More Disorder or More Diagnosis? What the ADHD Surge Actually Shows

The question behind the numbers

Diagnosed attention-deficit/hyperactivity disorder (ADHD) has risen almost everywhere it is measured, and the rise has accelerated. In the United States, the share of children and adolescents with a diagnosis went from roughly 6% in the late 1990s to about 10% by the mid-2010s, and to 11.4% by 2022. Stimulant prescriptions climbed with it. Yet when researchers pool studies that measure ADHD the same way in the general population, the number is remarkably stable: somewhere between 5% and 8%, year after year, country after country, decade after decade. That divergence is the entire argument. It means the headline "ADHD is rising" and the headline "ADHD prevalence is flat" can both be true, because they are measuring different things.

The natural question is whether the extra diagnoses reflect more people who genuinely have the disorder, or more people being seen, counted, and labeled. The honest answer, on the current evidence, is that the second dominates — but the framing "more disorder versus more diagnosis" is itself the trap. A diagnosis is not a natural object lying in the world. It is a line drawn through a continuous distribution of attention, activity, and impulse control, and the same data can support two different lines. When the line moves, the count of "disordered" people moves with it. The real question is not whether the surge is real, but which part of it is biology and which part is bookkeeping.

Two rates that move for different reasons

It helps to keep two denominators straight. Measured community prevalence is what you get when a study samples a population and applies fixed, explicit criteria with validated instruments — usually a two-stage design with clinician confirmation. It is expensive, slow, and methodologically fussy, but it is designed to be comparable across time and place. Diagnosed prevalence is what you get from surveys of parents and from administrative records: how many people have been told by a clinician that they have ADHD. It is cheap, fast, and exquisitely sensitive to who has access to care, what parents and teachers believe, and what clinics are paid to do.

These two rates have behaved very differently. The pooled worldwide community estimate was 5.29% in a 2007 meta-analysis of 102 studies spanning 1985 to 2004 [1]. A 2015 meta-analysis of 175 studies landed at 7.2%, and — crucially — found no significant difference between studies using older and newer diagnostic criteria [3]. A 2023 umbrella review of meta-analyses put the child and adolescent figure at 8.0% [4]. A 2024 meta-analysis reported pooled survey prevalence near 5.0%, but only 1.6% in studies that relied on routine registers [5]. That last contrast is the whole story in miniature: same construct, same country, and the administrative number is a third of the survey number.

The most instructive study tested the trend directly. A 2014 meta-regression of community-based studies across three decades found no evidence that ADHD prevalence had increased over time once procedures were held constant [2]. When you fix the ruler, the thing being measured stops growing. A 2025 systematic review of post-2020 work reached the same verdict from the other direction: across 40 studies in 17 countries, it found no significant rise in prevalence, only some pandemic-era wobble in incidence — and it cautioned that just four of the 40 studies were at low risk of bias [6]. The flat prevalence finding is not an artifact of old data; it keeps reappearing.

Meanwhile the diagnosed rate marched upward. In the United States it went from 6.1% in 1997-1998 to 10.2% in 2015-2016 [7], and national survey data put it at 11.4% ever-diagnosed and 10.5% currently by 2022 [8], with a separate federal brief reporting 11.3% for ages 5-17 in 2020-2022 [9]. The two rates are not in contradiction. They are a measurement of the net, and a measurement of the fish.

What has stayed flat: prevalence under a fixed ruler

It is worth pausing on why the flat-prevalence result is credible rather than convenient. The meta-regression that first established it was not trying to protect ADHD; it was testing a hypothesis and failing to find a trend [2]. Subsequent reviews with different inclusion rules kept recovering estimates in the same band. And the flat result survives the obvious objection that studies got better at finding cases: if improved instruments were driving the count, the improvement would show up as a trend within the standardized subsets, and it does not. What the stable estimate implies is that the underlying trait — whatever mix of genetics, neurodevelopment, and environment produces the behaviors we call ADHD — has not become dramatically more common in the population. The capacity to notice it has.

