Missed and late diagnosis

Pooling 56 studies from 40 countries published between 2012 and 2019, the mean age at autism diagnosis was 60.5 months, a little over 5 years, though individual studies ranged from about 31 to 235 months. In studies limited to children aged 10 or under it was 43 months.¹ An earlier review of 42 studies found means between 38 and 120 months, falling over time.² In two large UK databases of families taking part in research, the median was 55 months, and it did not fall between 2004 and 2014.³

In US surveillance of 8-year-olds in 2022, the median age of the earliest known diagnosis was 47 months, ranging from 36 months in California to 69.5 in Laredo, Texas. Children born in 2018 were 1.7 times as likely as children born in 2014 to have a diagnosis or autism special-education eligibility by age 4.⁴ In a nationally representative UK cohort, with timing reported by parents, 79% of children diagnosed by age 14 were diagnosed after starting school and 28% not until secondary school. Three in four of that last group had been flagged by a parent or teacher at age 5.⁵

A single large study suggests that autism diagnosed earlier and autism diagnosed later have partly different developmental and genetic profiles. Using longitudinal data from four population birth cohorts, Zhang and colleagues (2025) found that autistic children follow two different social and behavioral trajectories that go with the age at diagnosis. In separate genetic analyses, common genetic variants explained about 11% of the variation in the age at diagnosis, similar to any one sociodemographic or clinical factor, which typically explains under 15%, and the autism polygenic signal split into two modestly correlated factors. One was linked to an earlier diagnosis and to lower social and communication abilities in early childhood; the other to a later diagnosis and to more emotional and behavioral difficulty in adolescence, with stronger genetic correlations with ADHD and mental-health conditions.⁴²

Autism was first described in children, most of them boys. Kanner's 1943 report described 11 children, 8 of them boys, and Asperger's 1944 report described boys.⁷ When autism entered the DSM in 1980 as "infantile autism", people who no longer met the criteria were given a "residual" label, and a later review judged that the problems of autistic adults were not adequately represented.⁷ DSM-IV and ICD-10 required signs before age 3. DSM-5 and ICD-11 require them in the early developmental period and note that some may not show fully until social demands exceed a person's capacities.⁷

Among several possible explanations for girls being missed, the authors of that review list measures that were normed mainly in male-dominated samples and may be less sensitive in girls, particularly to milder repetitive behavior.⁷ In one US lab sample of 145 adults (50 women, 95 men) who already had a community diagnosis, using the ADOS as a gate for research excluded women 2.5 times as often as men.⁸ A pooled analysis of 27 sites, with the largest sample of autistic girls at the time, looked at children who had been diagnosed: after allowing for age, nonverbal IQ and language, scores on the main measures differed little by sex, with small to moderate differences emerging in adolescence. Its authors note it could not test people who were never diagnosed.⁹ In a review of studies of adults diagnosed late, participants in one study described diagnostic practice as built on the behavior of young boys, and others said it overlooked gender differences.¹⁰

A meta-analysis of 54 studies found about 4.2 autistic boys for every autistic girl overall. In studies that screened whole populations the ratio was 3.25, against 4.56 in studies that counted only people already diagnosed. Results varied widely between studies. The authors conclude that the true ratio is nearer 3 to 1, and that there appears to be a diagnostic gender bias that leaves girls who meet the criteria disproportionately undiagnosed.¹¹ A boy-to-girl difference remains even in population screening, and how much of it reflects missed girls is unresolved. A 2022 review of prevalence studies worldwide still found a median ratio of 4.2.¹² In Norway's national registers, the ratio was 3.67 among children and 2.57 among adults; the authors suggest later diagnosis in females, diagnostic bias and diagnostic trends as possible reasons.¹³ In a population sample of 10- to 12-year-olds, girls who met diagnostic criteria more often had an intellectual disability or behavioral difficulties than girls with equally high autistic traits who did not meet them. Boys showed no such difference. The authors say this could reflect bias in diagnosis or better adaptation in girls.¹⁴ In an English survey of adults, male sex strongly predicted autism only among those without an intellectual disability or with a mild one, and the authors suggest the measures may miss more women. Few autistic adults were found, so the estimate is imprecise.¹⁵

Findings on timing are mixed. Females were diagnosed later than males in some subgroups only of a Dutch survey of 2,275 people, among children in a Scottish clinical sample of 150, in 61 Italian adults without an intellectual disability, and in two of three US samples, but not in the third.^16,17,18,19 In US surveillance in 2002, being male went with earlier identification.²⁰ In the Scottish study of 150 people, assessment took no longer for females than for males, which the authors take to suggest the delay arises before referral.¹⁷ The meta-analysis of age at diagnosis, though, found that the factors studied, gender among them, often gave conflicting or inconclusive results.¹ Gender-diverse autistic people were diagnosed later than cisgender people in two US samples, and later than cisgender men in a study of 2,722 Dutch adults, in which cisgender women were the oldest at diagnosis.^19,21

A systematic review of barriers for girls and young women found five, largely as perceived by parents: compensatory behavior, parents' concerns, other people's perceptions, a lack of information and clinician bias.²² Camouflaging, and the studies linking it to later diagnosis, are covered in Masking.

