To end the Coronavirus pandemic, we have a clear and simple goal: cases need to go to zero everywhere.
Viruses don’t respect borders – even the 1918 influenza pandemic reached remote islands within months, and that was long before the days of global air travel. It is therefore the entire world that needs to make progress against the virus if we want to prevent a situation where countries either need to lock themselves off from the rest of the world or suffer recurring COVID-19 outbreaks.
Only if we end the pandemic everywhere can the pandemic end anywhere.
For this reason our work on the pandemic at Our World in Data is always global. We do not only focus on a few countries, but are instead building the database and visualization tools to allow everyone to monitor the global situation.
In this post I want to introduce our new visualisation to allow everyone to monitor the global fight against the pandemic.
It would be straightforward to build the appropriate visualization if the total number of cases was known. We would simply report new cases over time – a chart that the literature refers to as the epidemic curve, often shortened to epi curve.
However, with the COVID-19 pandemic we are in the unfortunate situation that the number of total cases is not known. Only a fraction of total cases – those confirmed by a test – is known.
This is why we spent recent months building the database and the visualization tool to make this variation of the epidemic curve possible.
Below I will explain in more detail what we can learn from this chart, but the quick summary is that each line in this chart shows you new confirmed cases over time – that part of the chart is simply the classic epi curve – while the line color shows you the quality of the data at each point in time. Darker shades of blue indicate that here the confirmed case count is likely closer to the true total number of cases.
This allows us to see whether the world is making progress against the pandemic: what the world needs to achieve is that all lines turn into dark blue and hit zero.
Per capita: The same chart, but showing cases per capita, can be found here.
As the chart title says, the epi curve for each country shows the number of confirmed cases only: those cases that were confirmed by a laboratory test.
Confirmed cases are only a fraction of the number of total cases. The question is: how big of a fraction of the total number of cases are confirmed? This is not known, but we can get an indication by looking at the extent of testing.
How big a fraction of total cases get confirmed depends on how much a country actually tests. To understand the spread of the disease we need to interpret the number of cases – the epidemic curve – in light of how much testing for COVID-19 the country actually does.
For this reason we at Our World in Data built a global database on testing. We made it available for everyone – epidemiologists, the WHO, the UN, and many governments rely on our database for their daily work. We update it continuously.
When we decided to build the testing database we did so because we wanted to build this chart: one that brings together the epi curve with the relevant information about how much a country actually tests to allow everyone to monitor the quality of the reported data.
We can only include countries in this chart for which we have data on testing. You find the complete overview – including a detailed description of each source – in our testing database here. Our data covers currently 66% of the world population.
Just as it is not informative to look at case counts in isolation it is also not informative to look at the number of tests in isolation. We need to see the number of tests in relation to the size of the outbreak: countries with large outbreaks need to do much more testing to monitor the spread of the pandemic than countries where the disease is under control. See the recent work of my colleague Joe Hasell who looked at this in detail.
Based on this principle we calculate a quality metric for the case counts that answers the question: How many tests does a country do to find one COVID-19 case?
The world map here shows this data – it is exactly the same data shown as the line color in the chart above.
By moving the time-line below the map you can see how this metric has changed around the world, by clicking on a country you can see how it changed in every country.
The world map shows enormous differences across countries:
- Some countries, like Australia, South Korea and Slovenia do hundreds, or even thousands of tests for each case they find.
- Others, such as Brazil, Mexico, and Pakistan, do very few tests – five or fewer – for every confirmed case.
Countries that do very few tests per confirmed case are unlikely to be testing widely enough to find all cases. The WHO has suggested around 10 – 30 tests per confirmed case as a general benchmark of adequate testing.1
The countries that do more than 30 tests per case are shown in shades of blue. Those that find a case for every 30 tests or fewer are shown in shades of orange and red.
In countries that test very little in relation to their outbreak – shown in shades of red in the chart – many cases are likely to go unreported. In these countries, the number of confirmed cases indicated may represent only a fraction of the total number of cases.
The widely available data on the number of confirmed cases only becomes meaningful when it can be interpreted in light of how much a country tests. This is what the chart shows.
Whether a country makes progress can be seen by looking at the two metrics the chart presents:
- The trajectories show the daily number of cases. The goal is for every line to bend towards zero.
- And line color gives an indication of the quality of a country’s data at each point in time.
If a country finds a case for every few tests they perform the line is shown in shades of red. Here it is likely that the unknown known number of cases is high. The quality of the data for these countries’ during these times is poor.
The darker shades of blue mean that a country does many tests for each case it finds; in these countries the share of undetected cases is likely smaller – the quality of the data, ‘better’. The goal is that a country tests widely in relation to its outbreak, shown by the line color turning into dark shades of blue.
By hovering over the epi curves – or finding a country via ‘Select countries’ in the bottom left – you can see how the testing relative to the size of the outbreak has changed over time for each country.
The charts below highlight two very different groups of countries.
- The data for Thailand, New Zealand, and South Korea shows that these countries monitored the outbreak well from the start or caught up rapidly after an initial outbreak. Eventually they were able to bend the curve and bring down the number of confirmed cases, while increasing the ratio of tests to confirmed cases. These are not the only countries, that achieved this; you can add for example Austria, Iceland, Slovenia, Tunisia, or Latvia to the chart and you will find similar trajectories.
- The data for Brazil, Mexico, the United States, UK, Sweden, India, Pakistan, South Africa, and Nigeria shows that these countries test little relative to the size of the outbreak. Additionally these countries report unfortuantely still very high daily case counts – their lines are red and far from zero.
In the early phase of the outbreak there was not always a big difference between these two groups, but while the first group of countries managed to catch up with the outbreak, the second one did not achieve this yet.
We believe it is the most important chart to track the global outbreak of COVID-19. It shows us whether reported cases go to zero and crucially gives us an indication of how good that reporting is.
As all our work on the pandemic we will update this chart every day so that you can monitor whether the world is making progress to our global goal or not.
To be safe anywhere, every region in the world needs to make progress against the pandemic – and this means dark blue lines hitting zero. Currently we are far from it.
These two versions of the chart show not the number of days, but the cumulative count of confirmed cases on the x-axis: