Restraints used on people with learning disabilities and autistic people in inpatient units - what do seven years of statistics tell us?

Before I start, a quick warning. Most of my blogposts these days run through official statistics to try and build a picture of what's happening for (to?) people with learning disabilities. Through the distancing veil of numbers and graphs, this picture is usually quite upsetting. I found putting together the graphs for this blogpost more upsetting than any I've done recently - just thought I'd let you know this in advance when you're deciding whether to read it or not.

This blogpost will run through what the publicly available statistics say about what restraints (or 'restrictive interventions') autistic people and people with learning disabilities in mental health inpatient services in England have been subject since this information became publicly available around 7 years ago. I haven't looked properly at this information for the best part of five years, and the picture, in becoming more comprehensive, has become even more grim.

The dataset I used in that blogpost and in this one is the Mental Health Services Data Set (MHSDS) - this is a large dataset collected by all NHS-commissioned mental health services in England, and includes information on people flagged in these services as an autistic person or a person with learning disabilities. NHS England/DHSC extracts the information specifically related to people with learning disabilities and autistic people in any mental health inpatient service every month, including both the 'specialist' inpatient services that were largely the focus of Transforming Care/Building The Right Support, and general mental health inpatient services. Providing information for the MHSDS is mandatory for NHS funded service providers, including independent sector providers. The number of people with learning disabilities and autistic people in inpatient services according to the MHSDS is much higher than the number of people recorded in the Assuring Transformation dataset, which the government prefers to use when quoting how well it's doing.

There are going to be a lot of graphs. When going through them, I think there are at least three big things that have had an impact on the number of restraints reported in these mandatory statistics:

1) The Covid-19 pandemic. For reasons that will become clear later, it's hard to be completely confident in trends over time in the restraints data, but it looks like during the Covid-19 pandemic (at its peak in 2020-2021) there were fewer people being admitted to inpatient units and those people who were there were more likely to be subject to some types of restraint.

2) In 2022 and 2023, a reported cyber attack had an impact on the quantity and quality of data reported in the MHSDS, with some national information not reported for around a year. This means there are some gaps in what is reported, and some of the numbers might be less reliable than usual.

3) The completeness of reporting. When I first started looking at this information in 2020 and 2021, it was clear that many big inpatient service providers (including some very big independent sector inpatient providers) were routinely not reporting any restraints at all. The Mental Health Units (Use of Force) Act 2018 was passed into law in 2018 to improve this state of affairs, but statutory guidance from DHSC was not issued until December 2021. This means that, if reporting becomes more comprehensive, we should see this reflected in the restraint statistics. 

How many people with learning disabilities and autistic people are in inpatient services?

For pretty much all the graphs in this blogpost, I'll be sharing information from the MHSDS for every six months over a seven-year period, from March 2019 to March 2026. The first graph below shows information on the number of people with learning disabilities and autistic people in inpatient services at the end of the month in question, divided into people in independent sector services (the light blue line), NHS services (the dark blue line), and finally the total number of people in inpatient services (the black line).

The numbers here (typically 3,500-4,000 people) are much higher than those reported retrospectively by commissioners in the Assuring Transformation dataset (typically 2,000-2,500 people), possibly because the MHSDS is much better at identifying those flagged as a person with learning disabilities or an autistic person across the whole range of mental health inpatient services.

As I mentioned above, there seems to be a slight dip during the peak of the Covid-19 pandemic, then there is a year (where there is a gap in the black line) where the cyber attack makes the numbers a little unreliable. The introduction of the 2021 statutory guidance for the Mental Health Units (Use of Force) Act (2018) should not have an impact on the reporting of the number of people in inpatient services. However, there does seem to be a recent trend upwards in the number of people with learning disabilities and autistic people in both independent sector and NHS inpatient units, possibly partly due to the gradual demise of Transforming Care/Building The Right Support policy attention. At the end of March 2024 there were 3,540 people in inpatient units, rising to 3,930 people at the end of March 2026. This increase is in both independent sector inpatient services (from 1,070 people to 1,220 people) and in NHS inpatient services (from 2,425 people to 2,645 people).