Children whose autism comes with an intellectual disability, a language delay or greater support needs tend to be diagnosed earlier. In US surveillance of 8-year-olds at 13 sites in 2002, before the current criteria, the median age of identification was 5.7 years, and an IQ of 70 or below, being male and losing skills already gained each went with earlier identification.²⁰ In the UK research databases, an autism (rather than another spectrum) diagnosis, language delay, language regression and greater support needs went with earlier diagnosis.³ A review of 42 studies found greater severity went with earlier diagnosis, consistently across the studies that tested it.² In the UK cohort, children without a cognitive delay at age 5 were more likely to be diagnosed in secondary school than in primary school.⁵

In England, autism was far more common among adults with a moderate to profound intellectual disability than among other adults (odds ratio 63.5, with a wide margin of uncertainty; the intellectual disability sample was 290 adults).¹⁵ Among 161 adults, mostly men, first diagnosed autistic at two Italian specialist centers, intellectual disability was one of the most common earlier diagnoses.²³

In US surveillance in 2002, among 2,568 8-year-olds who met the criteria for autism, Black, Hispanic and other children were less likely than White children to have a documented diagnosis (odds ratios 0.79, 0.76 and 0.65, after adjusting for IQ, sex and other factors). For Black children the gap held regardless of IQ; for the other groups it was concentrated among children with an IQ below 70.²⁴ Among 406 US children on Medicaid, White children were 2.6 times as likely as Black children to be diagnosed autistic at their first specialist visit. ADHD was the most common first diagnosis for those who were not; Black children were 5.1 times as likely as White children to be given adjustment disorder rather than ADHD, and 2.4 times as likely to be given conduct disorder rather than ADHD.²⁵ Parents of 584 Black autistic children in a US genetics research repository recalled an average age at diagnosis of 64.9 months, 42.3 months after they first had concerns, with very wide variation.²⁶

The US pattern has since changed. Since 2020, the same surveillance network has found autism more common among Asian or Pacific Islander, Black and Hispanic 8-year-olds than among White 8-year-olds, and in 2022 among American Indian or Alaska Native children too. Among children with cognitive data in 2022, 52.8% of Black autistic 8-year-olds had an intellectual disability, against 32.7% of White autistic 8-year-olds.⁴ These figures describe the children who were identified, so they cannot show how common intellectual disability is among all autistic children in each group. Population counts are in Prevalence.

Findings on income differ by country. In a review of 42 studies, 19 of them from the US, 5 of 11 studies of socioeconomic status linked higher status to earlier diagnosis and 6 found no link.² In the UK, one study found that lower status went with earlier diagnosis, and another that children living in poverty were diagnosed later.^3,5

Many people diagnosed autistic later in life were first diagnosed with something else. In a US parent survey of 1,496 autistic children, about 1 in 5 had been diagnosed with ADHD first. They were diagnosed autistic about 3 years later than children whose ADHD was diagnosed at the same time or afterwards, and were nearly 30 times as likely to be diagnosed after age 6.²⁷ Among 258 Israeli children, mostly boys, first diagnosed at ages 6 to 18, earlier diagnoses of language delay and ADHD were common, and two thirds had been prescribed medication for behavioral or psychosocial conditions first.²⁸

In the 161 Italian adults, two thirds had at least one earlier psychiatric diagnosis, and the median gap from first mental-health contact to the autism diagnosis was 11 years.²³ In Quebec health records, 77.5% of adults first diagnosed autistic between 2012 and 2017 had an anxiety disorder recorded in the preceding years.²⁹ In a US sample of 4,657 autistic adults, those diagnosed at 21 or later were more likely to self-report psychiatric conditions than those diagnosed as children.³⁰ Some of these earlier diagnoses will be accurate co-occurring conditions and some mistaken, and the studies cannot separate the two. Nor can they show that an earlier diagnosis caused the delay: children diagnosed with ADHD first may also differ in how their autism shows.

