CoVID 19 Worldwide Growth Rates

Updated: 17th March 2020, 17:28 UTC (added US state graph, added Netherlands to graph 2).

This page presents historical SARS-CoV-2 coronavirus infection information. I try to present it in a way that helps inform about how the virus is progressing throughout the main outbreaks worldwide in a way that allows countries to be compared and perhaps lessons learned. The underlying infection and containment processes are complicated, and conclusions you or I may draw from these graphs are not true predictions. For that, you'd need much more data about what is being done in each country, and about how the data is collected. Comments are my interpretation only. Nevertheless, I hope you find it useful.

Mark Handley, UCL.

Contents

Graph 1: Italy, France, Germany, Spain, UK, Netherlands

Graph 2: Italy, France, Germany, Spain, Switzerland, UK, Netherlands

Graph 3: Lombardy, Italy, Switzerland, Austria

Graph 4: Italy, Denmark, Norway, Sweden, Finland, Iceland

Graph 5: Italy, Belgium, Ireland, Portugal

Graph 6: Italy, Greece, Poland, Czech Republic, Slovenia, Romania

Graph 7: Italy, France, Germany, Spain, Switzerland, UK, Netherlands (Linear Scale)

Graph 8: China, Italy, Iran, France, USA, South Korea, Japan

Graph 9: Italy, Iran, France, USA, South Korea, Singapore, Japan

Graph 10: Italy, Switzerland, Washington, New York, New Jersey, Colorado, California, Florida

Graph 11: Italy, Iran, France, USA, South Korea, Singapore, Japan

Graph 12: Italy, Spain, USA, Qatar, Thailand, Malaysia, Brazil, Bahrain, Indonesiam Kuwait, Egypt, India, Australia

Graph 13: Qatar, Thailand, Malaysia, Brazil, Bahrain, Indonesiam Kuwait, Egypt, India, Australia

Graph 14: China, Italy, Iran, France, USA, South Korea, Japan (Linear Scale)

Europe

• The graph shows number of confirmed cases, plotted on a log scale, against time. The country curves are shown offset by the amounts shown.
• Italy's daily increase rate continues to reduce, with today's increase being 14%. This data point is missing some data from Apulia and Trento, where rates recently have been rising, but this missing data is unlikely to change the overall picture greatly.
• Over the past few days France, Germany, Netherlands and Spain have all become aligned with the 22 percent daily increase curve that Italy followed at this point, rather than the higher 35% curve all the large European countries initially followed..
• The UK results show a reduced daily increase rate the last two days. However, the UK switched to only testing hospital admissions recently, so these two datapoints are not directly comparable to earlier data. We will need to wait a few more days to establish a new trend in order to see whether there is any change in the original 35% increase rate or not.
• This graph previously showed Switzerland. I've broken this out separately

• The graph shows number of confirmed cases per million inhabitants, plotted on a log scale, against time. The country curves are shown offset by the amounts shown.
• Normalizing by population makes little difference for most of the countries shown.
• Switzerland has a relatively small population, and so the number of cases per million inhabitants is much higher. The graph places Switzerland on the 35% line seven days behind Italy, but yesterday's datapoint is closer to 3 days behind Italy as a whole. No absolute data was released today, so today's datapoint is missing. At the moment, the Swiss and Italian trend lines are diverging; there's no good way to plot this situation as Switzerland is no longer following the Italian track. Normalized by population, Switzerland follows about 9 days behind Lombardy (see next graph).
• Spain how seems to be firmly on the 22% curve. This places Spain 5.5 days behind Italy on this metric.

• The graph shows number of confirmed cases per million inhabitants, plotted on a log scale, against time. The country curves are shown offset by the amounts shown.
• I have aligned the 35% daily increase rate parts of the four graphs, so the points at which they diverge can be seen. As a result the graph shows Lombardy as being 4.5 days ahead of Italy, but a more realistic value based on today's data points is seven days ahead.
• Switzerland has now tracked along the 35% curve about six days further than Italy did, so comparing with Italy as a whole no longer seems to be the right baseline. I've added the Lombardy results as a new reference point, but Switzerland has now stayed on the 35% daily increase curve for two days longer than Lombardy did. Beware though that these time estimates depend on both Lombardy and Switzerland detecting roughly the same fraction of cases. Switzerland is only testing severe cases at this point, and Lombardy was clearly overloaded at the same point. In both cases it is likely that at least the 10-20% requiring medical attention are detected, but there could easily be 2-3 days error in these time estimates.
• Austria has been on the 35% daily increase curve, but today's datapoint is lower. If this reduction continues, Austria may follow the Lombardy curve, but it is still too early to tell.

