Why the daily headlines for Covid-19 hospital deaths are wrong

If I were still a staff journalist, fighting for space in my paper, or time on my TV programme, my task would be easy. I would show my editor evidence that ministers are dishing out dodgy data daily on the Coronavirus pandemic; that around 1,500 more people have died in hospital from Covid-19 than the government’s latest headline figures suggest; that some deaths take weeks to find their way into the official statistics; that the true number of hospital deaths may well pass the 20,000 mark this week.

All these things are, indeed, true but — like the daily headline figures — not the whole truth. Given the prominence the daily numbers receive each afternoon, they should be presented clearly and understood properly. At the moment, they are not.

We all know now that the daily figures exclude death in care homes and the wider community. This is a terrible omission — terrible both in terms of the grisly truth of how the pandemic has killed some of our most vulnerable citizens, and in terms of the failure to gather vital data accurately and swiftly.

Here, though, my concern is with the figures that seem, on the face of it, more robust: deaths in our hospitals from Covid-19. What follows is my attempt to make sense of the Government’s statistics. Spoiler alert: I have not found, and do not believe there to be, any outright deceit. The quality of Britain’s statistics deserve their high international reputation. Indeed, my analysis is rooted in the detailed data that the Department of Health publishes each day. I have no reason to believe they are anything other than scrupulously accurate.

However, they need to be handled with care. On March 23, the day Boris Johnson announced the lockdown, we were told that the death toll in British hospitals by the day before had reached 335. In fact the true figure was at least double that. When the figures are finally in (one death on March 14 took more than a month to be included in the daily tally), the total up to March 22 is likely to be around 700.

As the numbers began to take off, the published figures lagged behind the truth. Since the start of April NHS England has provided detailed day-by-day data on when each death occurred, as well as when it was reported. The full truth has to be pieced together from separate daily data tables; but the task is fiddly rather than difficult.

Thus, on April 1, the headline figure for England was 2,166 deaths up to the end of March. But the next day, April 2, a further 477 March deaths were added to the total; on April 3, 338; on April 4, 204. As late as last Saturday, April 18, the daily headline figure included 26 March deaths. All in all, the current total of people who died in March from Covid-19 in English hospitals stands at 3,583; and this figure is likely to edge up a little further. The final total for March will be around two-thirds higher than we were told on April 1.

This is serious but not scandalous. If there is blame to be allocated, it is in the presentation of the data by ministers and journalists, not the data themselves. Hospitals generally take a day or two to log the daily numbers in their statistical returns. Perhaps they should be required to do so electronically within hours; but I’m not sure the most urgent demand on people who are doing all they can to minimise deaths is to divert them for even a few minutes so that our statistics are more up-to-date. Better for us to take more care over what the figures tell us when we get them.

Consider what happened on April 7. This was a fairly typical day for this month and, as it was a Tuesday, too early in the week for the numbers to be distorted by the tendency for hospitals to pay less attention to statistical reporting protocols at weekends. The latest count of deaths that day in English hospitals is 730. But only 135 were included in the national figures reported the following day, April 8. The next day, April 9, another 284 deaths were added to the April 7 national total; on April 10, 154. Since then, a further 157 deaths from April 7 have come to light.

In short, it’s not that the daily published figures are wrong, but that they bear little relation to what the headlines indicate: they omit the great majority of deaths that actually happened the day before — but include a large number of deaths that took place days and even weeks earlier.

Indeed, not only do I emphatically not criticise NHS England for the quality of its data (though it could do more to make the numbers easier to analyse); I give it credit for something that, as far as I know, has gone unremarked. Hospitals seem to be speeding up the process of getting the data into the public domain. At the beginning of April, around 70 per cent of deaths were incorporated into the national statistics on the day of the death, or on one of the two days after that. Leaving Thursday and Friday deaths aside — many of them do not find their way into the national data until the following Monday or Tuesday — the proportion has risen to around 85 per cent. Put another way, the number of deaths that are not included in the national returns on days 1, 2 or 3 has halved from 30 per cent to 15 per cent. That’s some achievement. It deserves wider recognition.

The obvious question that arises from this is: how many deaths are currently in the statistical pipeline — those that occurred two or more days ago but are not yet included in the national data? By definition it is impossible to be precise, but my judgement, based on the reporting pattern of recent days, is that the headline figures understate the total number of deaths from Covid-19 in English hospitals by around 1,300, and across Britain as a whole by 1,400-1,500. Yesterday (Sunday) the Department of Health put the total number of deaths up to Saturday at 16,060. The true number is probably around 17,500.

Note that the gap between headline and reality is, in absolute numbers, slightly less than at the beginning of April and, proportionately, much less. The headlines are less misleading than they were.

But they are still not right. In the coming weeks, big decisions must be taken about when, and in what stages, to relax the current lockdown. To make that call, ministers will need the best possible information. This will include hard numbers — recorded deaths — but also soft numbers, which are the best estimates of what the recorded numbers don’t yet include. Perhaps these estimates are circulating within Whitehall. If not, they should be. And if they are, they should be shared with the rest of us.