Skip to content
The Neurodiversity Directory

Autism
and ADHD
overdiagnosis:
what is actually
being counted?

Are autism and ADHD overdiagnosed? Three BJPsych Bulletin opinion pieces, read together, suggest the question can’t yet be answered.

Three opinion papers in BJPsych Bulletin, and only one of them uses the word “overdiagnosis”

In October 2026, BJPsych Bulletin, which Cambridge University Press publishes on behalf of the Royal College of Psychiatrists, carried three short papers on autism and ADHD. All three are labelled “Opinion” on their first page. None reports new data. They were accepted on the same day, 28 September 2026, and went online between 21 and 26 October. Nothing in the papers says they were commissioned together or belong to a themed set, and this piece does not treat them as one.

The argument they land in is whether autism and ADHD are being overdiagnosed. In England it has run through the independent review into mental health conditions, ADHD and autism, which the Directory has followed as the Fonagy Review. As our piece on Fonagy’s evidence to MPs on overdiagnosis reported, the review’s chair told MPs the review had no evidence of overdiagnosis at scale, and also that a warning sign sat in the review’s own figures. Our reading of the Fonagy Review’s final report argued that the report finds the diagnosis gate on nearly every page.

Only one of the three papers uses the word. Allan House, Emeritus Professor of Liaison Psychiatry at the University of Leeds, puts overdiagnosis in his title and asks whether the late philosopher Ian Hacking can help with the neurodiversity and overdiagnosis debate. Aslı Kazgan Kılıçaslan, an associate professor in the Department of Psychiatry at Buca Seyfi Demirsoy Training and Research Hospital in İzmir, Türkiye, never mentions it. Kazgan Kılıçaslan proposes treating ADHD as a “neurobiological spectrum”. Nor do Sima Rafiei and three colleagues from Warwick Medical School and the Coventry and Warwickshire health system, whose subject is what England’s routine pathway data can and cannot see.

Read together, though, the three do something none of them sets out to do. One says the category has moved so far that nobody is sure what it names. One says it is a line drawn across a continuum, placed by a judgement about function. One says national reporting cannot yet reliably count how often an ADHD assessment ends in a diagnosis. An overdiagnosis debate needs an agreed object to be over-diagnosed. On the evidence of these three papers, there isn’t one.

About these sources

All three papers are labelled “Opinion” by BJPsych Bulletin and are open access under a CC BY 4.0 licence. None reports original data. House (Leeds) writes alone and declares no interests and no funding. Kazgan Kılıçaslan (İzmir) writes alone, makes no funding or interest declaration, and acknowledges using ChatGPT (OpenAI) for language editing. Rafiei, Cartwright, Atkins and Toro (Warwick, Coventry and Warwickshire) declare that two of them hold roles in organisations involved in local neurodevelopmental pathway discussions. Hacking’s own writings are reported here only as House presents them.

Allan House reads Ian Hacking: making up people, looping effects and an “advocacy disorder”

House opens with the review’s “tricky topics”, among them criteria loosened so that they draw in “too many people who may thereby be disadvantaged”. That is his summary of the questions before the review, not a finding. His final revision is dated 18 August 2026, and the only report from the review he discusses is the interim one.

As House reports him, Hacking called the creation of categories of people “making up people”: a dynamic process running between the people classified, those close to them, and the experts and institutions around them. Because people respond to being categorised, Hacking called them “interactive kinds”. He used the term “looping effect” for the way naming people in a category can change both the people and the category’s defining features. House is explicit that Hacking “was clear that all this was compatible with people in the category having a disorder that has a biological basis.” A looping effect does not say the condition is invented. It says the category moves.

In 2006, as House reports, Hacking wrote that “autism has become an advocacy disorder, which contributes to the increased rate of diagnosis”. House also reports him calling autism, among other things, a “bureaucratic concept used in the administration and management of awkward school children”. Advocacy that described autism as a difference rather than a disorder he read as the final step in making up people, which he called resistance or reclaiming an identity. House adds that Hacking was writing “at a time when the quoted prevalence was less than half what it is now, when autism in girls was considered a rarity”.