• The graph shows number of confirmed cases per million inhabitants, plotted on a log scale, against time. The country curves are shown offset by the amounts shown.
• Norway had been tracking at roughly 35% daily growth, but the last five days have established a newer trend at around 22%. Today's increase was less, but it is too early to know if this will continue. Norway currently follows roughly four days behind Italy and about eleven days behind Lombardy. Beware though that, as with Switzerland, these time estimates depend on both Lombardy and Norway detecting roughly the same fraction of cases; likely there's 2-3 days of error in this estimate.
• Sweden appears to have departed the 35% increase curve, so I have aligned today's datapoint with the Italian curve. If the recent decrease continues, Sweden might avoid following Italy's track, but there was a dip 9 days ago, and it's possible the current dip is an echo of that. Today's datapoint puts Sweden seven days behind Italy.
• Finland had been experiencing a faster increase than 35% for several days. This was unusual, though we have seen it when a country has been catching up on testing missed cases, or when it is importing a large number of cases from abroad. The last two data points show a marked reduction in the daily increase rate. While this seems promising, it is reported that the testing strategy has changed, so we will have to wait a few more days to be able to observe the underlying increase rate again.
• Iceland suffered a large number of imported cases relative to its very small population. After that burst stopped, the increase rate had been a little lower than 22% for a few days, but the last few days data now show Iceland on roughly the same curve as Italy, two days ahead. This places Iceland five days behind Lombardy. Today's data is not yet available. Note that the absolute number of cases in Iceland, 161, is still relatively low, with most of them being imported cases. Also of note is that 2500 people are estimated to be self-isolating, which is nearly 1% of the population.
• Denmark imported a large burst of cases, particularly from Austria. The rate went from 35% initially, to ~150% daily increase for three days. Significant mitigation measures were very rapidly introduced. These will, however, not show in the graph for another 2-5 days. The last three datapoints show much a lower increase rate, but testing policy changed on March 12, so it is hard to say whether this reduction indicates an actual reduction in the underlying increase rate. It is also possible that the slowdown is because the initial burst of imported cases has now passed, and there will be an echo of this burst in several days time, as untested people they infected become symptomatic.

• The graph shows number of confirmed cases per million inhabitants, plotted on a log scale, against time. The country curves are shown offset by the amounts shown.
• Austria has been tracking up the 35% per day increase curve for quite some time, though for some reason the data is noisier than neighbouring countries. No obvious signs of mitigation efforts in the data yet.
• In Belgium, the inital early ramp-up appears to be mostly due to imported cases from Italian ski resorts. Since these imports stopped, Belgium seems to be on a similar 22% daily increase curve to that followed by Italy.
• The Irish data was initially somewhat noisy, as the absolute counts were low but for the last seven days Ireland has tracked along the 35% daily increase curve.

• The Greek data is very noisy; it is hard to establish any trend, but it vaguely follows the 35% increase regime over a two week timescale.
• Slovenia had been been seeing sustained faster than expected increases. This may be due to importing cases from a neighbour with a higher infection rate. The last four datapoints suggest Slovenia has moved to the 22% daily incease rate that Italy followed.
• I had previously though The Czech Republic had followed Italy onto the 22% daily increase curve, but the last two days data suggests thc Czech Republic is still on the 35% curve.
• Romania had been following the 35% increase rate line fairly consistently, but now appears to have moved to the 22% daily increase curve that Italy followed.
• Poland is in reasonable shape at the moment. Not only are cases per million inhabitants pretty low, but Poland has already taken strong mitigation efforts. I don't expect this to show in the data til around March 19th though.

• The graph shows number of confirmed cases, plotted on a linear scale, against time. The country curves are shown offset by the amounts shown.
• I normally use a log scale on the y-axis because exponential growth gives a straight line on a log-linear graph. This allows exponential growth rates to be compared. However, many readers don't like log-linear graphs, so this one is for them. It has to be truncated, or you cannot see any information. It doesn't really show anything additional over the log graphs, except it looks scarier.

