Homes, Not Hospitals: 2,125 People Still in Mental Health Hospitals

At the end of August 2026 there were 2,125 people with a learning disability and autistic people in mental health hospitals in England, and almost half of them had been in hospital for more than two years^[1]. Most are now autistic people who do not have a learning disability — the one group that has grown, not shrunk, since the programme to bring people home began.

These are not brief admissions for treatment that could not happen anywhere else. They are lives paused, often at the age when other people are starting careers and homes of their own. Many people arrive because community support ran out at the worst moment, then find the door back out jammed — not by their health, but by the absence of a house, a support team and a commissioner able to agree who pays. England has promised to end this four times in fourteen years.

Four national promises, and a count that has barely moved since 2020

England counts this group every month, which is unusual and useful: the Assuring Transformation collection, set up after the abuse uncovered at Winterbourne View hospital, asks commissioners to report every person with a learning disability or autism in inpatient mental health care^[1]. In December 2012 the Department of Health said everyone in hospital inappropriately would move to community support "no later than June 2014"^[3]. In 2015, Building the right support planned that "35% - 50% of inpatient provision will be closing nationally" by March 2019^[4]. The NHS Long Term Plan promised that by March 2023/24 provision would be "less than half of 2015 levels", with no more than 30 adults and 12 to 15 children per million in inpatient care^[5]. For 2025/26 NHS England asked for "a minimum 10% reduction", repeated in October 2025 as "a minimum 10% reduction year-on-year" to 2028/29^[6][7].

Against those promises, here is the series: 2,915 people in hospital in March 2015 and 2,245 in March 2026 on the latest revised figures — a fall of just under a quarter over eleven years, achieved almost entirely before 2020^[2]. A peer-reviewed analysis of the national data put it plainly: the 2015 target "and subsequent targets, were missed", and over eleven years "the planned closure of psychiatric inpatient beds has been unsuccessful"^[8].

Two cautions belong with that chart. The newest month is provisional: NHS England's own table of revisions shows March 2024 was first collected as 2,045 people and now stands at 2,365, so expect the latest count to rise^[2]. And this is the narrower of two counts: the Mental Health Services Data Set, which covers any mental health hospital, recorded 3,985 people in July 2026^[9].

The Long Term Plan's population targets are the clearest measure of the gap. NHS England's own rates, as first reported, give 39 adults per million in March 2024 against a target of no more than 30, and 19 under-18s per million against 12 to 15. By August 2026 the adult rate had risen to 42 per million^[2].

The group left behind is autistic people without a learning disability

The national picture is two trends moving in opposite directions. Inpatients with a learning disability and no autism diagnosis fell from 1,625 in March 2015 to 550 in August 2026. Over the same period autistic inpatients without a learning disability rose from 455 to 1,240, and their share went from 16% to 58%^[2]. NHS England cautions that autistic people were under-recorded before 2017, so the honest comparison starts there: 595 then, 1,240 now^[10][2]. An independent time-series analysis found the overall reduction came from "fewer inpatients with intellectual disabilities without autism over time, rather than fewer autistic inpatients without intellectual disabilities"^[11].

Among children it is starker. Of the 215 under-18s in hospital in August 2026, 200 were autistic without a learning disability^[2] — a shift NHS England's own board paper called "comparable to – and more profound than – the adult population"^[10]. A service model built in the 2010s around people with a learning disability is now mostly holding autistic people, many of them teenagers, and it was not designed for them.

People arrive in crisis, then stay for years

The data record why people were admitted, and more than one reason can be given. In August 2026, assessment or treatment of a mental health condition was recorded for 1,190 people (56%), harm to self for 585 (28%), risk of harm to others for 595 (28%), offending behaviour for 585 (28%) and suicidal thoughts for 280 (13%)^[2]. Most people come through a crisis door — and crises are what community services can prevent.