In a Dutch register of 1,211 autistic adults, most of them well educated, 24.6% reported at least one earlier psychiatric diagnosis they saw as a misdiagnosis: 31.7% of women and 16.7% of men. Personality disorders were the most common, followed by anxiety and mood disorders.³¹ In a later study of Dutch adults, very likely from the same register, gender-diverse adults reported the highest rates.²¹ Review authors recommend that clinicians assessing adults explore possible misdiagnosis, especially in women.³² Personality-disorder diagnoses, and what autistic people say about being given one before an autism assessment, are covered in Personality-disorder diagnoses and misdiagnosis.

A UK matched cohort drawn from primary-care records adds rate-based measures. Among 22,112 autistic adults in the IQVIA Medical Research Database, compared with 221,120 people without an autism diagnosis matched on age, sex and practice (median age 20, about three-quarters men), autistic adults without an intellectual disability were about 14 times as likely to already have a personality-disorder diagnosis on record at the start of follow-up (14.0 for men, 14.9 for women), and about 5 times as likely to have a new one recorded during it (4.8 for men, 4.6 for women); follow-up starts at cohort entry, the latest of the autism diagnosis and the data's start date.⁴³ In autistic adults with an intellectual disability the pattern differed by sex: the new-diagnosis ratio was higher in women (8.3 times) but lower in men (2.0 times) than without an intellectual disability. These are diagnoses recorded by clinicians, so they show how often the diagnosis is given, not whether it was wrong.

These earlier diagnoses are hardest to separate from borderline personality disorder (BPD), which shares many features with autism. A systematic review of studies published between 2020 and 2024, covering 684 participants, including 134 health-care professionals, found that the overlap, especially with BPD, undermines diagnostic accuracy, and the authors conclude that the lack of agreed criteria makes errors and delays more likely.⁴⁴ In a cross-sectional comparison, 51 autistic women and people assigned female at birth with few or subclinical BPD traits and 51 with a BPD diagnosis who were not autistic answered self-report questionnaires. The largest single differences were identity disruption, higher in the BPD group, and sensory processing, higher in the autistic group; a set of measures that included social cognition and camouflaging classified 95.1% of the participants correctly in that sample.⁴⁵ The authors note that autistic people may still resemble the BPD group more than the general population does, and the reverse.

The same problem of missed autism appears in clinics where adults are assessed for another condition first. In a UK specialist clinic for functional neurological disorder, 8 of 16 consecutively assessed adults met the criteria for clinically probable autism, and the autistic group had a younger age at symptom onset and a longer wait from onset to the FND diagnosis.⁴⁶ The sample is small and the authors call the findings preliminary.

Lai and Baron-Cohen describe the idea of a "lost generation": adults who had been excluded from a diagnosis of classic autism. They tie first diagnosis in adulthood to greater awareness, broader criteria and the idea of a spectrum.³² A 2007 survey of English households found about 1% of adults autistic (9.8 per 1,000, estimated from a small number of cases and so with a wide margin of uncertainty), with no difference by age, and the autistic adults it found tended to be unrecognised.³³

In English primary-care records in 2018, 2.94% of 10- to 14-year-olds had an autism diagnosis, against 0.02% of people aged 70 and over. From these data the authors project, as an exploratory estimate, that 59 to 72% of autistic people in England may have been undiagnosed. The projection depends on an assumed figure for how common autism is.³⁴ New diagnoses that year ran at about 1 in 250 children aged 5 to 9, 1 in 4,000 adults aged 20 to 49 and 1 in 18,000 people aged 50 and over.³⁴ In another UK primary-care database, the rate of recorded new diagnoses rose 787% (a relative rise, from a low starting rate) between 1998 and 2018, most steeply among adults and more among females than males. The authors judge greater recognition and wider use of the diagnosis the more likely explanation than more autism.³⁵ What undiagnosed older adults mean for aging research is covered in Aging.

Review authors argue that diagnosing adults can be harder: there may be no one who can describe the person's early development, learned camouflaging can hide traits, and co-occurring conditions are common.³²

A systematic review of 26 studies of adults' experiences, most of them qualitative, found no single route to diagnosis in adulthood. Many adults learned about autism through friends and acquaintances, the media, other autistic adults or their own child's diagnosis, recognized themselves, and then sought a professional to confirm it. Others were diagnosed after a crisis or a life transition, when seeking help for mental-health problems led to the autism diagnosis.¹⁰ In two small UK interview studies, 9 people diagnosed after 50 reported earlier treatment for anxiety and depression, and 11 women diagnosed after 40 described being mislabelled.^36,37