What does the MHSDS say about restrictive interventions?

The MHSDS includes information on both the number of people subject to any restrictive intervention in any single month, and the number of restrictive interventions people have been subject to in any single month. It also provides information on a range of specific types of physical restraint, and on mechanical restraint, chemical restraints, seclusion and segregation.

What do the MHSDS statistics say about restrictive interventions overall? The graph below shows the number of people with learning disabilities and autistic people in inpatient units who have been subject to at least one restrictive intervention in that month, from March 2019 to March 2026. The figures for the total number of people subject to restrictive interventions (the black line) is again broken down into independent sector inpatient services (the light blue line) and NHS inpatient services (the dark blue line). 

During the Covid-19 pandemic, the number of people subject to restrictive interventions was reported by independent sector services to have dipped, while there was an increase reported in NHS inpatient services. Was this a 'real' difference, or were independent sector services less likely to be reporting restrictive interventions to the MHSDS than NHS services?

A clue to answering this question is in what happened to reported numbers of people subject to restrictive interventions in relation to the statutory guidance for the Mental Health Units (Use of Force) Act, issued in December 2021. The graph below shows that from September 2021 to September 2022 there were big increases reported by independent sector services but not by NHS services. The cyber attack in 2022-23 makes interpretation over time difficult, but there are clear upward trends over the last few years in the number of people subject to restrictive interventions in both independent sector and NHS inpatient services.

By March 2026, within that one month 855 people with learning disabilities or autistic people in inpatient services were subject to at least one restrictive intervention - 250 people in independent sector inpatient services and 580 people in NHS inpatient services.

Because the number of people reported to be in inpatient services also fluctuates over this time period, we can also look at what percentage of people in inpatient units in each month are subject to at least one restrictive intervention. The graph below shows these percentages for people in independent sector services, people in NHS services, and everyone combined. Even given patchy reporting (the change in the percentage of people reported to be subject to restrictive interventions in independent sector inpatient services increased by over 500% in a year from September 2021 to September 2022, with the introduction of the statutory guidance) and patchy data due to cyber attacks, the overall trend is clearly upwards. What looked like a discrepancy between the independent sector and the NHS early on seems highly likely to be a function of incomplete reporting by independent sector providers, with recent data being similar for both sectors.

I just want to take a moment to reflect on what this graph is saying - in the month of March 2026, 21.8% (more than 1 in 5) of people with learning disabilities and autistic people in an inpatient service were subject to restraint at least once.

As well as the number of people subject to restrictive interventions, we know that people can be subject to multiple restraints, so the graph below shows the total number of restrictive interventions that autistic people and people with learning disabilities in inpatient services were subject to each month, again broken down by whether people are in independent sector or NHS inpatient services.

We've seen from the graphs above that the number of people in inpatient units has started to gradually increase, and the number of people reported as subject to restraints has increased over time (partly due to more comprehensive reporting), so the percentage of people in inpatient units reported as subject to restraints has also increased sharply over time, now standing at over 20% of people. 

The next graph below in some ways shows a similar picture - changes during the Covid-19 pandemic (a reported decrease in the independent sector, a reported increase in the NHS), a big increase reported in the independent sector with the introduction of the statutory guidance, and a general trend of continuing increases more recently. What is different is that despite having smaller numbers of people in their inpatient services and smaller numbers of people reported as subject to restraint, the total number of restrictive interventions reported by independent sector inpatient services is beginning to match those reported by NHS inpatient services.

What does this mean for a person being subject to restraint? The graph shows the average number of restrictive interventions in that month for every person that was subject to at least one. There are different trends over time for people in NHS services compared to people in independent sector services. 

In NHS inpatient services, there was a peak in the Covid-19 pandemic where people who were subject to at least one restraint were restrained on average nearly 13 times in the month - approaching once every 2 days. From this point, the average number of restraints per person for people in NHS inpatient services has reduced to between 7 and 8 times per month (still around once every 4 days).

