An excellent visual comparison of how COVID-19 has advanced in different countries on the personal website of Chris Billington, an Australian physicist and software developer.
The steeper upwards the yellow (cases) line is the faster COVID-19 is spreading in that country. With the exception of countries that tested very intensively early on (Taiwan, South Korea, Singapore, and Hong Kong) you can see that the number of deaths (red) will go up at the same rate after a delay of about two weeks. You know if the control measures in a country are working if the number cases (yellow) starts starts to flatten out. When this happens deaths will continue to go up for around two weeks before they flatten out.
Note that Billington’s data updates in real time.
Published on the author’s own website, Mar 2020
NB: These graphs are cut and pasted from the original format, as it was too wide for them to be posted here in the same way.
Click on this link to see in their original format.
Caveats and explanations
- Exponential projections are based on the growth rate over the most recent 5 days of data for each country. This is a rough answer to the question “What will happen if nothing changes about testing rates or control measures?” This can model both growth and decline in active cases, both of which should be approximately exponential while conditions remain fixed (and whilst most of the population has not been infected yet).
- The “ICU beds ≈ critical cases” line is based on the figure that 5% of diagnosed cases are critical. This is a measure of when each country’s healthcare system is completely overwhelmed. Healthcare systems will be overwhelmed to a great extent much before this point is reached, however, since ICU beds are also required for non-COVID-19 cases, are not geographically distributed identically to infections, and medical staff and ventilator availability limits healthcare capacity before beds actually run out. For example, some regions in Italy were already triaging patients for critical care when the national figures were a factor of ten below running out of beds.
- The “Δ active growth rate” and “Δ death growth rate” statistics are measures of whether the growth/death rates are speeding up, or slowing down. E.g., Δ active growth rate = -1.0%/day² implies that the growth rate of active cases is one percentage point lower each day than the previous day. These statistics are computed by comparing the fitted growth rate for the most recent five days to that of the preceding five days. This information is not used in the projection, since there is no particular reason to assume that a recent change in the growth rate represents an ongoing trend.
- Data quality is of limited by testing and reporting within each country, and the fact that these conditions are changing in time. Many countries seem to be either underreporting recoveries, or batching the reports only every few days. Data on recoveries is thus somewhat unreliable, and since active cases is computed as confirmed – recovered – deaths, this affects the active case count.
- 2020-03-23: Data source changed from Johns Hopkins to ulklc/covid19-timeseries, as Johns Hopkins is dropping recoveries from its datasets (among other issues).
Source for case numbers: ulklc/covid19-timeseries on gihub
Plots by Chris Billington. Contact: [email protected]
Python script for producing the plots can be found at https://github.com/chrisjbillington/chrisjbillington.github.io/blob/master/covid.py. The script is messy, as things are in a state of flux and I’ve been experimenting and switching data sources. It will likely be cleaner once it becomes clear which data source is best.