This is where a naive reading goes wrong in both directions. The "ADHD is an epidemic of sick children" reading overinterprets diagnosis counts as incidence. The "ADHD is a fabrication" reading overinterprets the definitional problem to deny that any of it is real. The stable-prevalence finding refutes the first and is silent on the second; it tells us the trait's frequency, not its validity. Measurement and diagnosis are the moving parts, and the next three sections are about how they move.

What has moved: the ruler itself

Diagnostic definitions of ADHD have been revised in ways that make the net wider, and this is not a conspiracy — each revision was argued on clinical grounds. The 1994 DSM-IV introduced the three presentational subtypes and permitted diagnosis in the presence of certain other conditions, replacing tighter earlier criteria. The 2013 DSM-5 loosened several hinges at once. It raised the required age of onset from "symptoms present before age 7" to "before age 12," and it changed the requirement that symptoms cause impairment for the age criterion to a requirement merely that symptoms be present. Each change trims away a slice of people previously excluded.

The age rule is the cleanest to quantify. A 2014 analysis of nationally representative US adolescent data found that shifting the onset criterion from age 7 to age 12 raised measured prevalence from 7.38% to 10.84% — an increase of about 47% — with the newly captured later-onset group no less impaired than the earlier-onset group, but more likely to be poor and from a minority background [10]. The same analysis showed the criteria change mechanically manufactures diagnoses. A review using a formal checklist of how disease definitions get modified documented the ADHD criteria changes as textbook definitional widening, with the same pattern seen in other conditions that later proved hard to rein in [11]. When a rule change alone moves prevalence by half, it is not a stretch to attribute a large share of the observed multi-decade rise to the accumulation of such changes.

The relative-age effect: an accidental experiment

If diagnostic boundaries were the only moving part, the story would be merely administrative. But there is a sharper natural experiment, and it is one of the most robust findings in child psychiatry: the youngest child in a school year is substantially more likely to be diagnosed with ADHD and to be prescribed stimulants than the oldest child in the same year, even though the two are the same age modulo a birthday cutoff. The mechanism is almost certainly comparative: within a classroom, a child who is nearly a year behind his or her peers in maturity looks more inattentive and more active, and the behavior reads as pathological.

A 2025 systematic review and meta-analysis confirmed the effect: the relative-age association is significant for diagnosis (relative risk about 1.38) and for medication (about 1.28) [12]. The effect appears in teacher ratings but not in parent ratings [12], which is exactly what an artifact of classroom comparison would predict — parents compare their child to siblings and neighbors, teachers compare to the assembled cohort. A separate individual-participant meta-analysis of 41 prospective studies, pooling more than 4,700 children, found that the relative-age effect did not predict persistence of ADHD symptoms [13]. That dissociation matters: if being young for the grade caused a real, enduring neurodevelopmental difference, it should also predict who stays symptomatic. It does not. It predicts who gets labeled and medicated, and then the label fades — which is what mislabeling looks like.

The relative-age effect is a proof of concept that diagnostic counts can be inflated by context alone, at a scale large enough to matter. It does not prove that every diagnosis is spurious — most children diagnosed with ADHD are not the youngest in their year, and the effect sizes are modest. It does prove that the boundary is socially porous, and that a meaningful fraction of the treated population is being treated for being comparatively immature rather than for a distinct disorder.

The genetic signature of a widening net

The newest and most elegant line of evidence closes the loop. If the rise in diagnoses were driven by a genuinely higher incidence of the underlying disorder, the average genetic risk of newly diagnosed people should stay flat or rise, because more of them would be the deeply affected cases a rising tide produces. If instead the criteria are widening to include milder cases, the average genetic liability of incident cases should fall over time — the bar is being lowered, so each new case sits a little closer to average.