The trend in independent sector inpatient services is the opposite. There was also an increase during the Covid-19 pandemic, but the average number of restrictive interventions per person subject to at least one has continued to increase ever since. In March 2026, a person subject to at least one restrictive intervention was on average restrained 14.2 times in that month - approaching once every 2 days.

What restrictive interventions are people subject to?

What do the statistics tell us about which types of restraint people with learning disabilities and autistic people in inpatient units have been subject to over the past seven years? The MHSDS records a lot of different types of restraint - for this blogpost I've grouped them into three sets of graphs:

1) Three forms of restraint which have been the subject of specific policy and legal attention: prone physical restraint, seclusion and segregation.

2) Various types of chemical restraint, and also mechanical restraint.

3) Various types of physical restraint (to help interpretation, I've included prone restraint in this as well as in the first set of graphs).

For each of these groups of restraints, there are two graphs; one about the number of people being subject to these types of restraint, and one about the number of instances of these types of restraint being used. With all of these graphs, the comprehensiveness of reporting really matters, so it is difficult to know the extent to which any increases over time are about more people in inpatient services, better reporting, or more people being subject to more restraints.

Prone physical restraint, seclusion and segregation

Prone physical restraint has been the subject of much policy and practice attention, with the intention of prone restraint no longer being used in inpatient services. The graphs below show that, with some fluctuations, the trend is very gradually downwards (the purple lines), although in March 2026, 65 people were still subject to 150 instances of prone physical restraint.

Seclusion and segregation have both been described as forms of solitary confinement, again with policy aims of reducing these to as near to zero as possible. If anything, the graphs report increases over time in both seclusion and segregation. Although interpreting changes over time is difficult, we can say that the current picture is grim. In March 2026, 170 people were subject to 340 instances of seclusion (the grey lines); and 65 people were subject to 95 instances of segregation (the red lines). The huge increases in the number of people reported to be subject to segregation with the introduction of the statutory guidance in 2022-23 shows just how many people were being hidden from official statistics.

Chemical restraints and mechanical restraint

The two graphs below use the same format for specific types of chemical restraint and for mechanical restraint. 

Restraint using rapid tranquiliser injections (the orange lines) are the most commonly used form of chemical restraint. Apart from a sharp jump during the Covid-19 pandemic and an apparent dip in 2022/23, the number of people subject to rapid tranquiliser injections is reported to have increased over time, reaching 120 people experiencing 410 rapid tranquiliser injections in March 2026.

Oral chemical restraint (the yellow lines) is also common. Again, there was a sharp jump in the reported use of oral chemical restraint during the Covid-19 pandemic, and fluctuating levels of oral chemical restraint reported ever since. In March 2026, 75 people were reported to have been subject to 180 instances of oral chemical restraint.

The number of people reported to have been subject to mechanical restraint (the light blue lines) seems to have gradually reduced over time (to 10 people in March 2026), with the number of instances of mechanical restraint generally fluctuating between 50 and 80 instances per month, but dropping to 20 in the month of March 2026.

Physical restraint

The graphs below report the same type of information for various types of physical restraint (including prone restraint). It's worth noticing here that vertical scales on these graphs are different to the previous two sets of graphs, as so many people are subject to so many instances of particular types of physical restraint. With the exception of prone restraint, pretty much all types of physical restraint are reported to have increased in the last seven years.

By March 2026, by far the most common form of physical restraint is standing restraint (the orange lines), exercised on 495 people a reported 2,315 times. Seated physical restraint (the purple lines) was experienced 1,860 times by 260 people in the month of March 2026. Also common are supine physical restraint (the lilac lines), experienced by 205 people on 825 occasions in March 2026, and restrictive escort (the yellow lines), experienced by 155 people on 465 occasions in March 2026. Despite apparent improvements in reporting over time, in March 2026 there were still 175 people experiencing 875 instances of 'other' physical restraint (the grey lines).

