select search filters
briefings
roundups & rapid reactions
Fiona fox's blog
The Office for National Statistics (ONS) have released their latest report on deaths involving COVID-19 by local area and socioeconomic deprivation.
Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:
“To me, the most interesting patterns in the data, which aren’t substantially different from in the previous release that came out in mid-June, are the huge differences in death rates depending on the type of place where people live, and depending on the measure of deprivation of where people live. On the type of place, death rates involving COVID-19 have been very much higher in urban areas than in rural areas, and it makes a big difference whether the area is classified as being ‘in a sparse setting’. That means it’s not close to other areas with denser population, and places in a sparse setting have much lower death rates involving COVID-19. As an extreme example, the age-standardised death rate from COVID-19 for people living in an urban major conurbation, say somewhere like London or Manchester, was 130 per 100,000 people, while for people living in “Rural hamlets and isolated dwellings in a sparse setting”, say in mid-Wales or the depths of rural Northumberland, it was 24 per 100,000. It’s true, for a lot of reasons, that death rates from all causes are lower in these rural places than in the biggest cities, but the difference is considerably more marked with deaths related to COVID-19. On deprivation, the ONS analysis is separate for England and for Wales, because the two countries use rather different measures of the deprivation of areas (though both measures are based on “factors such as income, employment, health, education, crime, the living environment and access to housing within an area”). In England, the death rate involving COVID-19 in the 10% of areas with the highest levels of deprivation was considerably more than double the rate in the 10% of areas with the lowest levels of deprivation. It’s already the case that death rates from all causes in the most deprived areas are a lot higher than in the least deprived areas, and that’s bad enough, but these data show that the position is even more extreme with deaths related to COVID-19. Broadly the results in Wales about deprivation are comparable to those in England.
“Overall, the patterns in the data covered in this ONS release are mostly very similar to those in the previous release. The reason is that most of the headline figures reported here are for deaths in the whole period from March to the end of June. The previous report covered March to the end of May, and in total covered about 46,700 deaths that involved COVID-19. This new release is based on data on about 50,600 deaths involving COVID-19, so includes only an extra 3,900 deaths. (Most of these extra deaths were in June, though about 550 were in March to May, and weren’t included before because of delays in death registration.) So, in this report as in the previous version, most of the strong patterns in the total across the pandemic so far are based on deaths from April when mortality was at its height, and the relatively small number of additional deaths in this new report haven’t changed things much.
“Two important strengths of the data in these releases are as follows. First, they are based on registrations of deaths, so won’t be affected by changes in how easy it was to be tested for the virus, or whether people were hospitalised. Second, they give death rates, not just counts of deaths, and these death rates are age-standardised. Other things being equal, you’d expect there to be a lot more deaths in London than in, say, Carlisle, from COVID-19 or anything else, simply because far more people live in London than in Carlisle. The rates shown in this release allow for the differences in population – they are rates per 100,000 people living in the place in question. The rates also allow for differences between different places in the proportions of men and of women in different age groups. If that wasn’t done, you might expect the death rate to be higher in a place with a relatively old population, simply because of the age of the inhabitants, and that’s going to be particularly true of deaths involving COVID-19, because it’s been known since the start of the pandemic that this is a disease that has proportionally much worse effects on older people. So when this data set tells us that the March-June age-standardised death rate for COVID-19 in London was 141.8 per 100,000, and for Carlisle it was 103.9 per 100,000, we can be confident that the lower rate in Carlisle isn’t simply because fewer people live there than in London, or because different proportions are older.
“Do we know why death rates from COVID-19 are higher in big cities than in rural areas, and higher in more deprived areas? Not directly from this data release, because the ONS analysis does not take account of other characteristics of people apart from where they live. So the data release can’t tell us whether, for example, the different ethnic mix in big cities compared to rural areas is related to the difference in mortality, even though other data sources have pointed out the worse effects of the new virus on ethnic minorities. You might speculate that, because many areas of high deprivation tend to be in parts of cities and large towns, the low levels of death related to COVID-19 in sparsely populated rural areas might have something to do with lower deprivation there. But these data don’t look at whether that’s true, and in any case there can be high levels of deprivation in certain rural areas, that aren’t always noticed by the majority of us that live in towns and cities.
“The new report does give some details of deaths in June specifically, and there’s much more detail, even for very small geographic areas, in the accompanying data set from ONS. But I’d recommend a considerable amount of caution in discussing those details, particularly for small areas. Because the total number of COVID-19 related deaths in June was fairly small, the numbers in many geographic areas will be very small indeed, and there is therefore very considerable statistical imprecision in the death rates that are reported. The ONS statisticians give measures of this imprecision by giving confidence intervals for the death rates, and for small areas in June, where the numbers of deaths are low, these confidence intervals are very wide. So in many cases, what might look like a fairly large difference in rates between two places might not be statistically significant, that is, it might plausibly be entirely due to chance and not to any real difference between the areas.”
Professor Richard Harris, Professor of Quantitative Social Geography, University of Bristol, said:
“The findings in June differ from the earlier ones in that London was hit first by Covid-19 and originally had the highest regional death rate. Since then, the disease has spread out and regions like the North West have taken the lead.
“Notable trends are that deprived and ethnically diverse areas remain at higher risks, as do places with greater overcrowding and care homes.
“The data do not in themselves say why deprivation raises risk but it isn’t difficult to imagine why – partly greater pre-existing health issues but also the links between deprivation, ethnicity and occupation types (jobs with greater risks of exposure) and household overcrowding.
“London has a young population on average but a part of that is driven by its ethnic diversity with many of those ethnic groups facing greater risk. That, and the fact that it is a world city with a large population, densely populated, that had greater exposure to the virus early on, will drive up the age standardised mortality rates. However, as the public health England report noted, that is a reversal of the usual health gradients (London usually has lower mortality rates).”
All our previous output on this subject can be seen at this weblink: www.sciencemediacentre.org/tag/covid-19
Declared interests
Prof Kevin McConway: “Prof McConway is a member of the SMC Advisory Committee, but his quote above is in his capacity as a professional statistician.”
None others received.