In an Australian survey, adults sought a diagnosis for an explanation and for support; cost and fear of not being taken seriously were the main barriers.³⁸ In a UK survey of 128 adults without an intellectual disability, routes varied widely, and 40% were dissatisfied with the process against 47% satisfied.³⁹ Among 130 adults who saw themselves as autistic without a formal diagnosis, self-identification often began with encouragement from friends or family, or with recognizing traits in autistic celebrities. Most wanted a formal diagnosis but were put off by waiting lists and problems with healthcare, and one in ten did not want one.⁴⁰ In US interviews with 65 queer and transgender autistic adults, some formally diagnosed and some self-identified, people described barriers to diagnosis, and being disbelieved both before and after it.⁴¹

Autistic writers have their own words for adults who find out they are autistic. Devon Price writes of "late-realized" and "self-realized" Autistics, Sol Smith of "late-identified autistics", and Megan Anna Neff of "late-in-life-identified Autistic people".^c3,c4,c6 The words cover people who recognize themselves as well as people diagnosed by a clinician; how they relate to formal assessment is covered in Assessment.

Many describe the discovery as a turning point. Price writes that "Almost every neurodiverse person I've spoken to for this book shared that discovering they were Autistic was a powerful aha moment, one that prompted them to rethink every narrative they'd believed about who they were."^c1 In Price's account, labels such as stupid, clueless or lazy gave way to an explanation, and lost friends, dropped classes and lost jobs were recast as a mismatch between a person's neurotype and what society expected.^c1,c3 Smith tells of a client who lost five jobs in his thirties with no clear reason; after realizing he was autistic, he felt "less like a failure and more like someone who performed well but was frequently misunderstood."^c4 Neff, asked how her own late diagnosis affected her, names "the awakening of self-compassion" first, and notes that many people identified late go on to reconsider their careers.^c5,c6

Price writes of having "heard hundreds of Autistic people" tell versions of one story: a child shows signs, the family pushes the label away, the child learns to hide, and after years of strain "they have some kind of breakdown" that leads to the discovery.^c2 Price's example is a woman whose months of burnout after college led to therapy and then an assessment, and whose family still did not believe it.^c2 Megan Griffith describes a different road: a diagnosis of bipolar disorder in college that she later came to doubt, then autism videos on TikTok that made "my whole life make sense", a self-diagnosis she withdrew after a therapist and a psychiatrist dismissed it, and finally a professional assessment for autism and ADHD.^c7 These are personal accounts; the routes above come from studies, which also describe diagnosis after a crisis.

Clinicians who assess adults have a name for one way autism is missed. In their guide to diagnosis, the clinicians Donna Henderson, Sarah Wayland and Jamell White list "diagnostic overshadowing" among the reasons clinicians miss autistic people: many have long carried diagnoses such as ADHD, anxiety, OCD or depression, and their autistic traits are put down to those.^c8 The same authors warn clinicians against assuming that "If she was autistic, someone else would have diagnosed it by now", or that an earlier rule-out of autism was sound.^c8 One of their examples is a medical student diagnosed with social anxiety disorder years before her autism diagnosis, who said of a missed cue in a restaurant that she was not anxious but "just didn't know that I was supposed to say something at that moment."^c9 Carly Jones, an autistic advocate writing for professionals, uses the term for disabled people whose every difficulty is put down to a first disability, such as deafness, so that autism is never considered.^c10

Autistic adults quoted in the same guide describe what a mistaken diagnosis cost them. Kayleigh R. was given several diagnoses in her twenties, borderline personality disorder among them, and treated with antidepressants and antipsychotics; after a friend's diagnosis led her to read how autism presents in women, she was diagnosed at 39 and describes herself as unmedicated and probably the most stable she has ever been.^c9 Reanna P. writes that the personality-disorder diagnosis "was overall invalidating to my lifelong struggles of [what I now know as] being autistic."^c9 Burford, writing for newly diagnosed adults, argues that a borderline personality disorder diagnosis "puts the cause of your suffering inside your personality, which is the one place it definitely wasn't."^c11 Melanie Yergeau, an autistic scholar of rhetoric, writes that an adult diagnosis "is often read as misdiagnosis", though many adult diagnoses replace or add to earlier ones such as intellectual disability, schizophrenia, ADHD or borderline personality disorder.^c12 The studies of earlier diagnoses are in the section above; these accounts do not show how often each happens.

Writers missed for decades describe not matching the picture of autism at the time. In a poem about her own childhood, Morénike Giwa Onaiwu describes a girl who "doesn't fit the profile", being "not of the gender and not of the race" then associated with autism; she sought her own evaluation after both her children were diagnosed.^c13 Michael, writing in a collection by autistic adults aged 20 to 70 and over, learned he was autistic at 68 when someone in an online group asked whether he realized it, and concluded that autistic people are "often invisible to themselves."^c14

Writers also say the change is slow. Price writes that "Piecing together a new identity as a disabled person and unlearning an entire culture's worth of anti-Autistic stigma take years for many of us."^c3