- The number of deaths from coronavirus (COVID-19) is often used to compare countries but is an unreliable metric for making meaningful comparisons. This is because the way in which COVID-19 deaths are counted varies across countries and may change over time.
- A better measure is excess deaths (or excess mortality) – the number of deaths in a given period over and above the number expected (such as the number in an average week). Comparing excess deaths to usual deaths across countries, regions and localities can provide valuable learning about differences in impact, and what has contributed to them.
- In England and Wales for the week ending 17 April – the peak week for death registrations – there were 113% more deaths than usual. This is second only to the figure seen during Spain’s peak week and exceeds that in Italy and France.
- London was the most badly affected region in the UK. But Madrid and Lombardy had higher peaks and were hit earlier and harder. The scale of the tragedy in New York City, however, dwarfs that in these other regions – at the peak, excess deaths were more than 630% higher than usual, compared to 240% in London.
- The variation between countries and regions in the UK is less than the variation seen in other countries. This could be because of the UK’s population density and inter-connectedness, or it could be because the UK was relatively slow to lockdown.
- The chance of a Londoner dying in the week ending 17 April 2020 – the peak week in the capital – was almost 240% higher than in previous years. The West Midlands had the next highest peak, while the south-west had the lowest, with a similar profile to Wales.
- At a local authority level, those with the highest relative risk of dying are in London, with Brent (141%), Harrow (137%) and Newham (126%) being the worst affected.
- While the reasons driving the different levels of impact from COVID-19 are complex and still being explored, this analysis concludes by highlighting five major factors that contribute towards an area’s excess death rate.
Coronavirus (COVID-19) is directly and indirectly responsible for the deaths of more than 60,000 people in the UK and many more thousands across the world. There is also still much to learn about this new disease, including how best to fight it. Comparing the impact of the disease in different countries and regions is an important first step in identifying variations in impact and what might have contributed to these. Understanding how geographical areas across the UK have been affected also sheds light on where and when the virus has spread and its impact in different regions.
The number of deaths from COVID-19 is an obvious international comparator but ultimately an unreliable metric for direct comparisons. This is because different countries count COVID-19 deaths in different ways. A better measure is excess deaths – the number of deaths in a given period over and above the number expected (such as the number in an average week). Expressing excess deaths as a proportion of usual deaths is a useful metric for comparing geographical areas (see Box 1).
Here we take a closer look at what the excess deaths data can tell us about COVID-19’s impact. We look at comparisons between countries, cities and regions, as well as certain areas where the impact has been felt most sharply. We also explore how far the virus has spread across particular regions and areas of the UK.
Excess deaths as a proportion of expected or usual deaths: Measures the relative risk of dying in a period. For example, if the average number of deaths in a particular week of the year is 10 (say for a population of 100,000), and in the same week in 2020 it has risen to 20, the risk of dying in that week has increased by 100%. If deaths are concentrated in very different population groups to those in historic periods (eg working age rather than older population), the usefulness of this metric to make comparisons between groups is more limited.
Excess deaths per capita: Measures the increase in the absolute risk of dying across the population. This helps to make comparisons between populations that are of different sizes. However, this can be misleading because different populations have different age structures (see the example below) and in early stages of a pandemic it can mask the extent of a local outbreak.
Excess deaths per capita, age- and sex-standardised: This adjusts for different age structures and is arguably the best measure as a comparator of risk between different groups. Age-standardised mortality rates are used to allow comparisons between populations that may contain varying proportions of people of different ages. This metric can be less useful in understanding the absolute scale of the impact.
Example: The table shows two regions each with populations of 100,000: region 1 with England’s age structure, and region 2 with London’s younger age structure. The usual deaths and excess deaths (over a period of the pandemic) are shown for each region, assuming that the rates of usual death by age and excess death by age are the same for each region. The increased risk of death for a person of a given age dying is therefore the same in each region (eg for those aged 65–74 just over 50% (12/23 or 8/15).