Then the spectrum. In House’s account, “an unhelpful professional contribution came with the idea of the autistic spectrum.” It carried “an essentialising assumption that everybody identified by the presence of any feature regarded as autistic must necessarily share some fundamental features.” The paper’s verdict follows: “This is most unlikely to be true”. Hacking preferred an autistic space or manifold, with no need to assume that those with little or no disability and those more incapacitated had the same disorder. He was worried, House writes, that the spectrum metaphor “may be actively harmful by leading to unnecessary diagnosis.”

For our present purposes, looping effects have had such a profound impact that we are no longer clear what we are talking about when we talk about autism.

Allan House, BJPsych Bulletin (2026)

That sentence is House’s, not Hacking’s, and it is the centre of the paper. If an important function of the category is to inform how bureaucratic responses are designed, he argues, “we are bound to be defeated by this degree of uncertainty about the nature of group membership”.

The turn comes next. “Looping is a two-way street”, House writes. Public discussion of cultural and social influences on autism diagnosis rates, he says, has tended to focus on advocacy by those diagnosed. Hacking’s ideas about looping should prompt a more thorough look at “the role of those who claim professional expertise”. House points to remarks by Uta Frith, as he reports them, about the over-ready acceptance of ideas like masking or camouflaging and autistic burnout.

The solutions House draws from Hacking are three, and modest: move from a single spectrum to a space in which clusters of people share some autistic characteristics and may differ in every other respect from another cluster; look harder at professional expertise, not only at advocacy; and attend to how the debate is conducted, not just what it is about. Without that, he ends, “we are unlikely to achieve the rapprochement necessary for a satisfactory outcome for all concerned.”

Aslı Kazgan Kılıçaslan’s calibration perspective: ADHD as a continuum with a line drawn at impairment

Kazgan Kılıçaslan starts from the other end of the problem. ADHD traits, the paper argues, are continuously distributed in the population, and “clinical ADHD represents the end of this distribution associated with functional impairment.” Diagnosis keeps its job: it is a categorical decision taken across dimensional biology.

The paper grounds the continuum in other people’s work, reported here as the paper presents it. A population-based genome-wide meta-analysis reported an estimated genetic correlation of 0.96 between continuous ADHD symptom scores and case–control ADHD. In one large adolescent sample, striatal connectivity varied with symptom scores while the categorical comparison of ADHD and controls was not significant.

The paper’s analogy is blood pressure: “Blood pressure is a parameter present in everyone but, when it causes organ damage, it becomes a disease requiring intervention.” Kazgan Kılıçaslan guards it at once. “Since there is no single objective biomarker in ADHD corresponding to blood pressure, the clinical threshold cannot be determined by symptom count alone”.

“Calibration” is the paper’s name for graded differences in regulating attention, arousal, motivation and cognitive control, and their context-dependent expression. The paper’s “neurobiological spectrum” means graded, partly overlapping variation “rather than to a single biological axis.” Kazgan Kılıçaslan concludes that “current evidence does not support reducing calibration to a single biological mechanism.”

The clinical heart of the paper is one sentence: “Within this framework, functional impairment is the bridge between dimensional biology and categorical clinical decision-making.” Two things follow. “Two individuals with similar levels of liability may have different functional outcomes”: a structured setting with rapid feedback may reduce some difficulties, while prolonged monotonous attention or delayed reward may make them more pronounced. And “a person may show symptoms below the diagnostic threshold and still require support.”

The paper draws its own limits. Those contextual differences “do not mean that the diagnosis is solely a product of the environment”. Kazgan Kılıçaslan calls the framework “not a new definition of the disorder”. On medication, the paper follows current clinical guidelines as it reports them, citing NICE: medication for adults whose symptoms still cause significant functional impairment in at least one area of life despite environmental modifications.

Against a single spectrum: House, reading Hacking, on autism

The spectrum carries an assumption that everyone with any autistic feature shares some fundamental features, which the paper calls most unlikely to be true. Hacking preferred a space of clusters that may differ in every other respect from one another, and worried, House reports, that the spectrum metaphor may lead to unnecessary diagnosis.

