The Ministry of Justice keeps a file called Deaths in prison custody, 1978 to 2024. It is exactly what it sounds like. Yet the biggest figure in its first table, for males and females combined in the most recent available year, 2024, is 87,256. (Source: Ministry of Justice, Safety in Custody -- Deaths_in_prison_custody_1978_to_2024_accessible -- Table_1_1)

That is not a death toll. It is the population in custody: the living people held in prisons in England and Wales. It sits in a table about dying because it is the denominator. Every rate you have ever seen quoted about safety behind bars, deaths per 1,000 prisoners, is a small number divided by that big one.

Hold the two ideas next to each other and the tension shows up fast. One column counts lives ended. The other counts lives held. And the second column moves far more, far faster, than the first.

Here is why that matters, and I'll flag clearly that this next bit is my reading rather than a published finding. If the number of deaths in a year stays flat while the custody population rises, the death rate per 1,000 falls. Nothing improved. Nobody was saved. The maths simply had more people to spread the loss across. Run it the other way, with a shrinking population and the same number of deaths, and the rate climbs while the reality is unchanged. The denominator is the lever, and it is the part almost nobody quotes.

So take the scale seriously. 87,256 is bigger than most English market towns. Against an England and Wales population of roughly 60 million, my rough arithmetic puts it at about one in every 690 people in those two nations being in a prison cell. That is a mid-sized town's worth of people whose air, medication, cell door and cellmate are all decided by the state.

Workplace safety figures usually cover people who clock off. This population cannot. It is the only group in the country where the state holds total responsibility for whether someone survives the night, which is precisely why the series stretches back 47 years and why the denominator deserves to be printed as loudly as the outcome.

The honest counter-argument to all of this: 87,256 on its own tells you nothing about whether 2024 was a safer year or a worse one. It is one figure from one year in a long series, it is an input rather than a result, and the direction of travel only becomes visible when you place it beside the death counts and check how the population was measured, whether as a yearly average or a snapshot. It also covers England and Wales, not the whole UK, so it is not a national picture. Anyone waving a single year's rate around without saying which way the population moved is doing arithmetic, not analysis.

My view, plainly: the next time anyone tells you custody is getting safer, ask what happened to the 87,256 first. If the answer is that it went up, you have been handed a fraction, not a fact.