That is what a 2026 study of the Danish population found. Examining the polygenic scores of people diagnosed with ADHD and autism across decades of diagnosis, the researchers found a stable or declining genetic contribution over the years of diagnosis, consistent with broadening diagnostic criteria attracting progressively lower-risk individuals into the patient pool [14]. The direction of the genetic gradient is the tell. A real epidemic should show the opposite. Here the strongest available biological marker moves the way a definitional expansion predicts, not the way a rising disease predicts.

Administrative incidence: DSM-5 and the pandemic

Administrative records let us watch the net widen in real time. A 2026 cohort study of everyone aged 3 to 29 in British Columbia from 2003 to 2023 found that ADHD diagnosis rates accelerated after the DSM-5 changes and again after the COVID-19 pandemic, with the steepest rises in female adolescents and young adults [15]. A companion analysis of adults in the same province found incident diagnoses roughly doubled to tripled in the pandemic period, with women overtaking men as the fastest-growing group [16]. A 2018 multinational study of 13 countries and territories found medication-use prevalence rising almost everywhere [19].

Two things are worth noticing. First, the timing follows policy and diagnosis-seeking, not any known change in human biology — pandemics do not rewrite genomes. Second, the composition of the new cases is itself evidence of ascertainment: girls and women and adults were historically under-recognized, and the surge is concentrated in exactly those groups. That is what a catch-up looks like. It is possible that some genuine increase in impairment — screen-mediated attention problems, disrupted schooling, pandemic distress — is layered on top. But the editorial commentary on the British Columbia findings framed the increase as both a sign of better recognition and a source of concern about overdiagnosis, and noted that the arithmetic still leaves a large treatment gap: for every person taking ADHD medication without a formal diagnosis, roughly three diagnosed people who could benefit are not receiving medication [17]. Both errors are happening at once.

Medication: rising use, unevenly distributed

Stimulant and non-stimulant prescribing has climbed globally, but the rise is concentrated in rich countries and encodes access, not just need. A study of 64 countries found ADHD medication consumption rising about 9.7% per year from 2015 to 2019, but with a stark gradient: high-income countries consumed roughly 6.4 defined daily doses per 1,000 children per day, upper-middle-income countries about 0.37, and lower-middle-income countries about 0.02 — a gap of more than two orders of magnitude [18]. The same study found that, contrary to the long-run trend, US consumption growth had flattened. Rising prescriptions are therefore not a uniform global signal. They are mostly a marker of countries whose health systems can afford to find and treat a condition that remains wildly under-treated elsewhere.

National data track the same trajectory. A UK analysis reported that the number of ADHD prescriptions rose from about 155,000 in 2019-2020 to about 233,000 in 2022-2023, and characterized the rise as a system under strain [24]. The consistency across countries — rising use, a widening demographic, and persistent unmet need — is hard to reconcile with a pure overdiagnosis story, which is why the clean answer to the opening question keeps splitting into two.

What Brazil shows

Brazil is a useful lens because its data combine high self-reported diagnosis with very low measured treatment, which is the opposite of the wealthy-country pattern and makes the ascertainment-versus-burden distinction vivid. A nationally representative study of more than 7,000 children across 87 cities found that 7.1% of parents reported an ADHD diagnosis, but only 3.9% met full DSM-5 criteria on structured assessment, and only 1.9% were currently using a psychostimulant [21]. The agreement between reported diagnosis and probable disorder was low, and the agreement between reported diagnosis and actual treatment was lower still. Diagnosis was strongly graded by socioeconomic status, with poorer children less likely to be diagnosed [21].

The broader Brazilian literature is consistent. A 2012 review argued that ADHD was persistently under-treated in the country [22], and a 2025 analysis of a country-wide drug dataset documented a general rise in psychiatric medication use over the years [23]. Brazil's public system has expanded recognition of ADHD in schools and clinics, but access to diagnosis and to medication remains uneven and often routed through the private sector. The result is a population in which the diagnosis count can rise while a large share of the genuinely impaired still goes unmeasured and untreated. Brazil also reminds us that the relative-age mechanism is a live hypothesis rather than a curiosity: its school-entry cutoffs and its highly stratified public and private schooling make it a natural setting to test whether the youngest-in-class are over-diagnosed, and no consistent long-run national series currently settles the question.

