The Office for National Statistics (ONS) have released a comparison of all cause mortality between European countries and regions between January and June 2020.
Prof Sheila Bird, Formerly Programme Leader, MRC Biostatistics Unit, University of Cambridge, said:
“The Office for National Statistics (ONS) has produced a substantial & hugely informative report on excess mortality, using age and sex standardization to take account of European nations’ and cities’ different age/sex structure and population over that past 5 years versus 2020. Not allowed for are ethnic composition and international differences in co-morbidities, conditional on age and sex.
“Not addressed either: deaths in the UK are counted by the week the death was registered rather than, as in most European countries, the week of its occurrence. Only in Scotland where, by law, fact-of-death must be registered for all deaths within 8 days is the likely impact on shifting of the peak-week minimal (see table). The ONS report gives a wealth of weekly detail on relative age and sex standardized death-rates, as illustrated in the table below. Excess death registrations in week 19 (2 to 8 May) of 2020 seem to have been anomalously low in England but also in France.
“The ONS analysis of cities’ cumulative age and sex standardised excess mortality, reported as percentage of expected deaths to the end of week 24 (6 to 12 June 2020), is particularly poignant. Some striking examples are Madrid (26.3%), Barcelona (17.0%); London (14.9%), Birmingham (15.8%); Edinburgh (9.1%), Glasgow (8.6%), Paris (7.4%); versus Copenhagen (-4.1%).
“Importantly, ONS summarized European nations’ cumulative standardized excess mortality by broad age-group (under 65 years; 65+ years) for all persons; and by gender. Some detail is given here for two of the home nations versus Denmark and Spain. Scotland fares better than England uniformly and better than Spain, those under 65 years excepted where Scotland’s co-morbidities may imperil.
Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:
“This hugely detailed data release from ONS provides comparative information on the course of the pandemic across most of Europe, and it does it in ways that avoid most of the major pitfalls and biases that can arise in such comparisons. But that doesn’t mean that it can answer all the interesting and important questions.
“It’s mostly based on various measures of what are called ‘excess deaths’ from all causes. Using deaths from all causes avoids issues arising from different countries using different definitions of what’s meant by a death caused by, or linked to, COVID-19. It also means that deaths arising only indirectly from the pandemic will be included too – for instance, deaths that occur if someone who was never infected with the virus dies of something else, because their access to health services was restricted due to lockdowns or to their local health services being overwhelmed by COVID-19 cases. Deaths from all causes are relatively easy to count, because every European country has long-standing systems for recording and registering deaths. There can be quite large differences in mortality rates between two countries simply because the population in one of them includes a greater proportion of younger people who are, on average, further from their deaths – but in this report, the death rates have been age-standardised to allow for such differences and put the data onto a more comparable basis. And the term ‘excess deaths’ means that comparisons are made between the rates of death during 2020, including the coronavirus pandemic so far, and ‘usual’ rates of death in previous years, so we can pick out the overall effect of the pandemic on deaths, because there haven’t been any other causes of major excess mortality across the whole continent.
“But that doesn’t mean there aren’t still issues about what it all means. Counting only deaths means that we don’t get a measure of harms to people who don’t die from an infection, but might be very ill for a long time, possibly with lasting consequences to their health. Not looking at recorded causes of death means that, well, these data sets can’t tell us any detail about what actually caused excess deaths. The data can’t tell us directly about the reasons for different excess deaths rates in different countries, cities, and regions; analysis of that will have to include other data. And there are choices to be made in how data on excess deaths are defined. There isn’t a simple ‘usual’ rate of deaths in a country or city, because numbers and rates of deaths can vary quite a lot from year to year anyway, depending on how much seasonal influenza occurs, whether there are any extreme spells of hot or cold weather, and so on. In this report, the ‘usual’ rate is taken to be the average for the previous five years before 2020, but there are other ways to do it, that might give somewhat different results. And there are many different comparisons that can be made – between overall death rates themselves, death rates and excess death rates for different sex and age groups, death rates relative to the average for the past five years for individual weeks, and also cumulatively for the whole year so far, and so on. The ONS data files include all these and more, and they don’t always tell exactly the same story as one another.
“Despite the fact that a death is a death, there do remain differences between countries in the way deaths are recorded, particularly in terms of timing. Some countries record deaths according to the data on which they actually occurred. Some countries, in these data sets, instead record them by the date the death is registered, and registration dates can be delayed by different amounts and can be affected by things like registration offices being closed on public holidays. That’s the case for the UK countries in these data (though the UK agencies do, of course, also record the dates when people die.) Even within the UK, there are differences between the UK nations in death registration practice and timing. Probably these differences do not have a major effect on the main patterns in this report, but they might matter in some specific, more detailed, comparisons.
“And one mustn’t forget that there is a big hole in these data – no information from Germany. Germany has the largest population in Europe (apart from Russia), and has a population density (people per square kilometre or square mile) closer to that of the UK than is the case for other big European countries like France, Spain or Italy. We know from other data that the impact of COVID-19 in Germany has generally been a lot less than in the UK (or France, Spain or Italy). We might learn a lot from comparisons involving Germany and German cities – but we can’t do that from this report because the data aren’t there. Of course, Germany does collect and record data on deaths, just as is done by the countries that are represented in this report – but my understanding is that the delays in publishing the German data are much longer than elsewhere. I believe that the aim in Germany is to publish very accurate data, and not do as most other countries do and publish data more quickly, in provisional form that might be revised later. That can have advantages, but not when you’re trying to make sense of a pandemic that is still occurring.
“For me, perhaps the most interesting finding in this report is not that the scale and size of the pandemic (measured by excess deaths) seems to have been greater in England than in all the other countries except Spain, because that was already known from previous, less detailed, comparisons. It is that, in most of Western Europe, the peak ‘hot spots’ of the pandemic tended to be geographically isolated from one another, with gaps between them where death rates were lower, but that this happened much less in the UK than in other Western European countries (with some exceptions, particularly Belgium perhaps). As I’ve mentioned, nothing in these data tell us why that might have happened, though one can speculate that it has something to do with the fact that countries like France and Spain have regions of low population density (and Belgium does not, not in the same way anyway). Data from Germany, and data for regions within the Netherlands (which were also not available), might have helped make more sense of this comparison.
“There’s so much detail in these data sets that the temptation to look for small, interesting-sounding, differences between cities and local regions is great. But care is needed. In areas with relatively small populations, mortality rates can vary quite a lot for reasons that have much more to do with random chance than with the effect of COVID-19 or anything else specific – in a small region, just a relatively few deaths occurring, by chance, this month rather than last month can affect the age-standardised rates quite a lot.”
All our previous output on this subject can be seen at this weblink: