The Office for National Statistics (ONS) have released provisional counts of the number of deaths registered in England and Wales in the week ending 1 January 2021.
Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:
“The latest provisional death registration figures for England and Wales have been published by ONS. They update the position to the week ending on 1 January 2021. As always, counts based on death registration can allow us to look at changes in the numbers of deaths from all causes, not just those assigned to Covid-19 by a simple rule such as the 28-day rule. As well as including deaths from all causes, the deaths involving Covid-19 in these data all have a mention of Covid-19 on the death certificate completed by the doctor certifying the death, which is a more accurate indicator of what might have led to the death than a rule based on days since a positive test. But the latest data are harder to interpret than usual. I’ll describe why first, and then get on to the actual new data.
“Human calendar customs and holidays conspire to make it difficult or impossible to make some of the comparisons we’d normally make, for the latest week. There are several aspects of this. Deaths generally aren’t registered immediately after they occur, and that is more of an issue over the Christmas and New Year holiday period than at other times of the year. The registration offices are closed on some holiday days, and even without that, people’s patterns of behaviour are different than in the rest of the year and things may simply not be dealt with as quickly. These seasonal patterns of registration delay mean that comparisons based on the date of registration of death, and also comparisons based on the date when death actually occurred, can be misleading. In the longer term, over a few weeks from now, this issue will level out, because the deaths will all be registered in the end (though at any time of year, a small number of deaths might not be registered for many months). But for now, comparisons between the latest week, the week before that (which included Christmas Day), and previous weeks when there were no public holidays, become potentially very misleading. What makes things even more difficult to interpret this year is that the ONS data are categorised by the week of registration or occurrence of the death, and the weeks always end on a Friday. Statisticians and epidemiologists generally consider it more informative to look at excess deaths from all causes, rather than only at deaths allocated to a specific cause like Covid-19, and that involves making a comparison with numbers of deaths in the corresponding week in previous years. ONS make that comparison by looking at average numbers of deaths in the corresponding week in the five years from 2015 to 2019. However, at this time of year, those weeks in previous years differed in the number of bank holidays in the week – this year, the week including Christmas Day had only one, while sometimes it has two, and the week including New Year had two bank holidays, while often it has only one. That messes up the simple calculation of excess deaths. Further, the latest week is classified as week 53 of 2020 (because most of it was in 2020), but most years do not have a week 53, so the ONS comparison with previous years for this week is made with the average week 52 figures for the previous 5 years, the same baseline as last week, and this again contributes to making these weekly comparisons very hard or impossible to interpret.
“An excellent blog* from the Head of Mortality Analysis at ONS, Sarah Caul, throws light on some of the wider issue in counting deaths and coronavirus.
“What can the new figures still tell us? Well, given all the news about rapidly increasing numbers of cases of Covid-19, we’d expect the number of deaths from Covid-19 to be increasing fast. In fact, the number of deaths registered in week 53 in England and Wales was down on the number in week 52, by 1,451 (going by the date deaths were registered). But that’s not a real decline – it happened just because of the issues over registration delays. ONS also provide numbers categorized by the week when the death occurred. Those figures are also difficult to interpret in the short term, at any time of year, because registration delays mean that the number of death occurrences in the last few weeks will be considerably incomplete because some are not registered yet. This shortfall will be greatest for the most recent week. But in fact the number of deaths that occurred in week 53, and were registered by the cut-off date of 9 January, was slightly higher than the number that occurred in week 52. That increase will become much bigger as the data are revised when more of the week 53 deaths are registered. So the decrease in death registrations for the most recent week does not correspond to a real decrease in deaths (from all causes). This isn’t good news at all. The number of deaths in the week ending 1 January was over 2,000 higher than the average number of the past five years – more than a quarter higher. The detail of that comparison is problematic this week, but it’s certainly the case that deaths are occurring in much greater numbers that would be expected in a typical year.
“On deaths involving Covid-19, the news is also not good. In England and Wales in the week ending 1 January, 3,144 deaths were registered for which Covid-19 was on the death certificate. That’s more than the previous week, and these Covid-related deaths account for nearly one third of all the deaths registered that week. People are sometimes concerned that even deaths with Covid-19 mentioned by the certifying doctor might not be caused by Covid-19 (though it would not be mentioned on the certificate if it did not contribute to causing the death in some significant way). Because of this, ONS now publish the number of these deaths for which Covid-19 is recorded as the underlying cause. For the week ending 1 January, that applies to nearly 9 in 10 of them (87.2%, or 2,741 deaths). That’s broadly been the position since September, as it was during the first pandemic wave. Even during the summer when the numbers of deaths involving Covid-19 were much lower, around three-quarters of deaths involving Covid-19 had Covid-19 as their underlying cause. There are simply not huge numbers of people registered as dying of Covid-19 but who really were killed in a car crash or by an unrelated heart attack. These are very predominantly deaths that would simply not have occurred at this time if the person had not been infected with the virus.
“Deaths from all causes registered in the week ending January 1 were lower than in the previous week for Wales and for all regions of England except London, but again this isn’t a true decrease. It is a consequence of the registration issues. What’s notable is that, despite the complication from registration delays, the number is still up in London, which illustrates how serious the pandemic is there at present. In London that week, more than two registered deaths in five (43%) involved Covid-19, and that was the case in Wales too (also 43%). In the other regions where cases were particularly high a couple of weeks before, the East of England and the South-East, nearly a third of registered deaths involved Covid-19, and in all the other regions but one, the proportion was more than a quarter. The exception was the South West, but even there, 17% of registered deaths involved Covid-19. Those numbers are all really high, even in the South West. Too many are being killed by this horrible disease.
“Finally I’ll mention that numbers of deaths in people’s homes continue to be around 100 a day above what would be expected for the time of year. I STILL don’t know why this is occurring.”
Deaths registered weekly in England and Wales, provisional: week ending 1 January 2021
All our previous output on this subject can be seen at this weblink:
Prof Kevin McConway: “I am a Trustee of the SMC and a member of the Advisory Committee, but my quote above is in my capacity as a professional statistician.”