Summary

So, what do I think the MHSDS information tells us about restraints being used on people with learning disabilities and autistic people in inpatient services over the last seven years? A few things:

• There has to be a real lack of confidence in the comprehensiveness of the data on restraints being reported by inpatient services - this really hampers any interpretation of reported changes over time or differences between NHS and independent sector organisations. Clearly the belated implementation of the Use of Force Act (2018) has resulted in a big improvement in the comprehensiveness of the data provided, but we still have no way of knowing how complete the information is.
• Whether due to better reporting, changes over time, or a combination of both, the picture for people with learning disabilities and autistic people in inpatient services with regard to restraints is grim, and if anything getting worse rather than better.
• Overall, over 20% of people with learning disabilities and autistic people in inpatient services will be subject to an average of 10 restrictive interventions per person in any given month. That's over 850 people being subject to a restrictive intervention roughly once every three days.
• With more comprehensive reporting, it seems that NHS inpatient services and independent sector inpatient services are similar in the percentage of people being subject to restraint, but independent sector inpatient services on average restrain people more often (14 times in a month) than NHS inpatient services (8 times in a month).
• Standing, seated, supine, restrictive escort and 'other' physical restraints are most commonly reported (at least 150 people are subject to each of these types of physical restraint every month), and possibly increasing over time.
• Policy statements about reducing levels of prone restraint, seclusion and segregation are not working. Although the number of people reported to be subject to prone restraint has gradually reduced over time, around 65 people every month are still subject to prone restraint. The trends for seclusion (170 people) and segregation (65 people) suggest increases rather than decreases over time.
• Injected rapid tranquilisers and oral chemical restraint are the most commonly reported forms of chemical restraint, with the use of injected rapid tranquilisers (120 people in March 2026) possibly increasing over time.

This is such a grim, grim picture. What chance do people have in services with long-standing cultures of restraint, combined with support for people all over falling apart, crisis-driven and inadequately funded mental health services, and policy attention that studiously avoids focusing on what matters?

Children with learning disabilities in schools - 2026 update

This blogpost updates previous blogposts I've written about Department for Education annual statistics concerning children and young people identified within the English education system as children/young people with learning disabilities, recorded in an annual census of schools that takes place in January each year.

In the Special Educational Needs (SEN) statistics there are a number of mutually exclusive categories of SEN, three of which concern children with learning disabilities – Moderate Learning Difficulties (MLD), Severe Learning Difficulties (SLD) and Profound & Multiple Learning Difficulties (PMLD). 

From 2024/25, an additional mutually exclusive category of Down syndrome has been added, related to the Down Syndrome Act (2022). I highly recommend this column by Sharon Smith in Special Needs Jungle which clearly explains the context of the Act, and the potential implications of adding this new category to the SEN statistics.

There are a number of other SEN categories recorded within these statistics (Specific Learning Difficulties; Speech, language and communication needs; Social, emotional and mental health; Autistic spectrum disorder; Visual impairment; Hearing impairment; Multisensory impairment; Physical disability).

Within the annual census, a child can be classified as having a ‘primary need’ in one of these categories, and optionally classified as having an additional, ‘secondary need’ in another category. While the categories of MLD, SLD and PMLD can be both primary and secondary needs, it appears from my reading of the statistics so far that Down syndrome has only been used as a primary need.

Most importantly in terms of how children are supported, children may have a special educational need that has been judged to require specific support in the form of an Education, Health and Care (EHC) plan (previously an SEN Statement). Beyond that, DfE statistics now only report an additional much larger number of children at a level of ‘SEN support’, which has no requirements to specifically support a child.

This blogpost simply goes through what some of these statistics say about the education of children and young people with learning disabilities, as identified within the education system. 

The first question is simply – how many children and young people with learning disabilities are recorded in DfE statistics?

The first graph below shows the number of children with a statement/EHC plan with a ‘primary SEN need’ of MLD, SLD and PMLD (and, from 2025, Down syndrome), from 2010 to 2026 (apologies for the acronyms).

The graph shows that in 2026, over 78,000 children in England had a statement/EHC plan and were identified as children with learning disabilities in one of these four categories. 