The two camps, steelmanned

The overdiagnosis camp is not cynical. Its strongest claim is that a stable true prevalence plus a rising diagnosed count mathematically implies that some fraction of new diagnoses are of people who would not, under the older and arguably more defensible line, have had the disorder. The relative-age evidence shows that a sizable slice of that fraction is being labeled for context rather than constitution. The falling genetic liability of new cohorts shows the injected cases are systematically milder. And the definitional changes were made without any way to know how many people they would newly capture. When the boundary is drawn by committee, the committee should expect to be wrong at the margin.

The under-recognition camp is not naive either. Its strongest claim is that the flat-prevalence studies measure a trait that was always widespread and largely ignored, and that the rise in diagnosis is a rise in justice: girls, adults, people of color, and the poor are finally being counted. The treatment gradient proves these groups were previously missed. The editorial arithmetic — three diagnosed people untreated for every one medicated without a diagnosis — proves that even in a well-resourced country, the dominant error is under-treatment, not over-treatment [17]. And in countries like Brazil, where 7.1% report a diagnosis but only 1.9% are medicated, calling the problem "overdiagnosis" would be a category error about which side of the ledger needs attention [21].

Both camps are right about different margins. The camps are not symmetric states of the world; they are two tails of the same distribution of error, and a country's income and access determine which tail dominates.

Verdict: probabilities, not a verdict

Weighing the evidence, the most defensible reading carries these calibrated bets:

• High confidence (~85%) that the multi-decade rise in diagnosed ADHD and stimulant treatment is driven mainly by ascertainment — recognition, service access, diagnosis-seeking, and policy — plus deliberate loosening of diagnostic criteria, rather than by a large increase in the incidence of the underlying trait [1,2,5,6,12,14].
• Moderate-to-high confidence (~70%) that there is little or no true increase in the incidence of the core trait, while a modest genuine increase — from changes in schooling, screens, prematurity survival, or environment — cannot be excluded and deserves continued monitoring [2,6,17].
• High confidence (~85%) that "more diagnosis" is not synonymous with "overdiagnosis": the relative-age and genetic evidence identify real mislabeling, while the treatment gradient and the Brazilian data identify large, real under-treatment, and both are simultaneously true [12,14,17,21].
• Moderate confidence (~70-80%) that in Brazil the rise in diagnosis reflects ascertainment, policy, and expanded (mostly private) access more than a rising true burden, with the public-data caveat that no consistent long-run national series exists.

The cleanest way to state the conclusion: ADHD did not become dramatically more common. The line drawn around it moved, the number of people who could see a clinician moved, and the two moved in the same direction at the same time. Some of that is correction of historic neglect. Some of it is the labeling of ordinary immaturity. The evidence cannot yet apportion the two with precision, and anyone who claims otherwise is overreading their preferred side of the ledger.

Reversal conditions

This reading would have to be revised if any of the following appeared. First, a population with a frozen diagnostic definition, a stable instrument, and consistent ascertainment that nonetheless shows a rising incidence of the trait after adjusting for relative age and service access. Second, a demonstrated environmental cause with a large effect size that plausibly shifted incidence — not correlations with screen time or pollution, but a mechanism that survives adjustment and replication across settings. Third, evidence that newly diagnosed cohorts do not carry falling genetic liability when measured with better instruments or in non-European ancestries. Each would move probability toward "more disorder" and away from "more diagnosis."

The gap that would settle it

The decisive study does not exist, and the gap is well defined. It would take a country with a long, stable, identical diagnostic procedure and a fixed threshold; measure incidence repeatedly; and decompose the change into (a) definitional drift, (b) service-access expansion, (c) the relative-age artifact, and (d) residual true change. No country currently runs such a series. The closest proxies — the Danish register work on genetic liability [14] and the Canadian administrative cohorts around the DSM-5 and pandemic [15,16] — are each suggestive but none is a clean fixed-ruler panel.

