Belmarsh
Reception prisons · HMPPS functional group: Reception · Public sector · Long-term and high-security estate
HMP Belmarsh is a public sector prison in the Reception group. It sits in the lower middle of 31 similar prisons, with a likely position of 12th to 24th. Its official HMPPS rating is "Good". Based on official data to March 2026.
At a glance
| Band among 31 reception prisons | Lower middle |
|---|---|
| Likely position |
12th to 24th of 31
Median 20th. The range is the 5th to 95th percentile over 1,000 re-weightings and measurement uncertainty. |
| Pillars | Safety Lower middle Decency Upper middle Regime Lower middle Resettlement Bottom quarter |
| Official HMPPS rating (context) | Good, score 63.5%
Not an input to our bands. |
The measures behind the bands
Position is where the prison sits among its 31 peers on that measure alone (lower incident rates are better; higher grades and outcomes are better). For incident outcomes the position is after adjusting for who the prison holds.
| Measure | This prison | Group median | Position in group | Data from |
|---|---|---|---|---|
| Safety pillar position: Lower middle | ||||
| Prisoner-on-prisoner assaults | 330 per 1,000 prisoners expected 175 | 357 | Bottom quarter | Apr 2025 to Mar 2026 |
| Assaults on staff | 178 per 1,000 prisoners expected 121 | 140 | Bottom quarter | Apr 2025 to Mar 2026 |
| Self-harm incidents | 604 per 1,000 prisoners expected 613 | 686 | Lower middle | Apr 2025 to Mar 2026 |
| HMIP healthy prison test: safety | 3 (reasonably good) | 2.0 | Top quarter | Inspected Jun 2024 |
| Prisoner survey: safety and security | 4 (highest) | 3.0 | Top quarter | Survey from Nov 2024 |
| Decency pillar position: Upper middle | ||||
| HMIP healthy prison test: respect | 3 (reasonably good) | 3.0 | Upper middle | Inspected Jun 2024 |
| Prisoner survey: respect | 3 | 3.0 | Lower middle | Survey from Nov 2024 |
| Regime pillar position: Lower middle | ||||
| HMIP healthy prison test: purposeful activity | 1 (poor) | 1.0 | Lower middle | Inspected Jun 2024 |
| Prisoner survey: purposeful activity | 3 | 3.0 | Lower middle | Survey from Nov 2024 |
| Prisoners in at least half-time purposeful activity | 40% | 46% | Lower middle | 2025/26 |
| Resettlement pillar position: Bottom quarter | ||||
| Housed on the first night after release | 65% | 74% | Bottom quarter | 2025/26 |
| In employment six weeks after release | 9% | 16% | Bottom quarter | 2025/26 |
| In employment six months after release | 22% | 26% | Lower middle | 2025/26 |
| HMIP healthy prison test: preparation for release | 3 (reasonably good) | 3.0 | Upper middle | Inspected Jun 2024 |
Incidents compared with what we would expect
Rates are incidents per 1,000 prisoners over April 2025 to March 2026. The expectation is the average of a statistical model of who the prison holds (sentence length, remand, recalls, age, turnover, crowding) and the average of its HMPPS comparator prisons. Neither includes the prison's own result.
| Outcome | Actual | Model expects | Similar prisons | Expected (used) | Actual ÷ expected | Reading |
|---|---|---|---|---|---|---|
| Prisoner-on-prisoner assaults | 330 | 259 | 117 | 175 | 1.89× | Higher than expected on one benchmark only: not flagged |
| Assaults on staff | 178 | 120 | 123 | 121 | 1.46× | Higher than expected, but within what chance and recording differences could produce |
| Self-harm incidents | 604 | 596 | 630 | 613 | 0.99× | About as expected |
A flag appears only when the result is outside statistical limits and in the same direction under the model, the comparators and their average. It is a prompt to look, not a finding of fault. Counts of one or two incidents are suppressed in the source and estimated here.
