COVID-19 Analytics

I have been fascinated by the data coming out of organisations such as the ECDC. So, I dusted off my trusty PowerBI and tried analysing the data for anomalies and insights.

I should note that I am not an epidemiologist, just a software engineer with a love of data and that this analysis is a personal project, not anything endorsed by my employer. Please don't use this analysis as medical advice, please listen to the advice of your medical experts, which generally says we should be avoiding physical contact with each other as much as possible.

• Worldwide analytics
• NZ Analytics

New Zealand

There's not much to say about New Zealand, I am very pleased that so far New Zealand has no deaths and the cases are so few that they have been captured manually on the Ministry of Health's website. Still it's interesting to look at the spread and the demographics of impacted people.

It's clear the cases are following the classic exponential curve for virus outbreaks and even at these relatively small numbers the rate of increase is concerning. So far there are no deaths. We have to accept that there will be deaths though, this is a global pandemic and we have to remain ever vigilant.

On to the world...

Summary observations worldwide:

My observations are:

• The "true" mortality rate appears to be around 1.4%
• China and South Korea appear to have the outbreak under control for now
• The outlook for Italy remains grim
• Spain is following rapidly down Italy's path
• The Netherlands is early on in their exposure but I am concerned about the increasing rate of fatalities at a higher rate than most other countries.
• France is later in their exposure but fatalities there are now increasing at an alarming rate, both France and the Netherlands have approximately the same number of deaths per million population (25).
• Indonesia is too soon to tell, though they appear to have a high fatality rate.
• Iran have a very high raw number of deaths, but their rate of increase in deaths shows that they are not likely to end up in Italy's situation.
• The UK appears to be managing the outbreak well, though they have many cases and it may be too early to tell...
• The US has a relatively low mortality rate right now but I expect this to regrettably rise dramatically due to their overwhelmingly exponential rise in cases. That said, when considering America's population, they are not doing any "worse" to manage the virus than other Western countries.
• Switzerland has more deaths and cases than one would expect for a country of their population.
• There are a number of countries that have kept fatalities low (e.g. Norway, Israel and Singapore) and this is heartening.

I'd like to start by talking about lethality or the mortality rate of cases. I am, listening to talking heads on UK TV lamenting that they have to stay at home and can't visit their friends and claiming that the death rate is likely to be "only" about 1%, which sounds pretty low.

Yes, that's true, it's likely that of the many people who will ultimately catch the virus, "only" 1% or so will actually die. I feel a bit uncomfortable saying "only" here because that's still a huge number of lives on a global scale.

The thing that the pundits somewhat overlook is the fact that we don't actually know the number of people who have truly contracted the disease, only the number of people who have been tested positive. Most countries don't test people unless they present with symptoms at a hospital with a history of direct contact with someone who had the disease.

Of all the countries in the world, South Korea has been the most aggressive with their testing, so their data is probably as accurate as we can expect.

South Korea's lethality

A naive way to calculate lethality would be to calculate the cumulative deaths so far divided by the cumulative cases so far. This overlooks the fact that deaths lag cases by some number of days. It's hard to know how many days the lag is (it's been estimated to be 10 days since hospital admission) so I have guessed at 5 days. The graph above shows the lethality and lethality offset (by 5 days) for South Korea. The big bump in late February is likely due to a lack of adequate testing, not some systemic flaw in health care. In fact all countries appear to have a similar "bump" in lethality when calculated as a lag with cases.

South Korea settled at about 1.4% which makes sense to me as the best rate we know for the lethality of the virus.

China's lethality

China has a similar looking curve. I find it interesting that the lagged lethality converges with the naive lethality over time. I suspect that when the curves meet, the rate of new infections has dropped to the point where the cases are coming back under control.

China's rate settled at 4%, which I presume is once again about the testing regime, though we have to acknowledge that some commentators have questioned the veracity of China's data.

Italy's lethality

Italy, as an example of a country that has been suffering greatly, has a lethality of 10%, with a lag value of 15%. This tells me, if my theory is correct, that Italy has a great deal more suffering to go before these values converge.

Countries with high lethality offset value:

What about the other countries that have a relatively high lethality offset, indicating they have more trouble ahead?

Indonesia appears to be on a very dangerous trajectory, though they have a rather small number of deaths at this stage, the offset indicator suggests they should be very concerned and are not testing effectively.

We all know that Spain and Italy are struggling, but this graph highlights the Netherlands, France and the United Kingdom as countries with alarmingly high death rates. France is an interesting case, which appears to have had a resurgence of lethality in recent times.

Iran is another country on an unusual trajectory but also with an alarming death rate.

Belgium and Turkey are surprising additions to this list.

Now these high death rates could be due to testing regimes. The way we could control by that is by looking at the rate of increase of deaths. All of these countries have about a 4-5% gap between lethality and the offset lethality, what this tells me is that all of these countries are unfortunately due considerably more deaths in future.

Looking at these same countries in terms of raw deaths we have the following:

Everyone knows that Spain is rapidly following Italy. In fact, this data suggests that Spain is accelerating in deaths far faster than Italy ever did. This could be due to testing regime or healthcare, or demographics, but the reason doesn't really matter if someone is dead.

France is accelerating slower than Italy but still at a very alarming rate.

The Netherlands and the UK are relatively low in terms of deaths right now, but I am convinced that this is simply because they are early on in their exposure to the virus. Looking at them from a pure case basis, we see the following:

Top Cases

Note that France and Iran are both sloped more gradually than other countries, indicating a slower rate of infection. The rest of the countries with the highest raw number of cases have a very high classic exponential curve.

This is the point at which the United States approaches our radar with their highest rate of infection in the world so far. America's population is definitely a factor here of course, so it's sensible to come up with a metric that's based on population.

So, I divided the cases by population of each country and came up with a metric of cases per thousand population (CPT):

Looking at this, the rate of infection in the US looks to be growing at the same rate as other Western countries when taking into account population. We also suddenly become aware that Switzerland, partly do to their smaller size has a significant number of infected people per capita. This combined with a lethality rate of about 2% for them presently means that Switzerland is of as much concern as potentially even Spain.

This is borne out when we compare deaths per million population:

What does this mean? I think that it's right for us to be worried about countries like the USA based on the sheer number of people infected, but I think it's clear there are many other countries that are on a very dangerous trajectory by the numbers. I don't think anyone can afford to be complacent.

Is there any good news? Yes, so far, the following countries have experienced more than 300 cases but have been able to avoid significant fatalities:

Some of these countries, like South Africa, have only recently had cases and so this may just be a case of not enough data, but other countries like Singapore have had cases for quite some time and remain resilient to fatalities, likely due to strict temperature checks they put in place early on. Israel and Norway are wonderful positive examples of countries that have managed to keep the fatality rate low.

I hope and pray that people continue to practice physical distancing and that as a global community we will overcome this with minimal loss of life.