Brazil is a promising natural laboratory for the relative-age component specifically, because school-entry cutoffs generate sharp within-year age variation and its public stimulant-dispensing and diagnosis data are collected at scale. Linking school-entry age to public dispensing records and to diagnosis, and comparing public and private systems, would convert a plausible mechanism into a measured share of the total. That is a bounded, feasible project rather than a research program, and it is the single highest-value next step this synthesis identifies. Until it is done, the best answer to "is it more disorder or more diagnosis?" remains: overwhelmingly more diagnosis, but the more diagnosis is partly a correction and partly a miscalculation, and the counting has barely begun.

References

1. Polanczyk GV, Willcutt EG, Salum GA, Kieling C, Rohde LA. Attention-deficit/hyperactivity disorder prevalence estimates across three decades: an updated systematic review and meta-regression analysis. International Journal of Epidemiology. 2014;43(2):434-442. doi:10.1093/ije/dyt261
2. Polanczyk GV, de Lima MS, Horta BL, Biederman J, Rohde LA. The worldwide prevalence of ADHD: a systematic review and metaregression analysis. American Journal of Psychiatry. 2007;164(6):942-948. doi:10.1176/ajp.2007.164.6.942
3. Thomas R, Sanders S, Doust J, Beller E, Glasziou P. Prevalence of attention-deficit/hyperactivity disorder: a systematic review and meta-analysis. Pediatrics. 2015;135(4):e994-e1001. doi:10.1542/peds.2014-3482
4. Ayano G, Demelash S, Gizachew Y, Tsegay L, Alati R. The global prevalence of attention deficit hyperactivity disorder in children and adolescents: an umbrella review of meta-analyses. Journal of Affective Disorders. 2023;339:860-866. doi:10.1016/j.jad.2023.07.071
5. Popit S, Serod K, Locatelli I, Stuhec M. Prevalence of attention-deficit/hyperactivity disorder (ADHD): systematic review and meta-analysis. European Psychiatry. 2024;67(1):e68. doi:10.1192/j.eurpsy.2024.1786
6. Martin AF, Rubin GJ, Rogers MB, Wessely S, Greenberg N, Hall CE, et al. The changing prevalence of ADHD? A systematic review. Journal of Affective Disorders. 2025;388:119427. doi:10.1016/j.jad.2025.119427
7. Xu G, Strathearn L, Liu B, Yang B, Bao W. Twenty-year trends in diagnosed attention-deficit/hyperactivity disorder among US children and adolescents, 1997-2016. JAMA Network Open. 2018;1(4):e181471. doi:10.1001/jamanetworkopen.2018.1471
8. Danielson ML, Claussen AH, Bitsko RH, Katz SM, Newsome K, Blumberg SJ, et al. ADHD prevalence among US children and adolescents in 2022: diagnosis, severity, co-occurring disorders, and treatment. Journal of Clinical Child & Adolescent Psychology. 2024;53(3):343-360. doi:10.1080/15374416.2024.2335625
9. Reuben C, Elgaddal N. Attention-deficit/hyperactivity disorder in children ages 5-17 years: United States, 2020-2022. NCHS Data Brief No. 499. National Center for Health Statistics; 2024. doi:10.15620/cdc/148043
10. Vande Voort JL, He JP, Jameson ND, Merikangas KR. Impact of the DSM-5 attention-deficit/hyperactivity disorder age-of-onset criterion in the US adolescent population. Journal of the American Academy of Child & Adolescent Psychiatry. 2014;53(7):736-744. doi:10.1016/j.jaac.2014.03.005
