1 April 2021
Lies, Damned Lies and Coronavirus
by David Chilvers
We are now one year since the original UK coronavirus lockdown. During this time, 126,670 individuals have died in this country with or from COVID-19, as evidenced on the daily Coronavirus dashboard using the “died within 28 days of a positive test” metric. According to the Worldometer run by John Hopkins University, the UK ranks sixth highest in terms of deaths per 100k from COVID-19, using this basis of measuring mortality. However, the practices involved in recording mortality data differ widely from country to country and, depending upon the measure used, the ranking of countries changes. So, it is important to understand the practices used before making international comparisons.
Most countries record a cause of death on their death certificates. When the initial COVID-19 wave began in early 2020, countries, to a varying degree, enhanced their classification to incorporate deaths from or with COVID-19. Whilst this data, if recorded well, can be useful, there is typically a considerable time delay in producing these statistics.
To provide more timely data, closer to real time, many countries introduced or adapted hospital systems to quickly record the numbers of deaths from or with COVID-19. This data usually only included those dying following a positive test and initially only covered deaths in hospital. Some countries expanded this later on to include deaths in nursing homes or care homes; rarely did the data include deaths at home.
So, whilst this second source provided more immediate data, which was very useful for understanding trends, it was not in all cases as comprehensive as the data collected from death certificates. And this latter metric also suffered from involving the judgement of the doctor completing the certificate, which was often done in the absence of a coronavirus test.
For these reasons, a third concept was introduced, that of excess deaths. We have covered this in the UK in previous articles, but basically the number of excess deaths is the total number of deaths in a period less the expected number of deaths extrapolated from previous mortality data. In many countries, including the UK, the expected number of deaths is a simple average of previous years data. Whilst this is a simple measure, it ignores any long-term trends in the numbers of deaths and/or any special factors relating to a particular year. The latter might include the impact of a heavy or light flu season.
We therefore have the three measures of COVID-19 mortality:
- Recorded at the time of death, usually following a positive coronavirus test and usually only covering hospital and care/nursing homes
- Recorded on the death certificate, involving some judgement by the doctor completing the certificate
- Obtained from comparing total deaths with an expected number
As of 30th March 2021, the UK figures were as follows:
|Metric||Number of deaths as at 19th March|
|At the time of death||126,279|
|On death certificate||150,116|
|Excess deaths 1||125,161|
1 The Excess deaths figure for the UK is estimated by taking the figure for England and Wales and pro rating by the UK population divided by the population for England and Wales.
As can be seen, the measure at the time of death and excess deaths are currently quite similar numbers. However, the first measure will increase as further cases come to light which have not been reported yet; the excess deaths figure will probably reduce, as it has been negative for each of the last two weeks. The metric from death certificates is significantly larger than the other two measures, but the numbers are all within the range 125k-150k.
For a wide range of countries, these three metrics are calculated. The Worldometer, as mentioned above, updates data from the first source on a daily basis. The Economist publishes statistic on deaths as recorded on death certificates and also the estimated number of excess deaths, for a range of countries. The countries covered are mainly in Europe, Australasia and the Americas.
In most of Western Europe, data for the three metrics is quite similar, suggesting that the systems put in place to measure real time deaths are performing very well. This data is for the period up until mid-February and covers deaths per 100k population.
|Country||real time||Death certificates||Excess deaths|
In other parts of the world, the correspondence between the three metrics is not so good and both the real time data and the data on death certificates considerably understates the estimated excess deaths. Here are a few examples:
|Country||real time||Death certificates||Excess deaths|
These countries are typically in Eastern Europe or South America, where it might be assumed that health statistics are less well developed. We would anticipate that the same situation would apply in much of Africa and Asia.
What this analysis suggests is that when looking at “league tables” of deaths from COVID-19, relying on the real time data from the Worldometer is likely to lead to nations in Western Europe appearing to be relatively worse than they actually are.
Taking the three metrics, the UK ranks 5th highest on real time deaths per 100k, 1st on deaths on certificates but only 16th on excess deaths – out of 69 countries. 16th is still in the top quartile but not in the top decile shown by the first two metrics. At the other extreme, Russia is 44th on real time deaths, 37th on deaths on certificates but 2nd on excess deaths, Peru is 15th, 5th and 1st.
The differences are large and raise the question as to how much, if any, of the variations in mortality rates are due to differences in recording processes, rather than real differences in outcomes. The London School of Economics has produced a useful working paper on the methods used to collect COVID-19 mortality statistics in a number of different countries in Western Europe.
As John Hopkins University, which produces the Worldometer data, notes “differences in mortality numbers can be caused by:
- Differences in the number of people tested: With more testing, more people with milder cases are identified. This lowers the case-fatality ratio.
- Demographics: For example, mortality tends to be higher in older populations.
- Characteristics of the healthcare system: For example, mortality may rise as hospitals become overwhelmed and have fewer resources.
- Other factors, many of which remain unknown.”
As we have seen, relying on real time data or even data on death certificates to measure overall mortality from the disease, can lead to significant understatement of eventual excess deaths, particularly in countries where healthcare statistics are less developed – apply caution when making these comparisons.
This article is one of a series – last week’s article Schools Back Part 2 is here.