New Prevent guidance says there is no proven link between mental illness and radicalisation. It never mentions that autism now accounts for a record 18 percent of referrals, a pattern a government-commissioned review once called a misallocation of resources.
Estimated reading time: 7 minutes
In short. On 19 October 2026 the Department of Health and Social Care published new Prevent guidance for NHS mental health services, replacing NHS England's 2017 version [1]. The guidance states plainly that "there is no established causal link between having a mental health condition and radicalisation" and that the same is true of neurodivergence [2]. Both statements are accurate as far as they go. What the guidance does not say anywhere across its four published documents is that the Home Office's own most recent Prevent statistics, published 6 August 2026, record autism as a factor in 18% of all referrals, up from 14% in the release six months earlier [4] [5], or that the government's own 2023 independent review of Prevent found this exact pattern to be "a serious misallocation of resources" [6]. The 2017 document it replaces never mentioned neurodivergence once [7]. Two documented mistakes: four stars.
Picked from this week's GOV.UK publications feed: new DHSC guidance, verified live with no withdrawal notice, scoring on three criteria, sector weight across every NHS mental health trust in England, a live, data-backed controversy about referral proportionality, and a first-of-kind framework shipped with no equality check or review clause. This is a Tier 2 structured teardown: the guidance's own words tested against the statistics and the independent review that sit either side of it.
Four pages, a glossary, and a silence
The new guidance runs across a main document and three shorter role-specific summaries: "Purpose of this guidance", "Managing risk of radicalisation in mental health services", and separate one-page briefings for frontline staff and for Prevent leads [1]. The Department of Health and Social Care commissioned RAND Europe, the independent policy research organisation, to help produce it [1]. It is non-statutory, sitting alongside the legal duty under section 26 of the Counter-Terrorism and Security Act 2015 that requires NHS trusts and foundation trusts to have regard to preventing people being drawn into terrorism [3]. Staff are told to use the "Notice, check, share" framework, to record their reasoning whether or not they refer, and, critically, never to let a patient's mental state count against making a referral [2]. None of that is unreasonable on its own terms. The guidance's problem is not what it says. It is what two of its central claims leave out.
The claims, tested
| The guidance's claim | What we found | Verdict |
|---|---|---|
| "Within the general population, there is no established causal link between having a mental health condition and radicalisation... we must avoid stigmatising people with mental health conditions" and, separately, "there is no established causal link between neurodivergence and radicalisation" [2] | The Home Office's own statistics, published the same year, show 36% of all Prevent referrals (3,633 of 9,957) carried at least one recorded mental health or neurodiversity condition, and autism alone featured in 18% (1,833 of 9,957), up from 14% (1,226 of 8,778) in the release six months prior [4] [5]. The government's own 2023 Independent Review of Prevent found that exact pattern to be "a serious misallocation of resources" that "risks diverting attention from the threat itself" (page 12, para 1.9) [6]. Neither document is cited anywhere in the new guidance | Scientifically accurate, but delivered with none of the context that would tell a reader the guidance sits inside a pattern its own government has already flagged as a problem |
| The guidance "replaces NHS England's 2017" document [1] and introduces detailed new sections instructing staff to "engage your professional curiosity" about undiagnosed neurodivergence [2] | The 2017 predecessor, a 29-page PDF, contains zero mentions of neurodivergence or autism anywhere in its text [7]. This is genuinely new ground, not an update to existing provisions. Across all four new documents there is no equality impact assessment, no stated review date and no commitment to monitor the guidance's own effect on referral or Channel-adoption patterns for this group; the only "monitoring" mentioned anywhere concerns individual Channel case reviews, not the policy itself | A first-of-kind expansion into the cohort already showing the fastest-growing referral share, shipped with no mechanism to check what it changes |
A record that predates the guidance by three years
The Home Office did not invent this pattern in 2026. In February 2023, William Shawcross's statutory Independent Review of Prevent put it in the plainest terms available to a government reviewer: "In my assessment, Prevent is carrying the weight for mental health services. Vulnerable people who do not necessarily pose a terrorism risk are being referred to Prevent to access other types of much-needed support. This is a serious misallocation of resources and risks diverting attention from the threat itself" [6]. That finding sat behind the government's subsequent Prevent reforms. Three and a half years later, DHSC has published guidance specifically for the NHS staff who sit at the centre of that finding, and the guidance does not mention it once. The closest the document comes is telling staff that concerns "must not be dismissed or accounted for based on a person's mental health or neurodivergence" [2], an instruction that reads, in isolation, like the opposite of Shawcross's concern. There is a coherent case that under-referring genuine risk is worse than over-referring a false one; the guidance does not make that case, because it never acknowledges there is a case to make.