The average total length of stay in August 2026 was 1,679 days — about four and a half years^[2]. Almost half of everyone in hospital, 1,025 people, had been there more than two years, and 310 people more than ten^[1]. Long stays are not confined to people with a learning disability: among the 1,240 autistic inpatients without a learning disability, 465 had been in hospital more than two years and 100 for more than ten^[2].

Distance compounds it: 525 people (29%) were more than 50km from home in August 2026 and 240 were 100km or more away. Some 815 (38%) were on a secure ward, and 1,930 of the 2,125 were detained under the Mental Health Act — 725 of them under Part III, the criminal-court and prison-transfer route^[1][2].

Leaving is slow even when clinicians agree the person is ready: in July 2026, 405 people (10%) in the MHSDS count had a delayed discharge^[9]. In the August 2026 Assuring Transformation return, the commonest primary reason recorded for people clinically ready to leave but still in hospital was housing, the median wait since being declared ready was 284 days; across all inpatients, only 18% had a date planned for leaving and 26% had gone six months without a review of their care^[1][2]. Research with 27 people in long-stay hospitals found "widespread frustration" and environments people felt were "not conducive to getting/staying well"^[12].

Restraint, seclusion and segregation are part of the picture

National data on restrictive practice is patchy, and that is itself a finding. In the MHSDS for July 2026, 880 people with a learning disability and autistic people in hospital were restrained at least once, and the tables record 8,670 restrictive-intervention types for this group that month (one episode can involve more than one type); 135 people were secluded, 60 segregated, and 110 of those restrained were under 18^[9][13]. NHS England warns these are "an undercount" and advises against using them to derive rates of restraint per person, so we have not^[13]. In Assuring Transformation the restrictive-intervention fields were fully completed for only 665 of 1,880 inpatients (35%) in July 2026^[14].

Where someone has looked closely, autistic people are over-represented at the most restrictive end. The CQC's 2020 review Out of sight — who cares? met 66 people in long-term segregation: 42 were autistic, and an information request found 67% of people in long-term segregation on learning disability and children's mental health wards were autistic^[15]. Time in long-term segregation ranged "from three days to 13 years", almost 71% of those reviewed had been segregated or secluded for three months or longer, and most wards visited "were not therapeutic environments"^[15]. Five years on, CQC monitoring still describes wards that are loud, bright and short of quiet space, and the monthly average of recorded restrictive interventions across all mental health inpatients rose 24% in a year, to 16,462 in 2024/25, a rise the CQC says may partly reflect better reporting^[16].

Progress has been thin. In March 2022 the CQC judged 13 of its 17 recommendations "not achieved"^[17]. The review programme chaired by Baroness Hollins examined the care of 191 people in long-term segregation between November 2019 and March 2023, found 115 in it at the start of its second phase "and a similar number ... at the end", and recommended that solitary confinement become a "never event" for under-18s and whenever it lasts more than 15 days^[18]. Of the 88 people the CQC has reviewed since taking the programme on in 2024, 35 had left long-term segregation by March 2026^[19].

Two studies from outside England point the same way. Among 21,275 children admitted to one US paediatric hospital, those with an autism diagnosis had 2.3 times the odds of being restrained as matched children without one (95% CI 1.6 to 3.4), adjusting for reason for admission, length of stay and co-occurring diagnoses^[20]. In Ontario, among more than 32,000 children and young people assessed in mental health services from 2012 to 2022, 8.4% of those with an autism diagnosis had been physically restrained in the previous three days against 3.1% of those without (adjusted odds ratio 1.70, 95% CI 1.43 to 2.02)^[21].

