Article 6 of 6

ADHD can be under-recognised nationally and overdiagnosed in particular people or settings. The evidence supports concern about both errors, but not a claim that Britain is experiencing a general epidemic of false adult diagnosis. The useful goal is better-distributed, auditable diagnosis — and support that does not always require a label.

This series has argued that Pemberton has not shown ADHD to be a myth. Rising referrals do not prove overdiagnosis. One disputed private assessment cannot estimate the accuracy of a market. Ordinary-sounding symptoms can form an enduring and impairing pattern. The absence of a diagnostic scan does not invalidate a clinical construct. Environmental influence and neurodevelopmental vulnerability can coexist.

It would now be tempting to move from that case has not been demonstrated to there is no case to answer.

The evidence does not permit that move either.

Diagnosis and prescribing have risen quickly. Thresholds can drift. Children’s relative age within a school year influences who is labelled. Social media spreads both recognition and misinformation. Private provision creates incentives that deserve audit. A major review in children and adolescents found evidence of overdiagnosis and overtreatment, especially at the milder edge.¹

At the same time, English health records show recognition below research prevalence estimates, long waits and striking inequalities by sex and age. Quiet children, women, older adults and people who compensate successfully may still be missed. Both errors can occur in the same system.

The honest question is not whether diagnosis is simply too high or too low. It is whether the right people receive the right conclusion for the right reasons — including no, another explanation and not yet certain.

What does “too much diagnosis” mean?

Several different claims are often compressed into overdiagnosis:

1. False-positive diagnosis: a person is labelled despite not meeting defensible criteria.
2. Threshold expansion: the criteria include milder difficulties than before.
3. Overmedicalisation: a real difficulty is framed too narrowly as an individual medical problem.
4. Overtreatment: diagnosis leads to intervention whose harms exceed its benefits.
5. Opportunity cost: assessment and treatment of lower-need people delays care for those at greater risk.

These can overlap but are not interchangeable. A mildly affected person may correctly meet current criteria yet gain little from diagnosis. A person with severe difficulties may be misdiagnosed because ADHD is not the best explanation. A valid diagnosis may lead to poor treatment, or no treatment at all.

Underdiagnosis also has several forms: never being referred, waiting without assessment, receiving the wrong alternative diagnosis, or meeting criteria but being unable to access treatment.

Counting labels cannot distinguish these processes.

The rise is real — and the measures are different

In England in 2025/26, NHS Business Services Authority statistics identified about 267,000 adults and 158,000 children who received community-dispensed central nervous system stimulants or other drugs used for ADHD. Those estimates rose by 40 per cent and 17 per cent respectively in one year.²

The figures are striking, but they are not a diagnosis census. They exclude private prescriptions and some other settings, include a drug group rather than a validated clinical sample, and tell us nothing directly about assessment quality.

Likewise, Channel 4’s 200 per cent rise in referrals between 2020 and 2025 measures people entering queues.³ Referrals can rise because previously missed people present, awareness changes, diagnostic thresholds shift, other services fail, or support becomes contingent on a label. The statistic describes demand, not its cause or eventual diagnostic yield.

The United States provides a useful warning about system effects. CDC data from a 2024 national parent survey estimate that 11.7 per cent of children aged 3–17 had a current ADHD diagnosis. State estimates ranged from 6 to 15 per cent; among diagnosed children, medication estimates ranged from 38 to 72 per cent.⁴

Such variation is unlikely to be explained by biology alone. Service structures, school expectations, public awareness, insurance, socioeconomic conditions and local practice affect who crosses the threshold. However, parent report of a professional diagnosis is not the same measure as a research assessment at one point in time. Variation is a reason to investigate, not by itself proof of error.

Evidence that some overdiagnosis occurs

The strongest case does not depend on one columnist’s assessment or an alarming graph.

Relative age

Children who are among the youngest in their school year are more likely than the oldest to receive an ADHD diagnosis and medication. A 2025 systematic review and meta-analysis estimated relative risks of 1.38 for diagnosis and 1.28 for medication, although results varied substantially between studies.⁵

Not every younger child diagnosed with ADHD is misdiagnosed. Relative immaturity can interact with genuine vulnerability, and the classroom influences impairment. Yet an administrative school cut-off should not alter underlying neurodevelopment. The finding is strong evidence that comparison groups and expectations influence referral and diagnosis.

Diagnosis rising faster than the measured phenotype

A Swedish population study found no increase from 2004 to 2014 in ADHD-like traits at the extreme end of the distribution, despite substantial growth in diagnosis and treatment. It did find small increases in normal and subthreshold traits.⁶

This pattern is compatible with better recognition of people who were previously missed. It is also compatible with a lower threshold. The study cannot allocate a percentage to either explanation, but it prevents us from assuming that all growth in recorded diagnosis reflects growth in the underlying phenotype.