For a dimensional spectrum: Kazgan Kılıçaslan, on ADHD

ADHD traits are continuously distributed, and clinical ADHD is the functionally impairing end. The spectrum is graded, partly overlapping variation, not a single biological axis, and the diagnostic line is drawn at functional impairment.

Staging these two as a fight would be false balance. They write about different conditions, and Kazgan Kılıçaslan’s refusal of “a single biological axis” already concedes part of what Hacking objected to. Where they part is on what the word “spectrum” does. For House’s Hacking, it essentialises. On our reading of Kazgan Kılıçaslan, it does the opposite: it turns a kind into a degree.

Rafiei and colleagues: a system that can see its backlog but not what its diagnoses mean

The Warwick paper asks a prior question: could anyone tell? The authors report national figures showing more than 200,000 people waiting for autism assessment and large numbers waiting for ADHD assessment. Backlog data alone, they argue, cannot show how people move through referral, assessment, support and follow-up, and the result is “pathway blindness: systems can see pressure, but not flow, fairness or diagnostic visibility.”

For anyone in that queue, the routes are set out in the Directory’s guides to getting an ADHD diagnosis in the UK and getting an autism diagnosis in the UK, and providers are listed under Assessments.

On the paper’s account, community paediatric autism assessments cannot yet be identified consistently in the Community Services Data Set, and so are excluded from the national autism waiting time statistics. For ADHD, national reporting cannot identify everyone waiting specifically for an ADHD assessment. It does not fully capture Right to Choose or private provider activity where providers do not report to the national mental health data set. And it “cannot yet reliably report how many people receive an ADHD diagnosis following assessment or what treatment they subsequently receive.”

For autism, the paper notes, NHS England already recommends that services consider conversion rates, the proportion diagnosed following assessment. For ADHD the equivalent rate is a national reporting priority, “although incomplete recording has limited its reliable calculation.” The number an argument about overdiagnosis would most obviously need is one the paper says national ADHD reporting cannot yet reliably produce. That inference is ours; the authors make no claim about overdiagnosis either way.

They go further, into what a diagnosis records. “A recorded ADHD diagnosis does not by itself describe symptom profile, functional impact, support need or treatment stage.” DSM-5-TR has mild, moderate and severe specifiers; ICD-11 distinguishes presentations without an equivalent severity framework; and “Neither classification provides a continuous severity scale as part of the diagnostic classification itself.” Current data, their summary adds, “struggle to distinguish confirmed diagnosis, suspected diagnosis, disclosure, self-identification and co-occurring needs.”

Without pathway visibility, Rafiei and colleagues argue, differences in recorded diagnosis cannot readily be told apart from differences in access, referral or progression.

Their remedy is unglamorous: a six-month pilot that Integrated Care Systems could run in three phases — map the intended, experienced and data pathways; test a minimum viable data set built first from fields existing data sets can already record; then use it to test a change aimed at one or two priority problems. They are explicit that this “should complement, rather than replace, action on workforce, capacity and waiting times.” The Directory made a cruder version of the same complaint with its own Freedom of Information map in The NHS won’t show you this data — so we will.

Where Hacking’s loop complicates the Directory’s own account

The Directory has argued positions close to all three papers, and some now need qualifying.

Start with the July piece on the real driver of autism and ADHD overdiagnosis. It argued that tightening criteria treats the problem as measurement error, and that the larger engine is upstream: neurodivergent became a desirable identity once “divergent” flipped from deviance to distinction. In The Neurodiversity Book that is the third of three doors into the same room, beside the constitutional and the environmentally shaped. Hacking, as House presents him, saw something of the same shape two decades earlier: advocacy as part of the loop, with reclaiming an identity as its final step. The third door has a philosophical ancestor, and the corpus should say so.

But House’s emphasis cuts against the July headline. Looping is a two-way street, he says, and public discussion has looked mostly at advocacy, not at those who claim professional expertise. A piece titled “the real driver” put the weight on the identity side. The book’s chapters on why the medical model arose when it did, and on the economics of the diagnostic loop, carry the other side: an apparatus that sorts, bills and renews. The honest position is that identity is one arm of a loop, not its driver, and the book’s own analysis was always closer to that than the July title was.

