UPDATE 29 May 2020
Yesterday (28 May) the Financial Times ran an article with the headline "UK suffers second-highest death rate from Coronavirus" which prompted much discussion on social media. (1)
The article details that in the UK "the virus has directly or indirectly killed 891 people per million" and that it is second only to Spain who have recently readjusted their death figures to a higher number. It also claims that in China, Brazil and Russia mortality rates are lower.
However as Professor David Spiegelhalter from Cambridge University said, “If we can believe the data from other countries, then the UK has done badly in terms of excess deaths."
That is an important point, the accuracy of China's figures continues to be questioned worldwide and Brazil and Russia have not yet reached the peak of their outbreaks so it is premature to compare them with any European country. (2) (3)
Studying the graphs used in the article reveals mathematical discrepancies. The graphs compare each country's excess deaths and how much they have increased this year.
(The total number of deaths above the historical average for any given date is known as "excess deaths". If in the UK for the first week of April the average number of deaths has been 1000 for the last ten years but this year the number of deaths for the same week was 1500, then we would say there have been 500 excess deaths or a 50% increase.)
By analysing the number of excess deaths, the article has claimed that the UK has a very high number of excess deaths, 891 per million people, compared to other countries and these "extra" deaths are from COVID-19. The official death toll from COVID-19 for the UK is 562 per 1 million people so the article goes on to imply that the UK is doing even worse than the official statistics say.
However, when looking at some of the other countries on the graphs a discrepancy is seen.
Every country, with the exception of Israel and South Africa, has had a large percentage increase in their excess death rate. This would be expected in a pandemic as it reflects the fact more people are dying. This number of excess deaths should be at least similar to, or more usually higher than, the number of deaths from COVID-19, it cannot be lower.
That would be like saying that on a given day there had been five more deaths than expected and of those five, nine were from COVID-19. That is mathematically impossible. (4)
So looking at the data in the article we find that, for example, France has had 370 per 1 million people extra deaths, but their official COVID-19 death figure is higher at 439 deaths per 1 million people. For Sweden, excess deaths are also charted as 370 per 1 million people but their official death figure from COVID-19 is higher at 423 per 1 million people.This is impossible. There are similar discrepancies throughout the graphs where the COVID-19 death rate is repeatedly higher than the total number of extra deaths.
The UK has without doubt had a high death toll from COVID-19 for multiple reasons. London is the largest populated city in Europe and one of the most global in terms of its ethnicity. The 2011 London census showed that 40% of London's population are from the BAME community (Black, Asian and Minority Ethnic) who have been disproportionately represented in the death figures as I discussed here. This population demographic has an enormous effect on virus transmission and outcome. (5)
These facts are reasons, not excuses, as to why the UK has seen a high death rate but when inconsistent and mathematically impossible data is used to compare it unfavourably with other nations it is of no benefit to anyone.
Original Piece 14 May 2020
Worldwide, media outlets are awash with COVID-19 headlines about how well some countries are doing compared to others. Comparisons of numbers of cases and numbers of deaths are continually being made that portray some countries, the UK included, in a poor light whilst depicting others as exemplary in their response to the pandemic. However, many of these claims gloss over the data and its complexities and do not stand up to closer scrutiny.
There are multiple reasons why it is not possible to truly compare countries at the moment and it would be wise to step back from such comparisons and consider how accurate any figures, for anywhere, are.
For example some countries are only reporting deaths where the patient tested positive for COVID-19, others are reporting all suspected COVID-19 cases irrespective of test results. Some report hospital deaths only, others deaths across the entire population. Some include deaths where the patient tested positive for COVID-19 even if they died from another cause such as terminal cancer.
The demographics of the country influence the data, a country with a higher population of elderly people, the most vulnerable group in this pandemic, or a denser population would expect to see a higher death toll than one with a younger or smaller population.
Exactly Which Deaths do Countries Report?
This is how each of the six worst affected countries in Europe and the USA are compiling their death figures:
The UK.
As a UK doctor I will look at our data in detail.
On 12 May the UK had had 32,692 deaths in total.
Until April 9th the number of COVID-19 cases reported was only hospital in-patients who had tested positive in hospital. It did not include anyone who tested positive outside of hospital.
From April 10th the case numbers began to include all the positive results for key workers and their families the vast majority of whom were at home and not in hospital.
This is why the number of cases appeared to double in 24 hours from 4344 on 9th April to 8681 on 10th April. As more testing is rolled out case numbers are reflecting this.
