select search filters
briefings
roundups & rapid reactions
before the headlines
Fiona fox's blog
The Office for National Statistics (ONS) have released provisional counts of the number of deaths registered in England and Wales in the week ending 9 April 2021.
Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:
“The provisional death registration figures from ONS are particularly awkward to interpret this week. That’s because the latest week included in the data is the week ending 9 April, and Easter Monday fell in that week. So, in terms of distortions of the trends by late registrations because of the bank holiday, more deaths than usual that would have been registered would have been registered the following week (or later) because of the Easter Monday holiday. But also some deaths that occurred before that week, that were registered late because of the Good Friday bank holiday the week before, would have been registered in the week ending 9 April. These two effects work in opposite directions on the number of registrations in the latest week, ending 9 April, but they’re very unlikely to cancel each other out, because the bank holidays involved come in different places in the weeks. Because of these complications, it certainly doesn’t concern me at all that the total number of deaths from all causes registered in England and Wales that week was slightly up on the previous week. And the fact that Easter is a moveable feast and falls on different dates in different years also complicated the comparisons of deaths this year with the average number of deaths in 2015-19, that ONS uses to give a measure of excess deaths. This time it’s because the corresponding weeks in those five years sometimes contained bank holidays, and sometimes didn’t, depending on the date of Easter. So I’m simply not going to comment on those numbers – things will be clearer in next week’s ONS bulletin.
“One way round these calendar difficulties could be to use data classified by the date when the deaths actually occurred, rather than the date when they were registered. ONS do publish data by occurrence date as well as by registration date, but that doesn’t deal with the issue of looking at short-term trends. That’s because deaths are not registered immediately after they occur – there’s usually a delay. The median time between the occurrence and registration of a death in England and Wales in 2019 was four days – that is, half of death were registered within four days, but the other half were not, and some of them took very much longer to register. (In 2019, 11% of deaths were still not registered two weeks after then person died, and half of those were still not registered three months after the death.) Delays are likely to be rather longer around bank holidays. So counts of deaths by date of occurrence around bank holidays will be incomplete for the most recent week for which data are published, and will be revised upwards later as the later registrations come in. Indeed revisions like that happen for all weeks, not just the most recent, though they are usually biggest for the most recent.
“I will comment briefly on a couple of aspects, though. Last week, there was some discussion about the fact that only about three-quarters of the deaths where Covid-19 was mentioned on the death certificate were actual recorded as having Covid-19 as the underlying cause of death. That’s the case for deaths registered in the most recent week too – 74% of deaths where Covid-19 was mentioned on the certificate had Covid-19 recorded as the underlying cause of death. But there are two things to bear in mind about that. First, this isn’t something new. During July and August last year, when the numbers of deaths with Covid-19 on the death certificate were low and generally falling, it was also that case that only about three-quarters of those deaths listed Covid-19 as the underlying cause. Before that, during the first peak, the percentage of Covid-19 deaths where Covid-19 was the underlying cause was much higher, over 90% much of the time. By mid-September last year, when deaths involving Covid-19 were beginning to rise again, the percentage where Covid-19 was recorded as the underlying cause also rose, and remained at roughly 90% till the middle of March 2021. Since then, it has fallen to somewhere around three-quarters each week. So this decrease in the percentage of Covid-19 related deaths where Covid-19 is given as the underlying cause isn’t new – it’s what happens when numbers of deaths involving Covid-19 are low. Second, though, it doesn’t mean that the other quarter of deaths didn’t involve Covid-19 at all. Typically, for a death that mentioned Covid-19 on the certificate but doesn’t record Covid-19 as the underlying cause, the doctor who certifies death would have listed it in the part of the certificate that includes “Other significant conditions contributing to death but not related to the disease or condition causing it.” So, for example, the patient’s death might indeed have been caused by something else, but the mention of Covid-19 could well be because the patient had Covid-19 and that made their death occur earlier than it otherwise would have, perhaps substantially earlier. So, for these deaths involving Covid-19 but not having Covid-19 as the underlying cause, the virus could have had a substantial effect on the patient, made their last days much more uncomfortable, or even shortened their life by a substantial amount. They very much aren’t deaths that have nothing to do with Covid-19, as some have said.
“There have also been various discussions recently on the extent to which vaccination has been responsible for the large falls in deaths involving Covid-19 since January this year. I’ll just make one comparison relevant to that. The peak week for deaths involving Covid-19 in the recent wave was the week ending 22 January. I looked at the numbers of deaths that took place in the ten weeks after that, up to the week ending 26 March, so before the current confusion caused by bank holidays. Also I used data on the date of occurrence of death, not the date of registration. Over those ten weeks, weekly deaths from all causes in England and Wales fell by 51%, and deaths with Covid-19 on the death certificate fell by 94%. Very impressive – but it can be compared with the corresponding data from the first wave, last year. Then, the week with the peak number of deaths involving Covid-19 was the week ending 17 April. Over the following ten weeks, to the week ending 19 June, weekly death occurrences from all causes fell by 55%, slightly bigger than the fall this year. Deaths with Covid-19 on the certificate fell by 92%, just slightly less than this year, but there’s evidence that deaths involving Covid-19 were undercounted to some extent at the peak of the first wave, so this fall is certainly very similar than the fall this year. In Spring 2020, of course, we had no vaccines, but we did have a lockdown that was slowly being lifted. I’m not arguing that vaccines have no effect – it’s clear that they do – and also the times of the year are different and there are other differences between the two periods as well. But I think this comparison makes it very clear that the reductions in deaths this year involve a lot more than the vaccine effect.”
All our previous output on this subject can be seen at this weblink:
www.sciencemediacentre.org/tag/covid-19
Declared interests
Prof Kevin McConway: “I am a Trustee of the SMC and a member of its Advisory Committee. I am also a member of the Public Data Advisory Group, which provides expert advice to the Cabinet Office on aspects of public understanding of data during the pandemic. My quote above is in my capacity as an independent professional statistician.”