But looking at excess deaths per capita suggests that the younger region 2 is faring much better with 62 per 100,000 population, compared to 91 per 100,000 in region 1 – though the underlying increase in risk for a person of a given age is the same in each region. Because we have assumed excess death rates by age are the same between the two regions, age-standardising would bring each region’s rates in line. Excess deaths as a proportion of usual deaths is 66% (91/137) in region 1, and 64% (62/97) in region 2. These values are very similar, showing that this measure is relatively insensitive to the different age structures and good for making comparisons.
|Region 1||Region 2|
|Age||Population||Usual deaths||Excess deaths||Population||Usual deaths||Excess deaths|
|Under 1 year||1,093||1||0||1,315||1||0|
International comparisons of excess deaths
The chart below compares weekly excess deaths for selected countries. It shows that in England and Wales for the week ending 17 April – the peak week for death registrations – there were 113% more deaths than usual. This is second only to Spain which had a peak of 153%, about 2 weeks earlier. The peak in France was around half that of England and Wales, while Germany’s peak was much lower at 13%.
Note: On 1 July 2020 we published an updated version of this chart. View the updated chart
Comparing excess deaths of different outbreak hotspots
Looking at the national figures alone masks the huge variation within countries. China’s outbreak was concentrated in Wuhan, while Lombardy (which includes the city of Milan) was the centre of the outbreak in Italy. In the USA, New York City has been hit far worse than anywhere else. In the UK, London bore the brunt, at least initially. All of these are globally well-connected major population centres, so the virus is likely to have reached them early and to have spread more widely before lockdown. The chart below compares these hotspots in terms of the increase in deaths beyond the expected level.
This shows that, despite London being the most badly affected region in the UK, Madrid and Lombardy had higher peaks and were hit earlier and harder. The scale of the tragedy in New York City, however, dwarfs that in these other regions – at the peak, excess deaths were more than 630% higher than usual, compared to 240% in London.
How far did the virus spread across the UK compared to other countries
The chart below looks at the variation between countries and regions of the UK, and regions of Italy, Spain and France. This shows that all regions and countries in the UK had excess death rates exceeding 30%, whereas this is the case in only 7 out of 20 regions in Italy, 9 out of 15 regions in Spain and 2 out of 13 regions in France. There is also much more clustering of regions and countries within the UK than elsewhere. The reason for these patterns could be because each region or country in the UK has areas that are densely populated and/or well connected. This may not be the case in the least affected regions of other countries. Another possibility is that every country/region in the UK was affected because the virus had spread across the nations before lockdown.
Which parts of the UK have seen the most excess deaths?
We can also look at the number of excess deaths within the different regions and countries of the UK. The chart below shows excess deaths as a proportion of the 5-year average by region.
This shows that the chance of a Londoner dying in the week ending 17 April 2020 – the peak week in the capital – was almost 240% higher than in previous years. The West Midlands had the next highest peak, while Northern Ireland had the lowest, with a similar profile to the south-west and Wales. The peak weeks in Yorkshire and Humber, the East Midlands, the south-east and the south-west were all a week later than elsewhere. In Scotland the peak appears to have occurred earlier than elsewhere, in the week ending 10 April, but there has been a slow decline since then. In the week ending 22 May, the northeast was the worst affected, with 41% more deaths than usual (compared to 30% or less everywhere else). The north-east has experienced a slower decline from the peak than London, the West Midlands and the north-west.
The table below shows excess deaths as a proportion of usual deaths by region over the 10-week period of the pandemic (week ending 20 March to week ending 22 May), ranked from highest to lowest. For added context the table also shows the total population, median age, percentage of population aged 65 and older, and population density of each region.
|Excess deaths as a share of usual deaths||Population, millions||Median age||Population aged 65 years and older||People per sq km|
|Yorkshire and The Humber||51%||5.5||40.1||19%||357|
Note: The pandemic period for England, Wales and Northern Ireland covers weeks ending 20 March to 22 May, while for Scotland it covers weeks beginning 16 March to 18 May (weeks ending 22 March to 24 May).
Over this period the chance of dying in England was 57% higher than in the same period in previous years; in Scotland it was 43%, 34% in Wales and 28% in Northern Ireland. There is clear variation between the English regions with the chance of dying across the 10-week period being 104% greater in London compared to 32% greater in the south-west.