11. Sanders S, Thomas R, Glasziou P, Doust J. A review of changes to the attention deficit/hyperactivity disorder age of onset criterion using the checklist for modifying disease definitions. BMC Psychiatry. 2019;19(1):357. doi:10.1186/s12888-019-2337-7
12. Frisira E, Holland J, Sayal K. Systematic review and meta-analysis: relative age in attention-deficit/hyperactivity disorder and autism spectrum disorder. European Child & Adolescent Psychiatry. 2025;34(2):381-401. doi:10.1007/s00787-024-02459-x
13. Gosling CJ, Caparos S, Pinabiaux C, Schwarzer G, Rücker G, Agha SS, et al. Association between relative age at school and persistence of ADHD in prospective studies: an individual participant data meta-analysis. The Lancet Psychiatry. 2023;10(12):922-933. doi:10.1016/S2215-0366(23)00272-9
14. LaBianca S, Lousdal ML, Dybdahl Krebs M, Sahlholdt Hansen O, Georgii Hellberg KL, Lundberg M, et al. Changes in genetic contributions to ASD and ADHD by year of diagnosis. JAMA Psychiatry. 2026;83(9):921-929. doi:10.1001/jamapsychiatry.2026.1450
15. Cui Z, Ambasta A, Thompson W, Bassett K, Carney G, Dormuth C. DSM-5 changes, COVID-19, and ADHD diagnosis rates in individuals younger than 30 years. JAMA Network Open. 2026;9(4):e265775. doi:10.1001/jamanetworkopen.2026.5775
16. Hu K, Desai R, Au S, Zhao B, Barbic S, Marchand K, et al. Trends of incident adult attention-deficit/hyperactivity disorder diagnoses before, during and after the pandemic provincial state of emergency in British Columbia (2013-2023): a population-based study. The Lancet Regional Health - Americas. 2025;51:101223. doi:10.1016/j.lana.2025.101223
17. Cortese S. Incidence of ADHD diagnoses on the rise—good or bad news? JAMA Network Open. 2026;9(4):e265784. doi:10.1001/jamanetworkopen.2026.5784
18. Chan AYL, Ma TT, Lau WCY, Ip P, Coghill D, Gao L, et al. Attention-deficit/hyperactivity disorder medication consumption in 64 countries and regions from 2015 to 2019: a longitudinal study. eClinicalMedicine. 2023;58:101780. doi:10.1016/j.eclinm.2022.101780
19. Raman SR, Man KKC, Bahmanyar S, Berard A, Bilder S, Boukhris T, et al. Trends in attention-deficit hyperactivity disorder medication use: a retrospective observational study using population-based databases. The Lancet Psychiatry. 2018;5(10):824-835. doi:10.1016/S2215-0366(18)30293-1
20. Abdelnour E, Jansen MO, Gold JA. ADHD diagnostic trends: increased recognition or overdiagnosis? Missouri Medicine. 2022;119(5):467-473. PMID:36337990
21. Arruda MA, Arruda R, Guidetti V, Bigal ME, Landeira-Fernandez J, Portugal AC, Anunciação L. Associated factors of attention-deficit/hyperactivity disorder diagnosis and psychostimulant use: a nationwide representative study. Pediatric Neurology. 2022;128:45-51. doi:10.1016/j.pediatrneurol.2021.11.008
22. Mattos P, Rohde LA, Polanczyk GV. ADHD is undertreated in Brazil. Brazilian Journal of Psychiatry. 2012;34(4):513-516. doi:10.1016/j.rbp.2012.04.002
23. Panzenhagen AC, Maffini ACS, Dantas R, Shimomura K, Ikeda MLR, Gelain DP, Shansis FM. Increased use of psychiatric drugs in Brazil over the years: evidence from a country-wide dataset. Trends in Psychiatry and Psychotherapy. 2025 (online ahead of print). doi:10.47626/2237-6089-2024-0914
24. Mahase E. ADHD: is prevalence increasing? BMJ. 2025;389:r1218. doi:10.1136/bmj.r1218