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 25 December 2020.
Prof Richard Harris, Professor of Quantitative Social Geography, University of Bristol, said:
“Unfortunately, it looks like the Government took the brakes off too quickly on London and the South East after the November lockdown. A large city, with a large population, high population density and with a functional influence on population movements over an even larger area was always going to risk spreading the disease rapidly when it became established in this region again, which is what has happened. Hopefully the new lockdown will reverse the current upward trends again.”
Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:
“The ONS weekly data on death registration in England and Wales, up to the week ending on Christmas Day (December 25), has been published. ONS always make it clear that these weekly numbers are provisional, and some of the numbers are revised as more information comes in over time. The main reason that the numbers are provisional is that what’s most important from an epidemiological point of view is when someone actually dies, but the figures come from death registrations, and registration doesn’t happen instantly on death. For most of the year, most of the delays in registration are short, a few days, though they can be very much longer (for example, if a death is referred to the coroner). However, registrations are often delayed more around bank holidays, because registration offices won’t be open on some days, and also because the timing of other factors will change, because people aren’t doing what they do at other times of year. This disruption to registrations can be particularly marked around Christmas. The main figures in the ONS bulletin relate to the dates on which deaths are registered, not the date when a death actually occurs. Data by date of occurrence of death are also available, but for the most recent weeks, they can also be affected markedly by late registrations, and again that’s more of a problem at holiday times. The data by occurrence includes deaths that occurred up to 25 December, as long as they were registered on or before 2 January 2021, but it’s clear from the data that a substantial number of deaths that occurred in the week ending 25 December still weren’t registered by 2 January. ONS estimate that somewhere between 12,683 and 16,547 deaths actually occurred in England and Wales in the week ending on 25 December, but the number occurring in that week that were registered up to 2 January is only 9,800.
“Statisticians and epidemiologists generally agree that the best way of measuring the effect of the pandemic on deaths is to look at excess deaths – that is, how much bigger the number of deaths from any cause that occur in a certain period of time is, compared to some measure of the number of deaths that would be expected to occur in that period on the basis of the past few years. However, there are additional problems around Christmas in using the ONS method of calculating excess deaths, that again relate to our pattern of bank holidays. To calculate the expected number of deaths in a given week, ONS take the average number of deaths registered (or occurring) that week in the five years from 2015 to 2019. But these weeks are not defined by the calendar date, but by the week number in the year (they always run from Saturday to Friday). The latest week in the data is week 52, and this year that ended on 25 December and had only one bank holiday in it (Christmas Day). But, as ONS point out, the five-year average for week 52, based on date of registration, is particularly low, because in four of the five years (all except 2015), there were two bank holidays in week 52 . So comparisons using the five-year average and using date of registration are going to be misleading for week 52 (and will again be problematic next week because there were two bank holidays that week this year). The counts by date of occurrence aren’t affected in the same way by holiday dates, but also can’t give a clear picture because the numbers for this year are incomplete for week 52 because of deaths that were not been registered by 2 January.
“There are other ways to set a baseline against which to measure excess deaths, and (for instance) Public Health England (PHE)1 use a statistical model that takes into account many things, including the pattern of registrations around Christmas. But the PHE report on excess deaths for the week ending 25 December is not yet available.
“So, having explained all that, I’m going to say less about the short-term trends in deaths than I sometimes do. Numbers of deaths in the whole of England and Wales for the most recent week were lower than the week before, whether measured by date of registration or date of occurrence, but these falls will have a lot to do with registration delays. My expectation is that, when we do have fuller data on deaths by date of occurrence, the general trend in numbers of deaths will be increasing. Excess deaths, measured by comparison with the five-year average, are far higher in the most recent week than in recent weeks, but that is likely largely to be because of the bank holiday pattern. Numbers of deaths involving Covid-19 (that is, where it is mentioned on the death certificate) are very roughly level, when measured by date of registration, and have been level since late November – and the same is true for the figures by date of occurrence, apart from the most recent week where the number is down (largely or entirely because of registration delays). This is not quite the same pattern as in the data on the dashboard at coronavirus.data.gov.uk, where deaths are counted if someone dies (from any cause) within 28 days of a positive test for the virus. Those numbers have been increasing, though quite slowly, over the period from late November. I don’t know the reason for this difference, which may have something to do with registration delays, or with differences in what is counted as a death related to Covid-19, or indeed something else. However, taking deaths by date of occurrence, and looking at the week ending 18 December (because the occurrence data from registrations for the week ending 25 December is too incomplete), the ONS number from registrations and the number from the dashboard are very similar (3,020 and 3,110 deaths respectively).
“The ONS figures on deaths do not (yet) reflect the large increases in confirmed cases of Covid-19. To some extent that’s because it takes time, typically a few weeks, between someone being infected and their death, if, sadly, that occurs. Confirmed cases have been rising pretty steeply, but only since late November, about the time that the second English lockdown ended (though the rise started a little earlier in some parts of the country). These ONS data only go up to four weeks after that, and the position in the most recent week is not clear because of the registration delay issue. Unfortunately, I’m expecting a rise in deaths registered over the next couple of weeks, following the rise in confirmed cases. I just hope that rise isn’t too great.
“The number of deaths (from all causes) is considerably higher than the five-year average in every region of England and in Wales, and very much so in most regions, though to a great extent that will be because of the timing of the bank holidays. But again I’d expect deaths to be running above the average level in all regions even without that anomaly.
“Again partly because of the bank holiday issue, total deaths are running above the five-year average in all the types of location that are reported by ONS (hospitals, care homes, private homes, and other places). It looks possible that, without that anomaly, deaths in care homes could be a little below average. But deaths in hospital would definitely be considerably above average. The long-term issue that deaths in people’s own homes are roughly a thousand a week above the average pattern is continuing, and it’s still not clear to me why that is happening.”
Dr Jonathan Pearson-Stuttard, Imperial Wellcome Trust Research Fellow, Imperial College London, said:
“Excess deaths are the most important mortality statistic to monitor throughout the pandemic because it captures the total mortality impacts – direct deaths from Covid-19, indirect deaths due to healthcare system and individual behaviour change and those wider economic and social impacts on health.
“Excess deaths from Week 52 are the highest they have been during this second wave however should at this stage be interpreted with caution because there was just one bank holiday in week 52 in 2020 compared to two bank holidays in that corresponding week in previous years. However, a very interesting but potentially worrying statistic is that, unlike in previous weeks, excess deaths are higher than deaths involving Covid-19 (3,566 vs 2,912). This is important as it may suggest an increase in excess deaths in that second group – indirect impacts of the pandemic upon health system pressures that impact care pathways for non-Covid-19 conditions. As hospitalisations have increased substantially over past weeks, and look set to continue, this could have a profoundly negative impact upon common causes of death such as heart disease, stroke, cancer and dementia in the short and longer term.
“Finally, the trend of excess deaths in the community has continued throughout the pandemic which is again of real concern and speaks to the ongoing indirect impacts of the pandemic on mortality.”
https://www.ons.gov.uk/releases/deathsregisteredinenglandandwalesprovisionalweekending25december2020
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 the Advisory Committee, but my quote above is in my capacity as a professional statistician.”
None others received.