Outside England, the counting mostly stops at the door

Few countries publish anything like England's monthly count, which makes comparison hard and England's data worth defending. Scotland's Coming Home report, in November 2018, found 705 people with learning disabilities placed out of their own area, 45% of them for more than ten years, and 67 people whose discharge was delayed; about half of those identified as a priority to return home were autistic, yet the survey excluded autistic people without a learning disability^[22]. A 2022 implementation report promised such placements and stays would be "greatly reduced" by March 2024^[23]. In June 2026, 67 adults with a learning disability were still delayed in hospital and 29 were in inappropriate out-of-area placements^[24]; a new action plan, published in March 2026, is meant to address the outstanding recommendations^[25]. Autistic adults without a learning disability are still not counted there^[24].

What the evidence asks of us

Treat admission as the last option, and check it in the numbers. NICE says to admit people with a learning disability and behaviour that challenges "only if assessment and care planning show that their needs cannot be met safely in the community", to place them "as close as possible to where the person usually lives", and to begin a discharge plan "as soon as the person is admitted", reviewed at least every three months^[27]. In August 2026 only 18% of people had a planned discharge date^[1] — a gap commissioners can be asked about locally, month by month, using this data.

Build the community side before the law changes. The Mental Health Act 2025 (Royal Assent 18 December 2025) will end detention for treatment under section 3 "on the basis of autism or learning disability" and limit Part 2 detention to 28 days for assessment^[28][29]. But those provisions were "not in force at Royal Assent", are commenced by regulations, and in September 2026 legislation.gov.uk still listed them as prospective; autism and learning disability also remain inside the Part 3 criminal-justice definition^[28]; with 725 of the 2,125 people detained under Part III^[2], changing section 3 alone will not empty these wards. The Act's other half matters as much: duties on commissioners to keep registers of people at risk and to "seek to ensure that the needs of people with autism or a learning disability can be met without detaining them" under Part 2^[28], backed by "an adequate supply of community services"^[29].

Unblock discharge, which is mostly housing and money. Housing is the commonest primary reason recorded for people clinically ready to leave who have not^[2], and the CQC found care for someone in prolonged seclusion or long-term segregation could cost £650 to about £2,300 a day, against £250 to £1,400 a day for the community placements it visited — figures it cautions are "not directly comparable", but which show hospital is not the cheap option^[15]. On the wider economics, see the lifetime cost of autism support and who pays.

Make restrictive practice visible, and count autistic people everywhere. A system that cannot say how often autistic people are restrained cannot show it is reducing it^[14]. England's split between autistic people with and without a learning disability is what revealed that progress had stopped for one group; Scotland's registers cannot show that^[22][24]. Where hospital care is genuinely needed, specialisation appears to help: a systematic review found admission was associated with improvements in mental health, and specialist intellectual disability services had "fewer out-of-area discharges and lower seclusion rates" than general wards^[30]. The goal is not that no autistic person ever receives hospital treatment. It is that nobody lives in a hospital for want of a home. Related reading: autism and intellectual disability, suicide and autism, and our Autism & Health research library.

Language: this piece uses identity-first language ("autistic people"), the preference of most autistic-led organisations, and clinical terms such as "autism spectrum disorder" or "learning disability" only when naming a dataset, category or quoted source.

Why we're publishing this

Health Insurance UK is a commercial health-insurance resource, not a charity or a campaign. We compiled this because the people doing the hard work — autistic people who have been on these wards, their families, and the charities and autistic-led organisations pressing commissioners and ministers — deserve the official numbers clean, current and in one place, ready to put in front of an MP or a commissioner. The argument is theirs; our part is to say plainly what England's own statistics show.

How to read this data

Two English datasets count overlapping but different groups: Assuring Transformation is a commissioner return covering specialist learning disability and autism inpatient care, while the Mental Health Services Data Set counts everyone flagged with a learning disability or autism in any mental health hospital. Each figure here is labelled with its source. Assuring Transformation counts are rounded to the nearest 5, percentages are calculated on unrounded figures, and the latest month is provisional and usually revised upwards; the rates per million use 2022 populations and are "as first reported". Two figures are our own sums of rounded counts: 465 autistic inpatients without a learning disability in hospital over two years (245 + 120 + 100) and 725 people detained under Part III (195 + 530). Restrictive-intervention figures are recorded counts that NHS England calls an undercount and says should not be turned into rates, so the comparative evidence here comes from published studies. The US and Ontario studies report odds ratios, not relative risks, and neither shows that being autistic causes restraint. Scottish and English figures come from different collections and are not comparable.

