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 26 November 2021.
Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:
“The latest weekly bulletin from ONS on provisional numbers of death registrations takes the data up to the week ending 26 November. In terms of death registrations where Covid-19 is mentioned on the certificate, the news looks positive – the number fell, compared to the week before, for the UK as a whole and also for England and for Wales. That’s the second week running when those numbers have fallen. However, some time always passes between a person become infected with the virus and their death, if, sadly, that happens, so the deaths registered in the most recent week relate to infections a few weeks before that, in early or mid November or even earlier in some cases. Deaths following from infections happening around now won’t show up in these figures for several weeks yet.
“I’m not going to predict how long this falling trend might continue, since we’re moving further towards the coldest part of the year, and because of all the uncertainty (so far) about the Omicron variant. Also, the number of death registrations involving Covid-19 did rise in the latest week compared to the week before, very slightly in Scotland (by just one death) and by a greater amount in Northern Ireland.
“However, my main reason for commenting this week is to say something about excess deaths. The number of death registrations from all causes in England and Wales fell in the most recent week compared to the week before, but the latest weekly number is still above the 5-year average for registrations in the corresponding weeks in 2015-19. So, on that measure, we have got excess deaths again in the most recent week, for the 21st week in a row.
“Also, the weekly number of excess death registrations is, again, larger than the number of registered deaths where Covid-19 is mentioned on the death certificate, as it has been throughout the 21 weeks of excess death registrations apart from one week, when registrations were affected by the August bank holiday.
“Several people (including, at times, me) have commented on the fact that not all the excess death registrations are accounted for by death registrations involving Covid-19, and some reports have discussed what might be causing this non-Covid excess. Some have raised the possibility that it could be due to indirect effects of the pandemic, such as people not receiving treatment for serious diseases such as cancers because of reduced access to health services (for whatever reason).
“I can’t say that these possible indirect effects are not happening – but I do want to express considerable caution about measuring non-Covid excess deaths in this way, and urge caution in proposing causes for the numbers. That’s for several reasons, as follows.
“The overall issue is that it’s not as straightforward as it might seem to measure excess deaths. The idea is simple enough – you work out how many deaths would have been expected to occur if the Covid-19 pandemic had not happened, you see how many deaths actually happened, and you get the excess by subtracting the expected number from the actual number. Measuring the pandemic’s death toll in this way avoids problems of how a Covid-related death is defined, and can include deaths that are caused indirectly by the pandemic.
“But we don’t actually know how many deaths would have occurred in a given week or month if the pandemic hadn’t happened, so that has to be estimated somehow. ONS often use the five-year average number of deaths in the corresponding period in 2015-19. That’s simple and straightforward to understand, but (for instance) it doesn’t take into account changes in the number of people in the country. Other things being equal, if there are more people alive, particularly more elderly people, there will be more deaths.
“Other organisations use different methods to find the expected number of deaths if Covid had not happened. The Institute and Faculty of Actuaries in their regular weekly monitoring reports for England and Wales1 uses a method that does take into account changes in the population size, and has other differences too. It has also been showing excess deaths since early July, as with the ONS figures, but for much of that time has not shown more excess deaths than the number of Covid-related deaths, and when it has shown excess non-Covid deaths, the number has generally been considerably smaller than the number calculated from the weekly ONS registration bulletins.
“We can’t say that the actuaries are definitely right and the ONS bulletin based figures are wrong, or vice versa, because obviously we don’t actually know how many deaths would have happened without Covid-19. There are other excess deaths calculations too, for example from the Office for Health Improvement and Disparities2 (OHID). Those figures cover just England, but use yet another different method of calculating that expected number of deaths.
“To get a figure for excess deaths, you need to know how many deaths actually happened, as well as an estimate of how many would have happened without Covid-19. But there can be issues about the actual numbers of deaths too. The weekly ONS bulletins concentrate on the number of death registrations that they record. Those numbers are generally very precise – but they record death registrations, not deaths. This means, for instance, that they can be disrupted in holiday weeks, because the offices that register deaths will be closed some of the time.
