Excess deaths in the time of Covid show how wrong our understanding of this crisis has been
With the UK now at the top of the European deaths table for Covid-19, simple comparisons between headline numbers have little left to tell. Several researchers have turned their attention to excess deaths to provide more insight: that is, the difference between the actual number of deaths taking place and the number usually recorded at this time of year.
Thanks to data that are updated weekly by the Office for National Statistics (ONS), keeping track of excess deaths in England and Wales is now straightforward. Before doing that, using three graphs which draw on these official statistics, we want to step back for a moment and reflect. Data on death in the time of Covid is not normal data – not even normal mortality data. Besides the shock that so many people are dying in this country from something that we’d barely thought about two months ago, the circumstances of a death from Covid-19 – the victim isolated, the family dispersed, the carers at risk of succumbing themselves – mean that the gravity of what is being measured here goes beyond any picture painted with numbers.
Yet the numbers are still worth paying attention to. And one of the things they show is that our image of what this crisis is about and where it is happening has not been right.
Figure 1 compares the actual number of deaths each week in England and Wales with the average number of deaths in that week over the last five years. The first bar shows the weekly average for the first 10 weeks of 2020, up to week ending (w/e) 6th March. The remaining bars show each of the weeks since, up to w/e 24th April, data for which was published on 5th May.
Three numbers are shown for each week:
- The actual number of deaths where Covid-19 has been mentioned on the death certificate (the grey part of the bar).
- The average number of deaths for that week over the past five years (the dotted line) – about 10,500 per week in March and April – which is taken as a measure of what we would normally have expected to see.
- The actual number of deaths where Covid-19 has not been mentioned: “other” deaths (the orange part of the bar). The segment which falls above the dotted line – “excess” other deaths – is a lighter shade of orange than the segment which falls below the line.
The number at the top of each bar shows excess deaths in the week, that is, the amount by which total actual deaths for the week, whether Covid-19 has been mentioned or not, exceed the five-year historic average deaths for the week. For the first 11 weeks of the 2020, to w/e 13th March, excess deaths had been negative (meaning there had been fewer deaths than in the previous years). In the week before the lockdown was announced, excess deaths numbered 70. In w/e 17th April, excess deaths peaked at 11,850. Over the seven weeks to w/e 24th April, excess deaths totalled 38,350.
Figure 1: total deaths and five-year average deaths, per week, England Wales, 2020
The most important point in the graph is that excess deaths are not just deaths where Covid-19 was mentioned on the death certificate. Over the seven weeks to April 24th, 27,340 death certificates mentioned Covid-19. All of these are treated here as excess deaths. There are then a further 11,010 additional deaths that make up the overall excess and which are part of the wider pool of other deaths that did not mention Covid-19. Those 11,010 add 40% to the Covid-19 total.
The same thing can be seen in Scotland, where over the seven weeks to w/e 24th April, 2,177 out of 3,146 excess deaths mentioned Covid-19 on the death certificate.
It is also clear from the graph that these “other” excess deaths only started to appear once people started to die from Covid-19. That some of these “other” deaths are really due to Covid-19 but weren’t classified as such must be beyond doubt. But unless all of them were, the inescapable conclusion is that some people have died not as a result of Covid-19 itself but as a result of the way that health and care services have changed to cope with Covid-19. Deaths in the time of Covid are not restricted to deaths directly attributable to Covid itself.
Figure 2 shows our estimate of how the 38,350 excess deaths in the first graph over the seven weeks to w/e 24th April are divided across the four settings in which a death may take place (hospital, care home, home or hospice and other). Again, all deaths where Covid-19 was mentioned are treated as excess. Each setting’s other excess is calculated as the total number of other deaths in the setting less the setting’s share of the five-year average.
Each setting’s share of the five-year average is an estimate, based on the assumption that the share is the same as the share of actual deaths in the first 10 weeks of 2020. This gives 46% for hospitals, 22% for care homes, 24% at home and 5% for hospices. Since these are also very close to the shares for 2018 as whole, the assumption looks robust. Whatever the particular assumption, the sum of the orange bars in figure 2 always equals the sum of the lightly shaded part of the orange bars in figure 1.
Figure 2: excess deaths by setting, England Wales, seven weeks to w/e 24th April
Again it is worth pausing, this time to reflect on where deaths normally occur. As the coronavirus crisis began to hit in late March, how many knew that in normal times, less than half of all deaths happened in hospitals, with nearly a quarter each in care homes and at home? If we had, wouldn’t we have been wondering about what might happen in care homes and at home from the very outset?
Five points stand out in figure 2. First, there were several thousand fewer non Covid-19 deaths in hospitals over the seven weeks than the five-year average number of deaths in hospital.
Second, deaths in care homes with Covid-19 on the certificate account for less than half of all excess deaths in care homes.
Third, the total number of excess deaths in care homes is only a few hundred less than the total number of excess deaths in hospitals.
Fourth, deaths at home with Covid-19 on the certificate account for less than one in five of all excess deaths at home.
Fifth, once deaths at home are set alongside deaths in care homes, around just two in five of all excess deaths in the time of Covid have actually happened in hospital.
How robust are these findings? It is true that the line between Covid-19 and other deaths is not clear when people have other underlying conditions. The negative orange bar for hospitals could be reduced, eliminated or even reversed if hospital deaths were classified differently. But that does not affect the total height of the hospital bar (grey minus orange here). Nor does it affect the other findings.
What is really driving this is something very simple: as figure 3 shows, at the height of the Covid crisis during April, hospitals’ share of all deaths in England and Wales fell from an average for January and February, and for March, of 46% to just 37% of all deaths in w/e 24th April. This statistic depends on no assumption nor is it sensitive to how deaths are classified.
Figure 3: the share of total deaths by setting, England Wales, 2020
What’s going on in these three graphs – where people have been dying and what they have been dying from – will be studied closely. Our concern right now is not to call for more understanding but to correct two misunderstandings.
The first is that excess deaths in the time of Covid are largely restricted to deaths where Covid is mentioned.
The second is that coping with the Covid crisis is mainly about hospitals coping.
If the first of these has led to the scale and variety of death in the time of Covid being underestimated, the second has led to a narrow view of where the threat lies, who is actually having to confront it and where support needs to go.
A proper view of where deaths happen in normal times might not have been enough to avoid the agony of the care homes but it would at least have meant that we were looking out for it from the first. The belated attention now paid to care homes needs to be extended to home too, to those dying there and those caring for them; to their need for protective personal equipment; to the conditions in which they are coping, often overcrowded and where self-isolation may not be practically possible.
 Office for National Statistics, Deaths registered weekly in England and Wales, provisional, weekly figures 2020 and Covid-19 – place of occurrence
 National Records of Scotland, Deaths involving coronavirus (COVID-19) in Scotland, figure 6 data (accessed 8th May)
 As above
 As above.