Interpreting change over time in very recent years is complicated by the addition of Down syndrome as a category of primary need in 2025, as presumably many of these children would have previously been placed in the MLD, SLD or PMLD primary need categories. It is also highly likely that this new category of primary need will take some time to bed in, so it is possible that the seeming numbers of children with Down syndrome will continue to increase quite sharply in these statistics for a while as children are recategorised.

For children with MLD there were 35,475 children in 2026, with a large decrease of 31% from 2010 to 2018 but an increase of 25% from 2018 to 2026 - the number of children in 2026 is still substantially less than the number of children in 2010. 

In 2026, there were 31,246 children with a statement/EHC plan and identified with a primary need of SLD, an increase in numbers of 24% from 2010 to 2021 but with numbers staying static from 2021 to 2026.

In 2026 there were 9,339 children with a statement/EHC plan and identified with a primary need of PMLD, an increase of 16% from 2010 to 2018 but with fluctuating numbers and a possible small decrease since.

It is early days for statistics concerning children with a primary need of Down syndrome, but in 2026 there there 2,242 children with a statement/EHCP and a primary need of Down syndrome.

 

The second graph below shows the number of children with a ‘primary SEN need’ of MLD, SLD and PMLD (and, from 2025, Down syndrome) at the level of SEN Support, from 2015 to 2026 (the reporting of statistics changed in 2015).

The graph shows that in 2026, 164,294 children in England were identified as children with learning disabilities at the level of SEN Support. For children with MLD this was 161,933 children in 2026, with a 16% increase from 2015 to 2016 and a 34% decrease from 2016 through to 2026.

In 2026, there were 1,592 children with a primary need of SLD at the level of SEN Support, with numbers increasing by 5% from 2015 to 2016 and decreasing by 52% from 2016 to 2026. 

In 2026 there were 503 children at the level of SEN Support with a primary need of PMLD, with an 11% increase from 2015 to 2018 and a 50% decrease from 2018 to 2026.

Finally, in 2026 there were 266 children at the level of SEN Support with a primary need of Down syndrome.

How many children with learning disabilities are being educated in mainstream schools or special schools? The graph below shows the percentage of children with statements/EHCPs in mainstream vs special schools from 2010 to 2026.

The percentage of children with a primary need of MLD and a statement/EHC plan in mainstream school dipped from 51% in 2010 to 43% in 2017, with the percentage increasing again to 63% in 2026, it's highest since statistics in this dataset began in 2010. 

For children with a statement/EHC plan and a primary need of SLD, the percentage of children in mainstream schools decreased from 17% in 2010 to 12% in 2018, and has stayed static from 2018 to 2026. 

For children with a statement/EHC plan and a primary need of PMLD, the percentage of children in mainstream school was around 14% from 2010 to 2016, 15% from 2017 to 2020, and 16% from 2021 to 2025, reducing again to 15% in 2026.

For children with a statement/EHC plan and a primary need of Down syndrome, the percentage of children in mainstream school was 64% in 2026.

For children identified at the level of SEN Support (I haven’t included a graph on this), in 2026 almost all the children with a primary need of MLD (99.7%), the vast majority of children with a primary need of SLD (91.7%), and substantial majorities of children with a primary need of PMLD (76.5%) or Down syndrome (83.5%) were in mainstream schools.

How many children are eligible for free school meals? Although it’s not an ideal marker of the financial circumstances of families, eligibility for free school meals is collected within DfE statistics. The graph below shows the proportion of children with statements/EHCPs associated with MLD, SLD. PMLD and Down syndrome (from 2025) and the proportion of children at the level of SEN Support eligible for free school meals, from 2016 to 2026. There are a number of trends in this graph:

1) There are big increases over time in the percentage of children eligible for free schools across MLD, SLD and PMLD groups in the graph

2) A greater percentage of children with a statement/EHCP are eligible for free school meals compared to children with the same label at the level of SEN Support

3) Children with a label of MLD are most likely to be eligible for free school meals, followed by children with a label of SLD, then children with a label of PMLD, and finally children with a label of Down syndrome.