“Some statisticians, including particularly Sheila Bird in SMC comments (and also me, occasionally), have rightly drawn attention to the fact that data on deaths classified by when the person actually died, which are often called “occurrence dates” rather than “registration dates”, can show a clearer picture for several purposes. ONS do release data classified by occurrence dates as well as by registration dates, and five-year averages by occurrence rather than by registration are also available. So it’s possible to reproduce the calculations of excess deaths (from all causes), and of non-Covid excess deaths, using data by occurrence rather than registration.
“There is, however a snag in working with occurrence dates, when looking at short-term recent trends. There can be delays3 between when a person dies and when their death is registered. In 2020, about three-quarters of deaths were registered within a week of the date when the person actually died, but that means that a quarter were not. The problem is that some deaths take much longer than that to be registered. For instance, if a death is referred to the coroner, and particularly if there has to be an inquest, the delay can get very long in some cases. Short registration delays of a week or less do mean that the data by occurrence date for the most recent week, and to some extent for the week before that, are incomplete (and will be revised upwards considerably in bulletins published later). But the existence of longer delays mean that some adjustments have to be made to the data by occurrence date.
“I’ve done that, and the results are interesting in terms of what might be causing the non-Covid excess deaths, and indeed how many there are. In the adjusted figures using occurrence dates, there have still been excess deaths from all causes since early July, as with registration dates. (I’m ignoring the most recent two weeks of the occurrence data, because of the impact of short registration delays on those figures.) The total number of excess deaths over that period has been greater than the number of deaths involving Covid-19, though not by quite as much as the results by date of registration.
“But what’s quite a lot different between the registration date and occurrence date calculations is the way numbers of death vary from week to week. In the data by date of occurrence, there are very noticeable spikes in the numbers of deaths from all causes in two specific weeks – the week 17-23 July and the week 4-10 September. For the July week, there were about 1,200 more deaths in that week than in the weeks either side, and that’s quite a difference for that time of year when in most weeks there are under 10,000 deaths (from all causes) in a week. For the September week, the difference is a bit smaller, more like 1,100, but that’s still very substantial. The spikes don’t seem to relate to there being extra Covid-related deaths in those weeks.
“What those two weeks have in common is that there were heatwaves. Temperatures rose to 30 degrees or above in many parts of England and Wales for a few days in both of those weeks. It is well known that heatwaves can have considerable short-term effects on mortality, particularly in older people. Public Health England (PHE) used to publish annual reports4 each November monitoring excess mortality during heatwaves. For 2020, they reported that there were about 2,500 excess deaths during heatwaves, and while that’s higher than many years, it’s comparable to the numbers in 2003 and 2006, long before Covid-19. The PHE estimation method is considerably more detailed and precise than mine, but the existence of these very short-term but quite substantial peaks in mortality, when the weather gets very hot, isn’t something new, and probably has little or nothing to do with Covid-19.
“Overall, it does look as if quite a substantial proportion of the non-Covid excess deaths since July can be explained by the bad effects on mortality of a couple of very hot weeks. I’m not saying that that is the only explanation for non-Covid excess deaths – I’m just pointing out that indirect effects of Covid-19 are certainly not the only explanation, and that what’s actually going on may be much more complicated.”
2 https://www.gov.uk/government/statistics/excess-mortality-in-england-weekly-reports
3 Indeed ONS have published a bulletin today on registration delays in 2020, at https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/articles/impactofregistrationdelaysonmortalitystatisticsinenglandandwales/2020
4 https://www.gov.uk/government/publications/phe-heatwave-mortality-monitoring. I assume that responsibility for this monitoring has moved to OHID or UKHSA, but I have not been able to find a report for the summer 2021 heatwaves.
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.”