• The graph shows number of confirmed cases, plotted on a log scale, against time. The country curves are shown offset by the amounts shown.
• The early part of the China curve is shown for comparison. I'm not sure how accurate the numbers are in the early stage because they didn't initially know what to look for, and the way of measuring changed part way through this period. The initial increase rate is fairly consistent with the 35% daily growth seen in Europe.
• Other than China, South Korea is the only country to have a high sustained increase rate, and then bring the virus under control.
• Iran tracked a little above a 35% increase rate until nine days ago. Since then, Iran has experienced consistent 11.5% exponential growth. This is likely to indicate that measures taken there have been effective, but exponential grown at 11.5% per day still indicates a doubling of cases every 6 days, as opposed to 2.5 days at 35% daily increase.
• The USA is tracking very closely along the 35% growth line. Testing in the US has been very limited, so the absolute numbers are likely to be undereported compared to other countries. Such under-reporting does not change the inferred daily increase percentage, but it does change the timeline. If the number of actual cases in the US is underreported by, say, 50% compared to Italy, this would move the US 2.5 days closer to Italy.
• Cases in the US are concentrated in a few states, in a similar way to how cases in Italy are concentrated in Lombardy.
• Japan is an enigma. The testing rate there is low, so cases may be being underreported. However consistently missing the same fraction of cases would not affect the exponential growth doubling time, which has been consistent for a long time. It seems more likely that early measures taken there are being effect at reducing the increase rate, but are not sufficient to avoid exponential growth altogether.

• The graph shows number of confirmed cases per million inhabitants, plotted on a log scale, against time. The country curves are shown offset by the amounts shown.
• Showing cases per million inhabitants does not greatly change the overall picture compared to the previous graph.
• In this view, the US moves a few days further behind Italy. This is probably deceptive, and it might be better to show data for Washington State, for example, on this graph.
• I've added a curve for Singapore, which contained the initial imported cases very well. The graph shows a worrying slow but steady super-exponential growth recently though, perhaps indicating that current measures are starting to not be effective, or that more cases are being imported again.

• The graph shows number of confirmed cases per million inhabitants, plotted on a log scale, against time. Each US state curve is shown offset by the amount shown in the key so as to align the curves.
• Most US states are a long way behind European countries, but several now have comparable rates per million inhabitants. I have plotted the states with most confirmed cases.
• Washington is ahead of many European countries, and only 2 days behind Switzerland. It is not exactly clear how to place this curve though - European countries moved from 35% daily increase rate to 22% over time, but Washington may have done the inverse. The data is too noisy though to have any certainty.
• Colorado, New York and Florida appear to be increasing at close to the 35% daily increase rate seen in Europe.
• California shows a slightly lower increase rate, and the data is very clean. It is possible the slower increase rate is due to warmer weather.
• New Jersey appears to be catching up on testing, so any underlying rate is impossible to determine.

• The graph shows number of confirmed cases per million inhabitants, plotted on a log scale, against time. The country curves are shown offset by the amounts shown.
• Canada and Israel are both tracking along the 35% daily increase curve that most other countries have followed before social isolation measures were introduced.
• It is too early to say if Israel's lower increase today was significant.

• Very few warm countries have enough cases to establish a clear increase rate trend. The graph shows number of confirmed case for some typical "cool" countries and some "warm" ones. During the timescale shown, Spain had experienced cool weather.
• Australia, Malaysia, India, Bahrain, Kuwait and Egypt are all experiencing roughly 14% daily growth. It is starting to see likely this is due to their warm climate.
• Brazil and Qatar have both recently experenced a very rapid increase in cases. This seems likely to be due to the virus spreading in some communities unnoticed, and testing is now catching up as happened in Italy, or due to a sudden influx of imported cases, as happened in Denmark. Qatar now has shown a four-day trend close to the 14% daily increase curve, but it will require several days more data to have confidence in this. The Brazilian data is still to noisy to establish a clear trend.
• Although Singapore has a warm climate, I have ommited it from this graph, because strong contact tracing there dwarfs any climate-dependent effect we might observe.

• The graph shows number of confirmed cases per million inhabitants, plotted on a log scale, against time. The country curves are shown offset by the amounts shown.
• This graph shows the same warm countries as the previous graph.
• Australia had been tracking along the 14% curve, but the last few day this has increased to 25% daily increase per day. Australia has been cooler recent as it moves into Autumn.
• In general, all the warm country data is quite noisy. Partly this is due to there being few cases in each country. All these countries are showing a lower average increase rate to that see in Europe or the US.