A category that did not exist in the old guidance
The 2017 NHS England document this guidance replaces ran to 29 pages and, checked directly by text extraction, used the words "neurodivergent", "neurodivergence" and "autism" zero times between them [7]. The new guidance gives neurodivergence its own dedicated section, its own glossary entries, and a specific instruction that staff should "engage your professional curiosity and consider whether someone may be neurodivergent, regardless of whether a formal diagnosis has been made" [2]. That instruction lands on a system where, by the Home Office's own account, autism's share of referrals has already grown by four percentage points in six months, and where autism's share rises further at each stage of the process, from 18% at referral to 23% once a case is discussed by a Channel panel and 24% once it is formally adopted [4]. Whether that rise reflects better detection, genuine risk, or a widening net is precisely the question an equality impact assessment exists to ask. This guidance does not include one.
Credit where due
The guidance is not careless about the science it does cite. Its statement that no causal link between mental illness, or neurodivergence, and radicalisation has been established is consistent with the published research literature, and the document repeatedly warns staff against treating a diagnosis as a shortcut for judgement: "each person should be considered on a case-by-case basis using... professional judgement", not a checklist [2]. It also recommends that senior staff, not just frontline workers, get deeper neurodivergence training, a proportionate response to its own observation that "awareness and understanding of neurodivergence may not be as widespread as it is with mental health conditions" [2]. And the Home Office's statistical bulletin, for its part, is unusually candid about the limits of its own data, noting that mental health and neurodiversity recording is "a new element on PCMT" and that "trends in this data over the next few years will need to be carefully monitored and interpreted as data coverage and quality improves" [4]. That caution exists somewhere in this story. It is just not in the document aimed at the staff making the referrals.
Verdict
Four stars, from two documented mistakes. This is not a finding that Prevent referrals from mental health settings are wrong, or that DHSC's new guidance gets the underlying science wrong. It is that a document built to reassure staff against stigmatising patients never once engages with the government's own evidence that the thing it is reassuring them not to worry about is already happening at a growing scale, and that the specific new framework it introduces for spotting undiagnosed neurodivergence ships with no mechanism anywhere to check what effect it has.
The star score counts two documented mistakes: the guidance's "no established causal link" framing for both mental health conditions and neurodivergence, presented without reference to the Home Office's own 6 August 2026 statistics (36% of referrals MH/ND-flagged, autism alone 18% and rising from 14% six months earlier) or the government's 2023 Independent Review of Prevent, which called the same pattern "a serious misallocation of resources" (page 12, para 1.9); and the introduction of an entirely new neurodivergence framework, confirmed absent from the 2017 predecessor guidance by direct text extraction, published with no equality impact assessment and no commitment to monitor its effect on referral patterns. Two falls in the 1 to 3 band: four stars; the bands are on the ratings page. Checked directly against all four published guidance documents, the Home Office's two most recent Prevent statistical releases, the 2017 NHS England predecessor PDF, and the 2023 Independent Review of Prevent PDF, fetched 20 October 2026. The document is tracked in the Daily Register; a revision will get a diff entry.
Sources
- Prevent: guidance for NHS mental health services, Department of Health and Social Care, published 19 October 2026
- Managing risk of radicalisation in mental health services, DHSC, published 19 October 2026
- Mental health professionals: your role under Prevent, DHSC, published 19 October 2026
- Individuals referred to and supported through the Prevent Programme, October 2024 to September 2025, Home Office, published 6 August 2026
- Individuals referred to and supported through the Prevent Programme, April 2024 to March 2025, Home Office, published 6 November 2025
- Independent Review of Prevent, William Shawcross, HC 1072, February 2023
- Guidance for mental health services in exercising duties to safeguard people from the risk of radicalisation, NHS England, published 2 November 2017