Milder cases and diminishing returns

A 2021 systematic scoping review mapped 334 studies of children and adolescents onto a framework for detecting overdiagnosis. It found increasing diagnosis and treatment, evidence that additional cases were often milder, and what the authors called “convincing evidence” of overdiagnosis and overtreatment.⁷

That conclusion should not be softened away. It should also be read precisely. The review concerned children and adolescents, not the current expansion in adult diagnosis. It defined overdiagnosis partly by whether the net effects of diagnosis were unfavourable, not solely whether criteria were incorrectly applied. Only five studies directly examined benefits and harms among the additional, milder cases, and evidence quality was often limited.

The review establishes a credible warning at the milder edge. It does not estimate what proportion of current British adult diagnoses are false.

Evidence that substantial need remains unmet

A UK primary-care study covering 7.66 million people found that, in records from 2000 to 2018, 0.5 per cent had an ADHD diagnosis and 0.2 per cent received at least two ADHD-medication prescriptions within a year.⁸ Recognition rose over time but remained below conventional research prevalence estimates.

The NHS England ADHD Taskforce reported administrative prevalence of 2.55 per cent in boys and 0.67 per cent in girls, and 0.74 per cent in adult men and 0.20 per cent in adult women. It compared these with reference estimates around 3–5 per cent in children and 2–3 per cent in adults, concluding that clinician-defined ADHD remains under-recognised and undertreated in England.⁹

Those comparisons are imperfect. Primary-care records may omit diagnoses managed elsewhere. Administrative diagnosis, prescriptions and population research do not measure the same thing. Research estimates vary with criteria, informants and impairment thresholds.

Even allowing for those limitations, the English picture is not one in which everyone who plausibly has ADHD has already been diagnosed.

Under-recognition is also patterned rather than random. Traditional referral routes have more readily detected disruptive hyperactivity than quiet inattention. Girls and women may present with less visible symptoms, more internal distress or stronger social pressure to compensate. Adults with high ability, structured families or an occupational niche may build elaborate systems that conceal cost until parenthood, promotion, illness or loss of structure overwhelms them. Older adults and some ethnic minority groups face further barriers in recognition and access.

None of these features proves ADHD in an individual. They explain why absence of childhood diagnosis, school exclusion or career failure cannot be used as automatic exclusions.

Both errors can coexist

Suppose 100 people in a population genuinely have a condition, but only 50 are diagnosed. If ten of those 50 are false positives, the system is simultaneously underdiagnosing and overdiagnosing: 60 genuine cases remain missed while ten other people carry an incorrect label.

That is a plausible model for ADHD in Britain. The total recognised prevalence can be too low while allocation is poor.

One group’s unmet need does not validate another person’s assessment. One false-positive diagnosis does not erase somebody else’s disability. National under-recognition is not an answer to concerns about a cursory clinic; a poor clinic is not an answer to population evidence.

The distribution also matters. People able to pay, articulate an ADHD narrative and navigate a fragmented system may reach assessment sooner. Others with severe impairment may struggle to complete forms, attend appointments or advocate for themselves. A service can expand diagnosis and remain inequitable.

Social media: plausible mechanism, unknown diagnostic effect

An analysis of 100 popular English-language TikTok videos about ADHD classified 52 per cent as misleading, 27 per cent as personal experience and 21 per cent as useful.¹⁰ Short-form media rewards recognition and certainty. “You may have ADHD if…” travels more easily than an account of developmental onset, cross-setting impairment, competing explanations and uncertainty.

This can prime how somebody interprets and reports procrastination, forgetfulness or overwhelm. Algorithms repeatedly supply confirming material. Other explanations — depression, anxiety, trauma, sleep disorders, substance use or an intolerable environment — may receive less attention.

It can also help people recognise genuine, previously missed patterns, especially when their experience does not resemble the disruptive-boy stereotype.

The crucial evidence is missing. The TikTok study did not follow viewers through high-quality assessment. The NHS Taskforce found no good evidence for the proportion of social-media-inspired referrals that eventually do or do not meet criteria.¹¹

Social media therefore provides a plausible route to both recognition and misattribution. “They saw it online” is a reason for careful assessment, not a conclusion in either direction.

Adult assessment carries distinctive uncertainty

An adult clinician must reconstruct a childhood-onset pattern retrospectively. School reports may be lost, relatives unavailable and memory changed by the explanatory framework the person has encountered. Current inattention may reflect sleep deprivation, mood, anxiety, trauma, bereavement, substance use, menopause, burnout or several coexisting conditions.

