One in four
pupils giving a
yes-or-no answer
called themselves
neurodivergent —
and were
more likely to
vape, smoke
and use drugs

One in four, and a warning about substances

The OxWell Student Survey is a large school-based survey of young people’s wellbeing. Its 2023 wave, drawn mainly from four areas of England, asked secondary school pupils whether they considered themselves neurodivergent. In a paper published in Child: Care, Health and Development in September 2026, John Headley Ward, Francesca Happé and Mina Fazel analysed the answers of 21,749 of those pupils against their reported use of vapes, cigarettes, alcohol and illicit drugs.

Two findings come out of it. The first is the one the paper was designed to test: pupils who identified as neurodivergent were more likely to be in the higher-risk categories for every substance studied. The second is the number that frames everything else. Of the pupils who answered yes or no, 25.1% said yes.

What the substance findings show

The headline is reassuring before it is worrying. Frequent substance use was generally low overall — the authors put it under 10%. But the gaps between the two groups were consistent.

Two things are worth drawing out. The first is how much the gap shrinks once other factors are accounted for. Adjusting for gender, age, ethnicity, socioeconomic status and family history of substance use cuts the odds ratios substantially; adjusting further for wellbeing and anxiety and depression scores shrinks them again. They stay significant — for these three measures, identifying as neurodivergent is associated with higher odds even after those adjustments — but a fourfold difference becomes something closer to twofold. The paper’s conclusion says neurodivergent young people are “three times more likely” to be in the high-use group; that is the unadjusted picture. For one alcohol measure, the number of days drinking, the difference was no longer significant once psychological measures were added.

The second is what the shrinkage suggests. The authors point to self-medication, discussed in both the ADHD and autism literature. In their limited data on why pupils vaped, drawn from free-text answers, 7.8% of neurodivergent pupils who had ever vaped, against 2% of their peers, said they did so to ease mental health symptoms, particularly stress and anxiety. The fact that the association weakened once wellbeing and anxiety/depression were accounted for, they write, “also lends weight to this hypothesis.” They call their overall findings “preliminary evidence”, and recommend school-based prevention designed to work for neurodivergent young people.

What the question actually asked

The finding is only as clear as the category it rests on, and the category is worth reading exactly. Pupils were asked:

Do you consider yourself to be dyslexic/dyspraxic, and/or autistic, and/or have ADHD (i.e., neurodivergent)?

Four labels, one box, and the operative words are consider yourself. No diagnosis was asked for, and the survey did not ask which of the four applied. Pupils who answered “not sure” or “prefer not to say” were excluded from the analysis. The 25.1% is a share of those who answered yes or no; counting everyone who answered the question, it is 19.4%.

The authors are open about what this means. The figure is “much higher” than the prevalence of ADHD and autism, which they put at around 5%, or of dyslexia and dyspraxia, estimated at 10% and 6%. Among the possible reasons their EHCP analysis found no effect, they list the construct validity of the survey’s questions on neurodiversity and EHCPs. And they note that because specific conditions were not asked about, they cannot say whether the substance association belongs to one or more of the groups — ADHD, for instance, where impulsivity is more prominent than in dyslexia.

The support data sharpen the picture. The survey asked whether pupils received support for special educational needs, such as an Education, Health and Care Plan. Only 28% of pupils who identified as neurodivergent said yes, yet they made up 70% of all pupils who did. In an exploratory analysis, receiving that support made no significant difference to the substance-use association.

What a self-applied label is measuring

A quarter of respondents is a large share — though the authors’ own introduction puts the proportion of the population identifying as neurodivergent at 15–20%, and the separate prevalence estimates they cite overlap. It is a large, mixed population, united by a word that, as a JAMA Psychiatry scoping review has already mapped, the research literature uses to denote very different things. In this survey it gathers pupils with a diagnosed condition, pupils who recognise themselves in one, pupils for whom the word names a real difficulty that no service has assessed, and, plausibly, some for whom it names something else: distress, a sense of not fitting, a vocabulary picked up because it was the one available.

That is not a reason to dismiss the finding. The study’s own logic points the other way. If a quarter of pupils see themselves in this category, most of them reporting no special educational needs support, and that group is more likely to be vaping, smoking and taking drugs, then something real is being picked up. What the study cannot tell us is what. The association survives adjustment for wellbeing and anxiety/depression, but shrinks under it, which is what you would expect if the word is partly carrying distress the other measures only half capture.

There is a pattern here worth naming. As self-identification claims multiply — some of them, as this Directory has noted, marketing-driven — research instruments are starting to use the word itself as the variable. That makes the findings easier to collect and harder to interpret. A category defined by self-recognition, bundling four conditions under one tick-box, is likely to produce associations with much that tracks struggle at school — if, as seems plausible, struggle at school is part of what the word has come to mean to the young people choosing it.

Prevention for a population nobody has defined

The authors’ recommendation follows from their data: scalable, universal school-based approaches to substance misuse that incorporate neurodiversity-affirming principles, co-designed with pupils, parents and teachers — while noting that the evidence for universal school-based prevention is currently limited. Their own figures show why it has to be universal. Prevention aimed at “neurodivergent pupils” is aimed at a quarter of a school, defined by a question each pupil answers for themselves, most of whom report no special educational needs support and, for all the survey can tell, no assessed need. The only intervention that can reach a group that size is one that changes the environment for everyone.

The structural point goes one step further. When a label grows to cover a quarter of a cohort, it stops describing a minority to be supported and starts describing the conditions the whole cohort is growing up in. The finding that these pupils vape and drink more is a warning worth heeding. The finding that one in four of them reached for the word in the first place is the one that ought to trouble the schools.

Citations

Ward, J. H., Happé, F. G. E. and Fazel, M. S. (2026) — Are School-Aged Children Who Self-Identify as Neurodivergent at Greater Risk of Substance Misuse? A Study of 20 000 Secondary School Students — Child: Care, Health and Development