The localities most and least impacted by COVID-19
The risk also varies substantially within the regions and countries of Great Britain.* The chart below shows the excess deaths over the whole pandemic period as a proportion of usual deaths for local authorities (health boards in Wales and council areas in Scotland). Most of the local authorities with the highest relative risk of dying are in London, with Brent (141%), Harrow (137%) and Newham (126%) being the worst affected. The least affected local authorities in London faced roughly the same risk as the worst affected health boards in Wales.
Outside of London, the relative risk of dying was also high in major cities with large population sizes, such as Birmingham (81%), Liverpool (74%), Manchester (65%), Sheffield (57%) and Leeds (48%). The least affected localities in England were north-east Lincolnshire (1%), Hastings (6%), and Torridge (6%).
In total, almost half (45%) of all local authorities had excess deaths exceeding 50% of usual deaths. Local authorities in the top quarter of those most affected had, on average, excess deaths exceeding 80% of usual deaths – those in the bottom quarter had average excess deaths of 21%. Differences between areas may partly be due to differences in vulnerability (eg all else being equal, younger populations would have lower ratios of excess to usual deaths; more deprived areas would have higher ratios). But these differences may also be due to variation in the proportions of people exposed to the virus.
A potential second wave of the pandemic may particularly impact areas less affected by the first wave if their populations have been exposed and developed some immunity. Total excess deaths in the UK over the 10 weeks of the pandemic to 22 May were around 62,000. If English local authorities in the bottom three-quarters of the distribution were to have the same death rate as the local authority at the top quartile, there would have been around 19,000 more deaths.
What might be driving higher rates of excess deaths in some regions?
This analysis has illustrated that there are major differences in excess deaths between countries, between hotspots within these countries, and between other regions and localities within countries. While the reasons driving the different levels of impact from COVID-19 are complex and still being explored, there are five major factors that contribute towards an area’s excess death rate:
- The proportion of the population infected before lockdown (which in the UK was on 23 March): We would expect this to be influenced by population density. This is borne out by analysis undertaken by ONS of the variation in excess death rates adjusted for age by local area. ONS found that the highest age mortality rate involving COVID-19 was in urban major conurbations, with 64.3 deaths per 100,000 population. The next two highest rates were urban minor conurbations with 34.3 deaths per 100,000 population, and urban cities and towns with 28.2 deaths per 100,000 population. The lowest rates were all found in sparse settings, rural hamlets and isolated dwellings, which all had the lowest age-adjusted mortality rate of 9.0 deaths per 100,000 population.
- The spread of the virus through the population after lockdown: This will reflect the extent to which populations are able to practice social-distancing and the share of the population at increased risk as a result of their occupation or circumstances (eg people living in larger households or in care homes).
- The composition and vulnerability of the population: COVID-19 does not affect everyone equally. Excess deaths as a proportion of usual deaths rise with age so, other things being equal, the older an area’s population the greater the excess deaths ratio. In addition, the ONS has found that socioeconomically deprived populations are more affected. The rate of deaths, adjusted for age, involving COVID-19 in the most deprived areas of England was 55.1 deaths per 100,000 population compared with 25.3 deaths per 100,000 population in the least deprived areas. One reason for this is that deprived populations are more likely to have underlying health conditions and are therefore at greater risk of being severely affected by COVID-19.
- Access to and the quality of health care received by patients with COVID-19: In Italy, the health care system in Lombardy was overwhelmed and may have contributed to the spread of infection across hospitals and the deaths of patients. In England, new capacity was rapidly built in Nightingale hospitals before it was needed and there is no evidence that patients in different areas of the country received significantly different quality of care.
- Health care for non-COVID-19 patients: Although, in England, some health care use by non-COVID-19 patients has fallen sharply since lockdown – and may have contributed towards excess deaths – there is not yet evidence that this varied significantly across different areas of the country. For example, attendances at major A&E departments fell by similar amounts across the country.
*Further analysis is needed to look at excess deaths in Northern Ireland at a local authority level.
Health Foundation @HealthFdn
RT @Gummifot: I’m excited to share the results from a study I’ve been working on as part of the #ImprovementAnalytics Unit at @HealthFdn. I…Follow us on Twitter
Work with us
We look for talented and passionate individuals as everyone at the Health Foundation has an important role to play.View current vacancies
The Q community
Q is an initiative connecting people with improvement expertise across the UK.Find out more