women without intellectual disability had been given the wrong diagnosis first, most often a personality disorder.

What the gap means in practice

Epidemiology only helps if it changes what we do in the exam room and in schools.

Stop treating a 4:1 boy-to-girl ratio as a prior that makes autism "unlikely" in a girl with clear social-communication and sensory history. CDC surveillance already sits at 3.4:1 for identified 8-year-olds. Careful ascertainment sits near 3:1. Adult health-system data show female diagnosis rates rising faster than male rates.

Take adult presentation seriously. A first autism diagnosis at 25 or 35 is not a fad; it is what you get when childhood systems only recognize children who cannot camouflage. Gesi's mean female diagnosis age of 29.4 years is one snapshot of that backlog.

Integrate autism into women's mental health pathways rather than sequestering it in pediatric developmental clinics. Perinatal mood visits and primary-care depression follow-ups are full of women describing lifelong sensory and social strain. Our page on postpartum depression statistics is adjacent terrain: the same women can carry both a perinatal mood episode and an unrecognized neurodevelopmental difference.

Hold two ideas at once. Autism is still more often identified in males, and some of that difference is likely biological. Autism is also still under-identified in females, through diagnostic bias, male-normed tools, and camouflaging. Self-identification is not a clinical diagnosis, and online quizzes are not a substitute for developmental assessment. Asking for that assessment is still a reasonable medical request when lifelong social effort, sensory load, and co-occurring anxiety have never been considered through an autism lens.

Frequently asked questions