• The graph shows number of confirmed cases, plotted on a linear scale, against time. The country curves are shown offset by the amounts shown. In the other graphs I use a log scale on the y-axis because exponential growth gives a straight line on a log-linear graph, and this makes changes in the growth rates visible and allows exponential growth rates to be compared. However, many readers have told me they don't like log-linear graphs, so this one is for them. It has to be truncated, or you cannot see any information. It doesn't really show anything additional over the log graphs, except it looks scarier and shows how much worse things are set to get if effective measures are not taken.
• I have shown the China curve until the daily increase rate drops below 0.1%. The glitch in the center of this curve is a change in measurement methodology. If Italy follows a similar curve, this wave of the epidemic may have abated in 30 days time, but this prediction should be tempered by the fact that Italy has not yet reached the logistics curve inflection point. On this basis, the worst is yet to come for many Italian hospitals.

Thoughts

While you're sitting pondering your mortality, think how astounding it is that one single viral particle from a bat can replicate so far, so fast, and cause so much trouble! Biology is truly amazing!

FAQ

Q: Where does the data come from?

Where possible, the data comes from the relevant national authorities, as they tend to be more up to date. In some cases I'm using the WHO daily briefing, but they lag the national authorities somewhat. The wikipedia pages contain links to the national authorities. Some are Finland, France, Germany, Ireland, Norway Qatar, Slovenia, South Korea, Spain, Switzerland, UK,

Q: Can I have your data?

A more complete dataset is the Johns Hopkins one. It tends to lag the national data sources and the WHO, but it is complete and machine readable.

Q: Can I use your graphs?

Yes, they're under a CC0 "No rights reserved" license. You can use them for any purpose. I'd prefer you link back here though, as the explanations add important context.

Q: Different countries are testing at different rates. How can you compare the data?

So long as the fraction of actual cases being detected does not change, this does not affect any inference we can make about the growth rate. 35% growth is still 35% growth, whether we measure 100% of the cases or 50%.

If, for example, Italy is detecting 50% of the cases and the US is detecting 25% of cases, this affects any predictions of how far the US is behind Italy At 35% growth, cases double every 2.5 days, so this undersampling would show the US 2.5 days further behind than it really is.

Likely no-one except Korea and Singapore are getting close to 100% of cases. Probably everyone is detecting at least 20% of cases, because those require medical attention. We don't really know what fraction of cases are missed, but the difference in sampling between countries might skew the delays by a few days in either direction.

Q: Comparing with Italy as a whole is flawed. Shouldn't you be comparing with Lombardy?

Perhaps. Certainly cases in Italy are concentrated in the North at the moment. Lombardy is running about seven days ahead of the rest of the country, as measured by cases per million inhabitants. I've shown this in graph 3. The problem is that cases in some other countries, notably France, Spain and the USA, are also concentrated, so comparing these entire countries with Lombardy is biased in the other direction. These graphs give a crude indication, but they're never going to be able to track problems down to the local level, where an individual town is overwhelmed by cases before this becomes commonplace in a country. Just remember, to compare with Lombardy, a rough approximation is to move the Italy curve seven days to the right.

Q: Shouldn't you break out China by province?

At the moment I'm only showing China on the graphs that show absolute counts, not counts per million inhabitants. As the vast majority of Chinese cases were in Hubei, there's not a great deal of difference on these graphs between showing Hubei and showing all of China. The population of Hubei provice is around 57 million, which is roughly the same as Italy. Thus the data for China is reasonably comparable with the data for Italy.

The Chinese data is somewhat suspect in the early days, so it's not very good for comparison until the epidemic gets a fair way along. Mostly I'm using Italy to provide a model for how other places might progress, as it's a better fit. But as of March 16th, I've changed the last (linear) graph to use the Chinese data to show one possible direction that Italy might folliow.

Elsewhere in China, cases that arose were stamped on quickly, and pretty much all of China shut down until that was complete. In Europe we failed to learn from that model, so we're all much further along than Beijing, Shanghai, or similar provinces ever got.

Q: Why didn't you show my country?

A1. Until the number of cases reaches around 100, the data tends to be too noisy to draw conclusions

A2. The graphs are pretty cluttered as it is.

A3. I'll add more as I find time.

Q: You aren't an epidemiologist. Why should I listen to you?

You probably shouldn't. I'm a computer scientist and I've spend decades analysing data, but you should talk to a real epidemiologist if you want to understand the underlying causes. Computer scientists do know a lot about exponential growth though.

Q: I'm a journalist. Will you appear on my TV show?

No. You should have a real epidemiologist on your TV show.

Mark Handley, UCL.