Until 28th April only hospital death numbers from proven COVID-19 infection were collated. From 29th April all proven COVID-19 deaths across all settings were included. This meant deaths at home, in care homes, hospices and prisons were included. (1)
To complicate things a little further Public Health England (PHE) reports deaths only where a swab was positive for COVID-19 but The Office for National Statistics (ONS) reports deaths where COVID-19 was suspected, even if a swab was either negative or no swab was taken.
This explains why PHE and ONS do not have the same figures for the numbers of deaths and why those of the ONS are higher. (2)
As an aside, recording the cause of death on a death certificate is more complicated than it first sounds. There are different entry levels for cause of death on the UK certificate, starting with the precise cause of death and tracing back over the conditions leading up to it.
For example, if an elderly person with diabetes and osteoporosis falls and breaks their hip, is taken into hospital where they acquire pneumonia and subsequently die what is the cause of death? The pneumonia, the hip fracture or the osteoporosis that weakened the bones initially? And should the diabetes be taken into account?
The certificate for this example would be completed as:
I (a) Pneumonia
I (b) Hip Fracture
I (c) Osteoporosis
II Diabetes
The diabetes may seem to be unconnected to the death but it can increase vulnerability to infection so is also included on the certificate but as “other significant condition”.
When analysing deaths associated with COVID-19, for some it will be recorded as the direct cause at I(a). For others it will be a contributing cause and entered at II (where the diabetes is in the above) meaning that whilst COVID-19 didn’t directly cause the death it probably hastened it.
The nature of death certificate reporting in the UK means a wide net is cast for COVID-19 related deaths. Combined with reporting deaths in every setting and suspected as well as proven COVID-19 deaths, one would expect our figures to be higher than countries with much tighter parameters on their death certificates. (3) (4)
Italy.
Italy was the first country in Europe to be overwhelmed by COVID-19.
It has now had 30,911 deaths as of 12 May 2020.
Italy only counts deaths where the person tested positive for COVID-19 and only records deaths in hospitals but includes every death where the person tested positive even if there is doubt over whether or not the actual cause of death was COVID-19. It is not known how many people have died outside of hospital or with unrecorded COVID-19. It also tests post-mortem for COVID-19 and these figures are included in its data.
Giorgio Gori, the mayor of Bergamo in the north of Italy said their data was “the tip of the iceberg…too many victims are not included in the reports because they die at home.”
And Eleonora Perobelli, from the Observatory on Long Term Care in Milan said, "the vast majority of officially confirmed Covid-19 related deaths are registered in hospitals".
(3) (4) (5) (11)
Spain.
Spain has recorded 27,104 COVID-19 deaths as of 12 May 2020.
The death toll was initially only based on deaths from proven cases of COVID-19. In mid-April Catalonia and Madrid started releasing figures that included suspected but not proven cases. If the rest of Spain was to record suspected deaths the nationwide death toll could be up to 70% higher.
Dr Fernando Simón, Director of the Spanish Coordinating Centre for Health Alerts and Emergencies said “It is hard to know what the real figure is, even with good statistics.” (3) (6)
France.
France has recorded 26,991 COVID-19 deaths as of 12 May 2020.
Its death toll appeared to rise steeply in April because it began including people who died in nursing and care homes from 1st April. It records both proven and suspected COVID-19 deaths. (3) (4)
Belgium.
As of 12 May 2020, Belgium has had 8848 COVID-19 deaths but adjusting for population size it has the highest death rate in the world at 763 per 1M population.
However, Belgium has always included suspected as well as proven COVID-19 deaths across all settings from the start of its outbreak. Its figures are thought to be extremely accurate with statisticians stating that it has successfully identified nearly all, around 97%, of the COVID-19 deaths in the country. (3) (4)
Germany.
As of 12 May 2020, Germany has had 7788 deaths which equates to 93 per 1M of the population making them appear to be doing much better than any of their European neighbours.
Deaths recorded are those only in people who tested positive for COVID-19, suspected cases are not included.
It is unclear if deaths outside of hospitals are included with some media, both inside and outside of Germany, reporting that care home deaths are included and others that they are not.
The Robert Koch Institute in Berlin that issues the figures breaks them down into the number of cases and deaths in different care settings and by outcome (recovered or died) but, on the 12 May 2020, of a total of 169,606 cases it said it had “no data available” for 54,833.
This may reflect the federal system of government which means the 16 different German states are reporting COVID-19 cases to their local health authority who in turn then transfer the data to the Robert Koch Institute, often with significant time lags. There is continuing controversy about how Germany’s figures are being collated because of the decentralised way deaths are being recorded. (11) (12)
The USA.