Poor assessment can turn present distress and common traits into lifelong ADHD. Overly rigid assessment makes the opposite error: demanding documents no adult can produce, treating achievement as proof of normal function, or ignoring the cost of compensation.

Collateral information should be sought where proportionate and interpreted rather than worshipped. Absence of a school report is not evidence of absence; a relative’s recollection is not infallible. The clinician’s task is to test developmental continuity, pervasiveness, impairment and alternatives, then state the degree of confidence honestly.

Medication response cannot adjudicate the case. Benefit, no benefit or an unpleasant response does not itself prove or disprove ADHD.

Diagnostic humility should allow review. If the subsequent course does not fit — impairment remains confined to one context, another explanation becomes stronger, or treatment goals make no sense — clinicians should be willing to reconsider.

Diagnosis has become a gateway to ordinary support

Demand is partly created outside medicine.

When extra time, coaching, disability support or workplace adjustments require a formal diagnosis, people are rationally encouraged to seek one. Somebody just below a threshold may have almost the same functional need as somebody just above it but receive nothing. Clinicians are asked to convert a continuum into a ticket for resources.

This loads the assessment with educational, financial and identity consequences. Patients may fear that describing good days will cost them help. Services become invested in categorical decisions because funding is organised around labels.

Basic support should instead follow functional need. Sleep advice, organisational help, study skills, quieter working, movement breaks and inclusive teaching do not require a psychiatrist’s permission. Medication and some specialist interventions still require clinical diagnosis; ordinary adjustments often should not.

Decoupling support from diagnosis would not abolish ADHD. It could reduce pressure at the threshold and allow assessment to answer a clinical question rather than ration every form of help.

How would we know whether practice is improving?

Britain needs data capable of detecting errors in both directions:

1. Representative prevalence studies using consistent assessment of symptoms, childhood onset, impairment and alternatives.
2. Diagnostic-yield data from NHS and private services: ADHD, no diagnosis, alternative diagnosis and explicit uncertainty, stratified by demographic and referral variables.
3. Quality audits against agreed standards for developmental history, cross-setting evidence, impairment, differential diagnosis, physical assessment and documented reasoning.
4. Independent reassessment studies, blinded where possible to provider and payment route, to locate unstable practice and reasons for disagreement.
5. Severity trends, asking whether newly diagnosed groups are progressively less impaired and whether net benefit changes at the margin.
6. Long-term outcomes in function, education, work, relationships, mental health, adverse effects, identity and quality of life — not symptom scores alone.
7. Measures of missed need so that reducing apparent false positives is not achieved by excluding difficult-to-assess people.

No study will discover a perfectly objective line through dimensional traits. Good data can show whether the line is being applied consistently, transparently and to people likely to benefit.

So, are we diagnosing too much ADHD?

In some settings, groups and milder cases, almost certainly yes. The relative-age effect, international variation, threshold changes, social-media misinformation and the child overdiagnosis review make complacency indefensible. Private and NHS services should be audited rather than trusted by reputation alone.

The stronger claim — that Britain is experiencing a general epidemic of false ADHD diagnosis, or that most newly diagnosed adults do not have the condition — has not been demonstrated. Recorded recognition in England remains below reference prevalence estimates, with pronounced inequalities and extraordinary waits. The evidence needed to quantify false-positive adult diagnosis across providers is largely absent.

The most defensible conclusion is therefore deliberately untidy.

We are diagnosing more ADHD. We are probably diagnosing some people who should not receive the label while continuing to miss many who should. Thresholds may be falling in some settings while neglected presentations are finally recognised in others. Some people may be prescribed too readily while others cannot access treatment at all.

Pemberton is right that expansion deserves scrutiny. He has not shown that diagnostic error makes the condition mythical. Advocates are right that historic neglect created substantial unmet need. That does not make every new diagnosis self-validating.

The task is not to choose belief or disbelief. It is to build a system able to say yes, no, another explanation and not yet certain for defensible reasons — while helping people with real difficulties even when the answer is not yes.

The response to badly distributed diagnosis is neither simply more nor simply less.

It is better diagnosis, less label-dependent support and treatment proportionate to need.

---

Series disclosure: I am a doctor diagnosed with ADHD in adulthood and take lisdexamfetamine. This is context, not evidence; the clinical claims in this series should stand or fall on the sources cited.

Sources and further reading

Series navigation

← Previous: Article 5 — Medication: benefit, harm and choice · Series index · End of series