Use this data

Free to cite with attribution to Health Insurance UK. Charities, campaigners and journalists are welcome to reuse these figures and charts. For the underlying figures as a spreadsheet, get in touch.

Important: This article summarises published research and official statistics for general information. It is not medical or legal advice and cannot be used to diagnose, treat or predict anything about an individual. If you have concerns about your own or someone else's health, please speak to a qualified health professional.

Sources

1. NHS England. Statistics on people with a learning disability and autistic people in mental health hospitals, AT: August 2026, MHSDS: July 2026 — summary findings from the AT dataset. NHS England Digital, 17 September 2026. digital.nhs.uk/…/at-august-2026-mhsds-july-2026
2. NHS England. Assuring Transformation data tables and CSV, August 2026 (tables 1.1, 2.2, 2.3, 2.7, 3.3, 4.1, 4.6, 4.8, 5.8, 6.2). NHS England Digital, 2026. files.digital.nhs.uk/…/LD_AT_DataTables_Aug2026.xlsx
3. Department of Health. Transforming care: a national response to Winterbourne View hospital. GOV.UK, 10 December 2012. gov.uk/…/winterbourne-view-hospital-department-of-health-review-and-response
4. NHS England, Local Government Association and ADASS. Building the right support: a national plan to develop community services and close inpatient facilities. NHS England, 30 October 2015. england.nhs.uk/…/ld-nat-imp-plan-oct15.pdf
5. NHS England. The NHS Long Term Plan, chapter 3, paragraph 3.34. NHS England, 2019 (archived). webarchive.nationalarchives.gov.uk/…/learning-disability-and-autism
6. NHS England. 2025/26 priorities and operational planning guidance. NHS England, 30 January 2025. england.nhs.uk/long-read/2025-26-priorities-and-operational-planning-guidance
7. NHS England. Medium Term Planning Framework — delivering change together 2026/27 to 2028/29. NHS England, 27 October 2025. england.nhs.uk/long-read/medium-term-planning-framework-delivering-change-together-2026-27-to-2028-29
8. Nisar A, Thompson PA, Boer H, Al-Delfi H, Langdon PE. Factors affecting psychiatric bed utilisation by people with intellectual disabilities: a time series analysis using the English national Mental Health Services Data Set. Journal of Intellectual Disability Research, 2025. europepmc.org/article/MED/40545879
9. NHS England. Statistics on people with a learning disability and autistic people in mental health hospitals — summary findings from MHSDS, July 2026. NHS England Digital, 17 September 2026. digital.nhs.uk/…/summary-report---mhsds
10. Cahill T. Update on Learning Disability and Autism Programme (board paper, 23 September 2025). NHS England, 2025. england.nhs.uk/long-read/update-on-learning-disability-and-autism-programme
11. Nisar A, Thompson PA, Boer H, Al-Delfi H, Langdon PE. Specialist psychiatric bed utilisation by people with intellectual disabilities and autistic people: a time-series analysis using the English Assuring Transformation dataset. Journal of Intellectual Disability Research, 2025. europepmc.org/article/MED/40506663
12. Glasby J, Miller R, Glasby AM, Ince R, Konteh F. 'Why are we stuck in hospital?' Barriers to people with learning disabilities/autistic people leaving 'long-stay' hospital: a mixed methods study. Health and Social Care Delivery Research, 2024. europepmc.org/article/MED/38343118
13. NHS England. Learning disability and autism MHSDS data tables, July 2026 (tables 17 and 18, restrictive interventions). NHS England Digital, 2026. files.digital.nhs.uk/…/LD_MHSDS_DataTables_Jul2026.xlsx
