Public Sector AI
Why doctors won't write your discharge letter yet
In August 2026, almost 14,000 adults a day in England's major hospitals were stuck in a bed they no longer needed. Everyone thinks they know why: there's nowhere for them to go. For patients who've been in a week or more, that's barely a third of the answer. Another third lies inside the hospital itself, in something any of us might do.
Over the past few months I've been interviewing NHS staff at all levels. One of them, a very senior leader at an acute hospital, told me how the wards there handle paperwork for people who are medically fit to leave (in NHS terms, they no longer meet the "criteria to reside"). Because nobody is completely sure when a patient will actually go, the work that could run early runs at the last minute. "The patient has no criteria to reside on Monday," I was told, but "they'll probably be with us for another week" so doctors generally don't write the discharge summary, the letter that goes to the GP, in advance. "I don't wanna write it twice" is the doctor's reasoning, as it was put to me. The paperwork waits for the decision.
I think about this mechanism every time I hear about social care shortages. Delayed discharge is usually discussed as a national shortage of care home beds and community care funding. Yet the same leader described part of it as "bad process" which ends up "being hidden behind" the very real capacity challenges.
In August 2026, across acute trusts in England with a Type 1 emergency department, an average of 13,833 adult patients a day1 no longer met the criteria to reside but remained in a bed. NHS England estimates the notional cost of these delayed bed days at £241m2 for that single month, though this is not literal cash to be saved. The figures are unvalidated management information2.
We summed NHS England's monthly estimates for September 2025 to August 2026 and they reach about £2.8bn3; that total is our sum, not an NHS England figure. It would be easy to assume the problem is universal. Actually, the problem is concentrated. In July 2026, across the 128 providers submitting acceptable data, most patients (84.7%4) left on their declared Discharge Ready Date.
However, for those who do not leave on time, the wait is significant. The delayed group waited 6.4 days on average4. While a third4 of these delayed patients only went one day late, they account for about 5% of the bed days lost4. More than a third of lost bed days4 sit in delays of three weeks or longer.
The intuitive explanation for these long delays is always capacity: there are simply no beds in rehabilitation units, or there is no staff for domiciliary care. This is true in part, as the others I interviewed each told me. A lack of social care does limit patient flow.
Yet NHS England's own delay-reason categories show a different distribution. Take the patients in August 2026 who had been in hospital a week or more and no longer needed to be there. By our count, NHS England's capacity reasons cover 30.8%1 of them. A similar share, 31.0%1, falls under interface process, where care arrangements are still underway. In practice it mixes paperwork with a genuine lack of beds. Together, those two leave a large remainder: about 38%1.
Most of that remainder is the hospital's own process and its care transfer hub. NHS England's two categories for it, hospital process and care transfer hub process, together make up 32.5%1 of patients stuck in an acute bed for a week or more. On an average weekly snapshot, that is about 4,000 patients1 of about 12,500.
The same leader pointed out that "discharge processes are remarkably nonclinical". Discharge is the fundamental coordination of moving a human being out of an institution. It requires awaiting medicines to take home, writing discharge letters, or waiting for a referral to the care transfer hub for supported discharge. It is also the hub itself waiting for confirmation of immediate care needs and deciding which pathway a patient should follow.
This is where we see the mechanical failure of delay. In many tasks, we wait until one step is finished before starting the next. You do not check out of a hotel until you have packed your bags. But discharge administration could run in parallel with the final days of recovery. The administrative bottleneck has little to do with whether the patient is well enough to go.
Instead, the paperwork acts as a rigidly dependent string of events. The doctor waits to write the summary because the departure date is uncertain. The care transfer hub waits for the referral before assigning a pathway. The summary is left to the last minute, because nobody wants to write it twice if the patient remains on the ward.
The result is that leaving hospital can be harder than getting in. As my interviewee put it, "because they're in hospitals, somehow we have a higher bar to discharge" than to admit. The institutional standard for a safe exit requires a level of bureaucratic certainty that the fluid environment of a ward actively prevents.
In a hospital ward, the fundamental uncertainty is the patient's exact date of departure. When the system cannot confidently predict this date, human staff manage the ambiguity by waiting. On my interviewee's account, staff hold back the paperwork rather than risk wasted effort on a patient whose plans might change. The result is a self-fulfilling cycle where the wait for certainty becomes a cause of the delay itself.
Can this coordination problem be solved? We often assume that repairing a broken institutional process requires some new technology. If we are struggling to manage discharge, we expect that we need a new national policy or a big new IT system to solve it. We look for a technological solution to an institutional problem.
But we do not need to invent entirely new tools to write letters or dispense tablets. The challenge is coordination and visibility. If a ward knew sooner which patients were genuinely likely to go home, the dependent sequence of paperwork could start earlier without the worry of wasted effort.
Resolving this does not mean dismissing the capacity crisis. Social care urgently needs more resources, and we should support efforts to build them. However, we also have to accept that about a third of these long stays come down to the hospital's own processes. These delays belong to the hospital and its care transfer hub.
Trusts can already see this. Acute trusts already record a delay reason for delayed patients who have been in a week or more1, in a weekly snapshot of their discharge situation report. If they can count the patients waiting for medicines, a discharge letter or a referral to the hub1, they can see where their coordination is failing. They can see the steps where the work is waiting for the decision.
Add up the months and it comes to almost 4.9 million delayed bed days3 in a year. That might look like a single national statistic about care homes. But each of those days is also a patient waiting on a sequence of decisions, and some of that sequence could start sooner. Social care will still get the blame. It just won't deserve all of it.
