COVID-19 Visualized: The power of effective visualizations for pandemic storytelling

Data is necessary for our understanding of the world, and particularly for the emergence of phenomena such as the COVID-19 outbreak. A viral pandemic is not a scenario in which intuition can provide a sense of how the spread is advancing, nor are feelings a sufficient approach to dealing with and ultimately defeating such an unseen enemy. Collecting, analyzing, sharing, and ultimately making use of data is what is needed.

This collected data needs to be efficiently conveyed to all sorts of individuals and groups, from lay people to experts and everything in between. Visualizing collected data can make its dissemination easy, and can help others understand quickly what has taken others so long to collect and analyze. After all, a picture is worth a thousand words.

I'm not an epidemiologist, and I have no interest in pretending to be one; there's enough of that going on right now. What I can do, however, is help bring to our readers some examples of great data visualizations in the time COVID-19, which is what I hope to do today.

John Burn-Murdoch works on data visualization (or, as his Twitter bio says, "Stories, stats & scatterplots") for the Financial Times. He has recently garnered a following and appreciation for his COVID-19 visualizations, which are exemplars of how quality data visualizations can clearly and succinctly tell a story. Personally, I have been relying on John's visualizations to understand the outbreak for the past couple of weeks, and generally refer them several times a day.

Though he seems to be adding additional visualizations to his daily repertoire as the outbreak progresses (and as additional data becomes available to do so), currently John produces the following visuals on a daily basis:

• Coronavirus mortality trajectories tracker for major countries
• Coronavirus mortality trajectories tracker for individual major countries, compared
• Coronavirus mortality trajectories in subnational regions
• Coronavirus case trajectories for major countries
• Coronavirus case trajectories for individual major countries, compared

If you want to get an idea of how great data visualizations can help us understand the underlying data, and succinctly convey a story, have a look at John's Coronavirus case trajectories for major countries visualization from March 26, 2020 — which charts the cumulative number of confirmed cases, by number of days since 100th case — from which much can be immediately gleaned.

Coronavirus case trajectories for major countries, March 26, 2020 (Credit: John Burn-Murdoch & Financial Times)

What are some helpful takeaways from the above?

• The US infection trajectory overtook China's on approximately the 14th day after the US reported its 100th case
• Turkey seems to currently be on an alarming trajectory, with a doubling rate of less than 2 days
• A number of Asian countries have had early success in flattening their curves
• Italy, once the undisputed epicenter of the disease, seems to be having some success in starting to flatten their curve as well
• It's easy to see confirmed case doubling rates, at current trajectories, given the overlayed broken lines, from daily, to every 2 days, to every 3 days, to weekly (US is doubling almost every 2 days; the UK is doubling approximately every 3 days)

Along with crafting great visualizations, John has been doing a good job at continually explaining a couple of important point related to visualizing this data. In particular, he explains the use of log scale and the choice of using absolute numbers rather than population-adjusted rates.

Here's a bit of what he has to say on log scale:

In the initial outbreak phase, a virus like this spreads exponentially not arithmetically, i.e a log scale is the natural way to track the spread
...
imo much concern over "reader don’t understand log scales" is misplaced.

When a reader ponders this chart, they’re asking "are these two countries on the same course", or "how many days til country X is at Y cases", not "how many pixels represent 100 cases" etc.

And here's what he has had to say on absolute numbers vs population-adjusted rates of infection:

Here’s the problem with per-capita stats for coronavirus:
• Switzerland has seen ~same pace of outbreak as everywhere else, but its per-cap figure is higher coz it has a low population
• Note "major country" in tweet. Include all countries? Iceland & Luxembourg go top
...
• Covid spreads from person to person; it’s not like cancer. It’s exogenous. Generally, and especially early in outbreak (first few weeks), higher per-capita numbers just mean smaller country, not anything different about how that country’s dealing with covid.

Here's another example of his visualizations, Coronavirus mortality trajectories for major countries — charting the cumulative number of deaths, by number of days since 10th death — also from March 26, 2020.

Coronavirus mortality trajectories for major countries, March 26, 2020 (Credit: John Burn-Murdoch & Financial Times)

A few takeaways from this visualization:

• UK is on a similar trajectory to that of Italy
• The US is crossing China in number of deaths on approximately the 21st day since its 10th death
• The current US mortality trajectory has them doubling their number of deaths approximately every 3 days
• The commencement of national lockdowns are clearly marked, making their results easier to look for over the subsequent time period

His work is featured in the continually-updated Financial Times article Coronavirus tracked: the latest figures as the pandemic spreads, which includes live versions of all of John's charts. Given the current pandemic situation, and Financial Times' assessment that the public should be able to access as much vital information on the topic, they have made the article free to read for non-subscribers. I would suggest keeping both John Burn-Murdoch's Twitter feed and the Financial Times article handy for future reading and reference.

I hope you have found this to be an effective example of COVID-19 data visualization. If you have other such examples, please feel free to drop them in the comments below for others to discover.

 
Related:

• Coronavirus Data and Poll Analysis – yes, there is hope, if we act now
• What is the most effective policy response to the new coronavirus pandemic?
• Covid-19, your community, and you — a data science perspective