Similar prisons
HMPPS names these prisons as the closest comparators, matching on age, sentence length, turnover, sex-offender share, complexity, building age and setting. Rates are per 1,000 prisoners; reoffending is the share of people released in April 2023 to March 2024 who reoffended within a year.
| Prison | Prisoner assaults | Staff assaults | Self-harm | Reoffending | Official rating |
|---|---|---|---|---|---|
| Belmarsh (this prison) | 330 | 178 | 604 | 44% | Good |
| Long Lartin | 177 | 215 | 798 | – | Good |
| Frankland | 130 | 59.0 | 682 | – | Good |
| Whitemoor | 60.7 | 152 | 382 | – | Good |
| Full Sutton | 86.1 | 52.2 | 657 | – | Good |
| Wakefield | 132 | 137 | 632 | – | Good |
| Average of similar prisons | 117 | 123 | 630 | – |
Who the prison holds
These describe the population, not the prison's performance. They are what the expectations above adjust for. Averages over the four quarter-ends of 2025/26.
| Share of population | This prison | Group median |
|---|---|---|
| Average population | 700 | 904 |
| On remand (untried or convicted, unsentenced) | 69% | 59% |
| Sentence under 12 months | 1% | 8% |
| Sentence 12 months to under 4 years | 4% | 9% |
| Sentence 4 years or more (determinate) | 9% | 6% |
| Life or IPP sentence | 7% | 0% |
| Recalled to custody | 9% | 15% |
| Aged under 25 | 21% | 16% |
| Aged 50 or over | 13% | 13% |
| Releases in the year, per prisoner held | 0.33 | 0.83 |
| Prisoners held in crowded accommodation | 0% | 56% |
Remand is decided by courts and recall by probation and the Parole Board. They are context here, not something a prison controls.
Also shown, not scored
Staff sickness and resignations
Sickness: 12.6 days per full-time employee (HMPPS target 9; median for public prisons in its group 13.3).
Resignation rate: 2.7%.
Position among public-sector prisons in its group: sickness lower middle, resignations lowest quarter. Related to quality in our tests, but the direction of cause is unclear, so it is not scored.
Reoffending after release
44.0% of 277 people released in April 2023 to March 2024 reoffended within a year. Expected given who is released and the mix of the prison and its comparators: 44.6%. Result: about as expected.
The Ministry of Justice says rates by releasing prison should not be used to assess prisons: they do not control for sentence or offender characteristics, and the prison of release may not be where the sentence was mainly served. About two years old.
Disorder incidents and finds
| Per 1,000 prisoners, 2025/26 | This prison | Group median |
|---|---|---|
| Incidents at height | 50.0 | 84.0 |
| Barricades | 55.8 | 15.5 |
| Drug finds | 330 | 338 |
| Weapon finds | 242 | 144 |
| Mobile phone finds | 38.6 | 92.7 |
Not scored. We cannot separate real disorder from differences in how incidents are recorded, and more finds can mean better searching rather than more contraband. Also recorded: 1 release in error; 1 active concerted indiscipline incident; 3 passive concerted indiscipline incidents.
Healthcare (Care Quality Commission)
Provider: Integris Healthcare Partners Limited; Practice Plus Group Health and Rehabilitation Services Limited.
Latest compliance judgement: no compliance judgement published.
The CQC does not give prison healthcare a graded rating, so this is context only.
How current is this?
| HMIP grades | Inspected Jun 2024 |
|---|---|
| Prisoner survey grades | Survey from Nov 2024 |
| Incident rates and workforce | April 2025 to March 2026 |
| Resettlement outcomes | 2025/26 performance year |
| Reoffending | People released April 2023 to March 2024, followed for a year plus a six-month wait for court outcomes |
| Population mix | Quarter-ends June 2025 to March 2026 |
Sources: data downloads and methodology. Think a figure is wrong? Tell us.