Coronavirus deaths

One of the most interesting things about this whole coronavirus epidemic, to me, is how the media (and the clever people) like to use stats. One of the most bandied stat is the number of tests done. From this, the theory is to trace recent contacts of anyone found to have the virus and then to test these contacts. Of course, what is supposed to follow from that is that everyone who has the virus is supposed to "self-isolate" for at least two weeks in order to prevent further spread of the virus.

The clever people now make all kinds of guesses from these numbers as to what is happening, and then what is going to happen. Unfortunately, the only thing they have right is that the virus is going to spread, and there is nothing we can do about it. There is much hype about a vaccine that is going to stop the general population from getting the virus, but as yet, there is no vaccine for the common cold nor the annual flu, so what are the chances of a vaccine for the coronavirus being discovered? As things stand at the moment, there are only two ways out after contracting the coronavirus: you recover because your immune system can overcome the disease; or you die. There is no medical cure.

There is much hype around slowing the virus to prevent overloading of the health care system (the buzzword is flattening the curve), but I have not heard a single word from any reporter, government spokesperson, nor clever person, as to why the health systems cannot handle serious cases. The answer to this is simple, money that should have gone to health care was spent elsewhere.

The unfortunate issue with this virus is that many people ending up in a hospital are more than likely terminal, so will probably die.

Because of what I have said above, I believe the most important stat that should be monitored is how many people are dying in a population. The total number of cases in a population is meaningless as it stands at the moment as it is impossible to test a whole population - there is no money nor facilities to do this job. South Korea had the facilties in place before the virus started, so was able to stop carriers very early in the epidemic. China seems to have gotten it right by brute force, but the rest of the world, by and large, got left behind.

The folowing is a series of charts of the daily deaths per million population in a few countries I have followed. These charts should be updated regularly (unless I too succumb to the virus). Note that the UK data has been manipulated to minimise the effect of the spike introduced on 30 April (see below).

Fig 1. Daily coronavirus deaths per million population of some countries in the northern hemisphere (see Note 7 below)

The graph above looks like a dog's breakfast, but if you look at the curves, you will note that there was a drop in deaths over Easter weekend in most of the Christian countries I have monitored. This was followed by a spike in deaths in the days following Easter. This seems to happen to the UK and Sweden every weekend, but more pronounced with Sweden!

In order to smooth the curves somewhat I have taken a five day moving average on the data. The average value is placed on the middle day of the set being averaged (thanks to Prof Niko Sauer for the idea).

Fig 2. Five day moving average of the daily coronavirus deaths per million population of some countries in the northern hemisphere

As you can see, the data now starts showing interesting trends (with the exception of Sweden (see Note 4 below)). Apart from Canada, Mexico and Russia (and also Brazil, since it was added), the countries monitored here are showing a downward trend in deaths.

It would appear (23 May), that Canada has passed its peak in deaths per day. The trend has been downward since 6 May.

Spain had a spike in the death rate on 23 May and a subsequent plunge on 26 May. The value on 26 May is lower than all values back to 12 May. As the number of deaths cited on Worldometer is cumulative, whatever led to the correction on 26 May goes quite a way back. As a result, I have stopped charting Spain until I can see what is going on.

Unfortunately, the UK changed their way of counting deaths on 29 April 2020. Up to this date, they had been using an isolated population of hospital deaths. They have now included deaths from other sources as well and these extra deaths were added on a single day resulting in a massive spike to their numbers on 30 April. These are shown on the chart below on the curve marked UK (R).

Fig 3. Five day moving average of the daily coronavirus deaths per million population in the UK

In order to make the data remain within the bounds of figs 1 and 2, I manipulated the data so that the extra deaths causing the spike is shared over the preceding six days. I used a linear least squares regression on those six days to predict the total number of hospital deaths on 30 April (22331 as opposed to the recorded total number of deaths, 26097). The difference in recorded and predicted deaths was then shared over the six days. This new curve is shown in Fig 3 as UK (M). This increased the range of Fig 2 from 14 to 20. In order to reduce this further, I spread these extra deaths over all data points preceding 30 April. This new curve is shown as UK (M2) in Fig 3, which gives a more acceptable range of 16 in Fig 2. It also makes the UK death rate the highest in my sample (Figs 1 and 2), however, the curve now follows the same trend as the one with hospital deaths alone.

I have taken the countries with a low daily death count from the above charts and plotted them on a new graph with a few other countries with a low death count. Here, I just show the graph with the five day moving average.

Fig 4. Five day moving average of the daily coronavirus deaths per million population of some countries with a low daily death rate (see Note 8 below)

Australia and South Africa showed similar death rates until 30 April, then South African rates started increasing and Australia decreasing (see Note 5 below). India, for the size of its population, shows a very low death rate, even if it is increasing slightly.

Around 17 May, South Africa's death rate started increasing drastically.

Since 28 May, Iran's death rate has been increasing, so much so that on 22 June it was higher than when I started recording the data. This increase started roughly one month after the start of Ramadan(23 April), so Ramadan has probably caused the increase.

As the South African curve moved beyond the confines of Fig. 4 on 8 July, I have started a new graph displaying the countries which started their peak after 11 April. I have included Iran, as it is going through a second peak.

Fig 5. Countries peaking after 11 April


1: The data collection was started, and sorted on total deaths per million population, on 23 April 2020.

2: For consistency, I try to collect the data daily at 11:00 SAST (GMT+2).

3: A much cited figure is that the USA has the highest number of deaths. From the graphs you can see that other contries are far worse off than the USA with regards to deaths per million population.

4: Sweden's stats are rather weird. It seems as though the pathologists take weekends off, then catch up during the week.

5: On 10 May, I stopped charting Australian results as the death rate had been zero for five consecutive days.

6: I have added Brazil to Fig 2 to compare it to the northern hemisphere countries I have been monitoring. Like Sweden, Brazil does not have a formal lockdown policy.

7: I have stopped updating Fig 1 as it is really getting too messy to understand. The curves on Fig 2 are still separated enough to make sense.

8: Germany, Portugal, France and Italy were added to Fig 4 when the daily death rate dropped below, and remained below, 2 deaths per million population per day.


The population numbers were taken from Google on 23 April 2020

The deaths per country is from Worldometer. Worldometer does have a table filled with various numbers relating to the coronavirus for the whole world, including deaths per million. However, I was interested in a much smaller subset.

Data preceding 23 April 2020 was collected off Worldometer's daily death graphs for each country shown above. I could find no other way to access this data.

(updated: 1 June 2020)