Given the size of the USA many argue it should be compared with Europe as a whole rather than individual European countries.
Given the highly devolved system of government across not only individual states but also counties and cities within those states, it is a herculean task to try and ascertain how and what data is being gathered.
The exception is New York City which has borne the brunt of the US’s COVID-19 outbreak with 351,000 cases and 27,000 deaths to date. New York adopted a similar approach to the European countries discussed above both in terms of its lockdown measures and its recording methodology. It is the only place in the USA that has provided weekly data. It includes suspected as well as proven cases of COVID-19 in its figures. (3) (4)
Demographic Differences Between Countries.
As well as differences in how individual countries collect their data in the first place, differing demographics, populations and cultures mean caution should be exercised when comparing them.
Age of Population.
In my piece here I discussed how the average age of COVID-19 cases is lower in some countries and how this may have helped keep the death rate down given the fact we know the virus attacks the elderly more than the young.
Italy has one of the oldest populations in the world and around 75% of its cases have been in people over the age of 50. This contrasts with Germany where just 22% of cases have been in the over 60s. The outbreak in Germany began in younger people who first became infected during skiing holidays and carnivals held in February. Hundreds of cases were traced back to a carnival in Langbroich and a ski centre in the Tyrol, again in young people.
Race.
Black and Ethnic Minority (BAME) people are disproportionately represented in the number of COVID-19 deaths worldwide as I discussed here. It is, as yet, unclear if there is a genetic predisposition in BAME individuals and/or if this disproportionate number of represents inequalities of health care and social provision.
However, it would seem reasonable to suggest that countries with a higher population of BAME people may have concomitant higher numbers of COVID-19 deaths. Not all countries collect statistics on ethnicity but, according to “World Atlas” figures, in France BAME people make up 15% of the population, in the UK 13%, in Italy 8% and in Germany 1%. (13)
Cultural Differences.
In Mediterranean Europe, multi-generational homes and socialising closely together are the norm. The common way to greet each other is with a kiss on each cheek, which is also an ideal way to transmit the virus. In countries such as Germany and Sweden people naturally afford each other a large amount of personal space and a type of social distancing is a natural characteristic. The UK probably sits within these two extremes with its approach.
It follows that cultural norms that afford regular, close contact between individual people and across generations provide the backdrop for more rapid spread of the virus which could be reflected in the COVID-19 data of that country.
Geography.
A country with a small and spread out population will find it easier to manage a pandemic than countries with large cities and dense populations. The denser the population the more easily the virus can spread.
Comparing population densities between countries shows some sharp contrasts. (The population density is the number of people per km2 in that country and is derived by dividing the number of people by the area of landmass they are living in.)
New Zealand has a population density of 18, Sweden 25 and Spain 94.
France has a population density of 119, Italy 206 and Germany 240.
The UK as a whole has a population density of 281 but this is very unevenly distributed between England, Scotland, Wales and Northern Ireland. The population density of England is 430 and of the UK’s COVID-19 cases 90% have been in England (29,673 from a UK total of 33,186.)
The UK has the highest population density of the countries it is continually compared with. However taking England alone its population density is even higher, at four times that of France and twice that of Italy and Germany.
In Conclusion.
Although the media and some world leaders are fixated on comparing countries, pitting them and their death rates against each other in this pandemic, this achieves very little in reality.
It can be seen that unless one knows exactly how many people have died across all settings, who was included and excluded as a “COVID-19 death”, ensure all countries are using the same data collecting measures AND be able to trust that no one country has manipulated its figures before presenting them, we cannot compare “like with like”, an essential corner stone of all scientific research and study. And that is before taking demographic and cultural differences between countries into account.
To quote Professor Spiegelhalter, President of the Royal Statistical Society, “It’s tempting to try to construct a league table, but we’ll have to wait months, if not years, for the true picture.” (15)
(4) https://www.economist.com/graphic-detail/2020/04/16/tracking-covid-19-excess-deaths-across-countries
(12) https://www.theguardian.com/world/2020/mar/22/germany-low-coronavirus-mortality-rate-puzzles-experts
Country comparison oddities
Spain and Portugal share a border and many demographic similarities.
Yet Spain has nearly 600 deaths per million (dpm) while Portugal has just over 100 dpm.
Similarly Austria shares a border with Italy. Yet Italy has over 500 dpm while Austria is just over 70.
As you suggest this is a complex situation which is being affected by as yet unappreciated factors.