In 2026, 53% of children with a label of MLD and a statement/EHCP were eligible for free school meals (48% for children with MLD at the level of SEN Support). 49% of children with a label of SLD and a statement/EHCP were eligible for free school meals (43% for children with SLD at the level of SEN Support). 42% of children with a label of PMLD and a statement/EHCP were eligible for free school meals (33% of children with PMLD at the level of SEN Support). Finally, 29% of children with a label of Down syndrome and a statement/EHCP were eligible for free school meals (18% for children with Down syndrome at the level of SEN support.

These figures compare to 27% of all children on the school rolls being eligible for free school meals in 2026 - the graph below shows this for 2026.

At what age are children with learning disabilities identified in schools? The graph below shows how many children for every 1,000 total children in school had a statement/EHCP at each age from age 5 to age 15 in 2026. 

For children with MLD, there were big increases in the number of children with a statement/EHCP through the later ages of primary school and into the early years of secondary school, with slight decreases beyond this point.

For children with SLD, there was a steady increase in the number of children with a statement/EHCP through the school years.

For children with PMLD and children with Down syndrome, the highest numbers of children with a statement/EHCP were in the earlier years of primary school, with slight decreases beyond this point.

I'm not in a position to offer a deep interpretation of these numbers, but I hope it is useful to see them set out like this. As Sharon Smith says in her column, the implications of the introduction of a new Down syndrome category of primary need are as yet unclear, and it is unclear whether using Down syndrome as a primary care need is more meaningful than existing categories in terms of working out what education and support would be most helpful for the child and their family (there are similar questions about the usefulness of the category of Moderate Learning Difficulty).

A very rough handle on the diversity of educational and support needs children with Down syndrome can be gained by looking at children's secondary needs. I'm not totally convinced by the quality of the data when it comes to the categorisation of primary and secondary needs. For example, in 2026 there were 40 children with a primary need of MLD recorded as having a secondary need of PMLD, which in my naivety seems unlikely. But the range of secondary needs recorded for the 2,242 children with a primary need of Down syndrome in 2026 is very broad. In descending order, they were: Speech, language and communication (395 children); severe learning difficulty (142 children); moderate learning difficulty (133 children); specific learning difficulty (108 children); hearing impairment (82 children); autistic spectrum disorder (62 children); physical disability (50 children): other difficulty or disability (44 children); profound and multiple learning difficulty (28 children); vision impairment (23 children); multi-sensory impairment (12 children); and social, emotional and mental health (11 children). 

In summary, I'm not seeing a massive and sustained increase in the number of children with the labels of MLD, SLD and PMLD, certainly taking 2010 as the baseline.  If anything, recent years suggest that the number of children with these labels being supported with an EHCP is flatlining, and the number of children in these categories is relatively small compared to the most common primary needs recorded, including autistic spectrum disorder (165,399 children in 2026); speech, language and communication needs (108,936 children); and social, emotional and mental health (83,339 children).

On the long view from 2010, there may be some signs of changes in the proportion of children with these labels in mainstream schools, with the proportion of children with MLD in mainstream schools increasing and the proportion of children with SLD and PMLD in mainstream schools gradually beginning to decrease. Poverty amongst children with the labels of MLD, SLD and PMLD is continuing to increase year on year, and to much higher levels than the overall figure for all children on the school rolls.

We have had only two years of statistics using the category of Down syndrome, so it is too soon to report any trends over time for this group. Children with Down syndrome, together with children with MLD, are more likely to be in mainstream schools compared to children with SLD or PMLD. Eligibility for free school meals is at a similar level for children with Down syndrome compared to all children on the school rolls, and is lower compared to children with a primary need of MLD, SLD and PMLD. These broad-brush statistics give little indication that legislation specifically focused on people with Down syndrome is required.

What all of these statistics miss is children who are not in school, for any number of reasons. Without having some understanding of this, we only have a partial picture of children with learning disabilities in England.