The Office for National Statistics (ONS) have released the latest data for deaths registered in England and Wales.
Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:
“This is one of the weeks when the weekly ONS bulletin giving provisional numbers of death registrations needs to be interpreted with particular care. The problem is that the latest week covered, the week 28 August to 3 September, includes the late summer bank holiday. That shouldn’t affect the figures for Scotland, where the August bank holiday is at the start of the month, but it certainly does affect the figures for the other three UK countries. Register offices are mostly closed on the bank holiday, and other people who might be involved in registering deaths might be more likely to be on holiday. Therefore deaths will tend to be registered later, and the number of deaths registered in a week with a bank holiday is lower than in comparable weeks without one. The deaths will eventually all be registered, but the bank holiday blip means that short term trends are very difficult to interpret. ONS do also publish data for England and Wales on numbers of deaths classified by when the person’s death actually occurred, rather than when their death was registered, but these are always incomplete for the most recent few weeks, particularly the most recent week, because of deaths that aren’t registered yet. So again they don’t help much in assessing short-term trends. (They will be amended in later bulletins – that’s one reason the bulletins are labelled as ‘provisional’.)
“However, there are a few points that can be made, and they mostly aren’t very encouraging. I’ll look first at deaths where Covid-19 is mentioned on the death certificate. Despite the issues about registrations in a bank holiday week, the number of registered deaths involving Covid-19 across the UK was slightly higher in the most recent week than the week before, at 781 compared to 769 the previous week. In Scotland, where the bank holiday disruption didn’t occur that week, the number was up by about a fifth. But the numbers of Covid-related deaths were also up by about a fifth in Northern Ireland, and by well over half (from 16 to 25) in Wales, and bank holiday issues do apply there (so if there were no underlying change, you’d expect the registrations to be lower for the latest week). The number of registrations is lower in England for the most recent week compared to the week before, but not much. It fell from 649 to 632, a fall of under 3%. I’d probably have expected it to fall a bit more in a bank holiday week if the underlying level hadn’t changed, but these things do vary from week to week anyway. But overall these figures tend to indicate an increasing trend in total numbers of Covid-related deaths across the UK. Another possible indication of an increasing trend in Covid-related deaths is that registrations of deaths involving Covid-19 in the most recent week made up 7.5% of all death registrations, or about 1 in 13 of all death registrations. That’s up from 1 in 15 the previous week, and 1 in 25 at the end of July.
“Very roughly, that increasing trend does match what we’ve seen in the daily counts of deaths (within 28 days of a positive Covid test) on the dashboard at coronavirus.data.gov.uk, for the corresponding dates, if I look at the dashboard figures classified by when the person passed away. The UK dashboard trend in the weeks corresponding to the latest ONS bulletin was definitely increasing, with the average daily number of deaths on the dashboard increasing from about 110 in the week ending 27 August to about 120 in the week ending 3 September. For England, the dashboard figures weren’t moving much at the time, though with a slight increasing trend. In Scotland and Wales, the trend on the dashboard figures was clearly increasing. It’s harder to assess the trend in the dashboard figures for Northern Ireland – it’s a small country and numbers of Covid-related deaths are averaging about 8 a day, so that random variation from day to day makes the picture less clear.
“It’s probably even harder to say much that’s meaningful about death registrations from all causes in England and Wales for the most recent week. The number is lower than for the previous week, by almost 1,500, but that could well all be because of the bank holiday moving registrations to later weeks. The number of registered deaths from all causes is still above the average for the corresponding week for the five years 2015-2019, before the pandemic, so by that measure there were excess deaths that week, 556 of them. That makes it the ninth consecutive week for which all-cause deaths have been above the five-year average. But the way in which ONS define the corresponding week in those previous years is such that the late August bank holiday didn’t always fall in the corresponding week to this year’s bank holiday week (which is week 35). It did fall in week 35 in four of the five years, but in 2015 it was in week 36. This could mean that the five year average is a bit too high to make a really fair comparison with this year’s figure – but that would mean that the number of excess deaths this year should really be a bit higher than 556, and certainly still a substantial excess. For the latest week, there were fewer excess deaths (556) than deaths involving Covid-19 (659), so if the numbers really showed the underlying trends, the excess deaths would all be explained by Covid-related deaths. But the numbers don’t show the underlying trends properly, and we’ll have to wait till later weeks to get a better idea of the real pattern.”
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 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.”