Ending corridor care for the dying: reverse boarding of end-of-life patients before and after a dedicated end-of-life ward
Abstract
Introduction
“Reverse boarding” — moving inpatients into corridors and other non-bed spaces to create capacity — is an acknowledged harm of hospital crowding, and NHS England has set a standard of zero. For patients dying in hospital the indignity is particularly acute. We examined whether opening a dedicated end-of-life care ward changed the exposure of dying patients to reverse boarding.
Methods
Routinely collected bed-management data (2021–2026) were analysed. End-of-life status was defined using stable markers held before the boarding event; ReSPECT documentation alone was deliberately excluded as a marker as it is present in around one in five of all spells. The share of reverse-boarded spells involving end-of-life patients was compared before and after ward opening (5,835 pre and 642 post spells), with rate ratios and 95% confidence intervals.
Results
Between 2021 and 2025, end-of-life patients comprised 10–15% of several hundred to over a thousand reverse-boarded spells annually. After the ward opened, the end-of-life share of reverse-boarded spells halved, from 17.4% to 8.9% (rate ratio 0.51, 95% CI 0.40–0.66; p<0.001); weekly end-of-life reverse-boarding events fell from 21.4 to 12.4. Total reverse boarding continued, indicating the change was specific to dying patients rather than a general easing of crowding.
Conclusions
A dedicated end-of-life care ward was associated with a halving of corridor care for the dying within weeks of opening. The post-period is short and confounding cannot be excluded; monthly monitoring now reports to our quality committee, with a bed-management flag prioritising dying patients for side rooms.
Comments
Ascertainment maturity and the move from 0.51 to 0.61
Thank you for sharing this and many congratulations on the research and wonderful poster. Several things make the result more believable than it could have been: defining end of life status from markers held before the boarding event, excluding ReSPECT because it is too common to discriminate, and showing that total reverse boarding held steady so the change was specific to dying patients rather than crowding easing generally. One question about your update. The before period is a closed calendar window, yet its end of life share moved from 17.4% to 18.6% between the two analyses. On a fixed denominator that is around 70 spells newly identified as end of life in a period that had already finished, roughly a 7% relative increase in ascertainment, presumably as palliative coding arrived. If so, the most recent spells will always be the least completely ascertained, which would bias the post opening share downwards and flatter the rate ratio. That is the direction your effect has already moved, from 0.51 to 0.61. Would it be possible to repeat the comparison using only spells with a fixed amount of coding maturity behind them, or markers available in real time? If the effect holds under that, it would be much harder to argue with. None of this takes away from the finding. A halving, even an attenuated one, is a lot for one ward opening.
Agreed, and a sensitivity analysis is planned
Thank you for reading the poster so carefully, and for the kind words. You are right, and we had reached the same conclusion when we updated the figures: the attenuation from 0.51 to 0.61 is partly real signal from a longer post period and partly ascertainment maturing, because one of our markers (death within a year of the spell) can only be complete once a year has passed, and the 2026 spells on both sides of the ward opening have had the least time to accrue it. That is why the pre period share crept up between the two runs, and the post period share will drift the same way; we said as much in the video summary, but your framing of it as a bias that flatters the rate ratio is the honest way to put it.
Two things give us some comfort in the meantime. The denominator (all reverse boarded spells) does not depend on ascertainment, so the bias sits only in the numerator, and the pre period has moved by about 7% relative over three months of maturing while the post share would need to rise by around 65% relative to close the gap. We also showed that total reverse boarding did not fall, so a general easing of crowding does not explain it.