Second, the spectrum. The book has been hard on the word, arguing that once autism became a spectrum, its grammar left nowhere to stand that was not somewhere on it. House’s Hacking agrees for a different reason. But the book’s chapter on universal cognitive mechanisms does its own ordering, unfurling the criteria into shared axes with ADHD and autism far out toward opposite poles. Hacking’s manifold is a warning to that scheme too. If autistic people fall into clusters that may differ in all other respects from one another, no single pole describes them. The book’s saving clause is that its axes are independent, which is why it reads AuDHD as arithmetic. Still, wherever the corpus slides into treating autism as simply the narrow pole, House’s reading of Hacking catches it.

Third, Kazgan Kılıçaslan, who agrees with the corpus that a diagnosis names a threshold crossed on a continuum. Our piece on ADHD as a construct rather than a natural kind said much the same. The paper’s account of context is close to what the corpus calls the demand structure, and its line that regulatory efficiency “may vary as task demands increase rather than reflect a fixed capacity deficit” could sit in the book’s chapter on the moving baseline. But the paper draws limits the corpus sometimes doesn’t: its continuum is genetic as well as behavioural, and clinically significant impairment can warrant medication, not merely mark a shortfall against someone else’s schedule. The blood-pressure analogy is where to argue with Kazgan Kılıçaslan. As the paper itself says, ADHD has no equivalent biomarker, so impairment has to be judged, and judged against particular demands. That is where that framework and ours meet and pull apart.

No agreed object: why the overdiagnosis argument can’t be settled with the instruments on the table

Ask what overdiagnosis would have to mean. More diagnoses than the true number of people with a fixed condition? House’s Hacking says the category is not fixed, because the category and the people in it move each other. More than the number impaired enough to cross the line? Kazgan Kılıçaslan sets the line at functional impairment, whose expression, the paper says, context can modify. More than some expected rate of diagnosis after assessment? Rafiei and colleagues say that rate cannot yet be reliably calculated for ADHD from national data. Each paper removes a different footing the debate needs, without setting out to.

That is how the review’s chair could tell MPs, as our earlier piece reported, both that the review had no evidence of overdiagnosis at scale and that, unless the system is changed, it is almost inevitable that the problem of overdiagnosis will arise. The word is being asked to carry a judgement none of the instruments in use was built to make. The code records that a threshold was crossed, not why or how far; the waiting list records pressure, not flow. And the category holds everyone who crossed under it, constitutional, environmentally shaped and identity-adopted alike, with no field to tell them apart.

Call this the instrument encoding the verdict; these papers sharpen it. The question is not only what the instruments find, but whether the thing being argued about holds still long enough to be measured. Here, it doesn’t. A debate that keeps asking whether there are too many or too few is rating how people are faring as if that settled who they are.

So the useful question is not whether autism and ADHD are overdiagnosed. It is whether anyone in the system will build what would make that answerable: pathway data that shows where people enter, wait, leave and end up; a record of severity and impact that a diagnostic code does not provide; and a classification that admits, as House’s Hacking does and as Kazgan Kılıçaslan’s refusal of a single biological axis partly does, that one name may cover more than one thing. Until then, every confident claim of overdiagnosis, and every confident denial, is a claim about an object nobody has agreed to define. That includes the claims made by people who share the Directory’s view.

Citations

House, A. (2026) — Making up autism: can reading philosopher Ian Hacking help us with the neurodiversity and overdiagnosis debate? — BJPsych Bulletin

Kazgan Kılıçaslan, A. (2026) — Reconceptualising ADHD as a neurobiological spectrum: calibration-based perspective — BJPsych Bulletin

Rafiei, S., Cartwright, G., Atkins, B. et al. (2026) — From backlog to blind spots: improving data visibility across autism and ADHD pathways in England — BJPsych Bulletin

News and updates.

Straight to your inbox.