14. NHS England. Assuring Transformation restrictive intervention data quality report, August 2026. NHS England Digital, 2026. files.digital.nhs.uk/…/Restrictive Intervention (AT) Data Quality report - August 2026.xlsx
15. Care Quality Commission. Out of sight — who cares? A review of restraint, seclusion and segregation for autistic people, and people with a learning disability and/or mental health condition. CQC, October 2020. cqc.org.uk/publications/themed-work/rssreview
16. Care Quality Commission. Monitoring the Mental Health Act in 2024/25. CQC, 29 January 2026. cqc.org.uk/publications/monitoring-mental-health-act/2024-2025
17. Care Quality Commission. Out of sight — who cares? Restraint, segregation and seclusion review: progress report. CQC, March 2022. cqc.org.uk/…/20220325_rssreview-progress-march_print.pdf
18. Hollins S. Baroness Hollins' final report: my heart breaks — solitary confinement in hospital has no therapeutic benefit for people with a learning disability and autistic people. Department of Health and Social Care, 8 November 2023. gov.uk/…/baroness-hollins-final-report
19. Care Quality Commission. Learning from Independent Care (Education) and Treatment Reviews in 2024 and 2025. CQC, 2026. cqc.org.uk/publications/icetr/2024-2025
20. Calabrese ME, Sideridis G, Weitzman C. Physical and pharmacologic restraint in hospitalized children with autism spectrum disorder. Pediatrics, 2024. europepmc.org/article/MED/38073325
21. Stewart SL, Drew AL, Fearon D. Risk factors for physical restraint use with children and youth in community and inpatient mental health settings in Ontario. Child and Adolescent Psychiatry and Mental Health, 2026. europepmc.org/article/MED/41992347
22. MacDonald A. Coming home: complex care needs and out of area placements 2018. Scottish Government, 22 November 2018. gov.scot/publications/coming-home-complex-care-needs-out-area-placements-report-2018
23. Scottish Government. Coming Home Implementation: report from the Working Group on Complex Care and Delayed Discharge. Scottish Government, 21 February 2022. gov.scot/publications/coming-home-implementation-report-working-group-complex-care-delayed-discharge
24. Public Health Scotland. Insights into learning disabilities and complex needs: statistics for Scotland, 22 September 2026 (census at 25 June 2026). Public Health Scotland, 2026. publichealthscotland.scot/…/22-september-2026
25. Scottish Government. Coming Home Action Plan 2026. Scottish Government, 19 March 2026. gov.scot/publications/coming-home-action-plan-2026
26. Lloyd-White S, Borschmann R, Rorke A, Stark E, Combe G. Demographics and clinical outcomes of autistic children and adolescents receiving psychiatric in-patient care in the Thames Valley from 2019 to 2024: retrospective cohort study. BJPsych Open, 2026. europepmc.org/article/MED/42494208
27. National Institute for Health and Care Excellence. Learning disabilities and behaviour that challenges: service design and delivery (NG93), recommendations 1.8.1 to 1.8.10. NICE, 28 March 2018. nice.org.uk/guidance/ng93/chapter/Recommendations
28. Mental Health Act 2025 (c. 33), section 3, Schedule 1, section 4 and section 56. legislation.gov.uk, Royal Assent 18 December 2025. legislation.gov.uk/ukpga/2025/33/crossheading/autism-and-learning-disability
29. Department of Health and Social Care. Mental Health Act 2025: explanatory notes. legislation.gov.uk, 2025. legislation.gov.uk/ukpga/2025/33/pdfs/ukpgaen_20250033_en.pdf
30. Melvin CL, Barnoux M, Alexander R, et al. A systematic review of in-patient psychiatric care for people with intellectual disabilities and/or autism: effectiveness, patient safety and experience. BJPsych Open, 2022. europepmc.org/article/MED/36268640