Visualizing worldwide influenza and SARS-CoV-2 activity over time
August 15, 2020: Major updates to article and animations
Influenza viruses travel across the globe in a wave. In much of the northern hemisphere influenza activity picks up in late November, early December, peaks around February, and then starts to go down, only to emerge in much of the Southern hemisphere around May before reaching its peak there around July or August, then decreasing and reemerging in the Northern hemisphere. In the tropics in both hemispheres the situation is different and more complex, with influenza activity in many areas roughly tracking the rainy season.
Remarkably perhaps, there is no scientific consensus on exactly why influenza follows this pattern, why it peaks in the winter months and in the rainy season. It could have to do with the climate itself: The virus may survive better in colder and drier weather, although the latter wouldn’t explain why it peaks in the rainy season as well. Or it may be due to the lifestyle changes colder and wetter weather induces: People spend more time indoors, which may cause vitamin D deficiency, making people more susceptible to disease. Or more time indoors means people are in closer contact with each other, increasing the risk of transmission. These are just some of the possible explanations that have been proposed.
What about SARS-CoV-2 and the disease it causes, Covid-19? Like influenza, Covid-19 is a respiratory illness caused by a virus. So does it follow a similar seasonal pattern as influenza?
The problem is that to the extent that we are not sure of the causes for influenza’s seasonality or know just to what extent SARS-CoV-2 is similar to influenza in relevant respects, we do not have a good theoretical basis for predicting whether it will follow a similar pattern.
But we can observe what has happened so far. And that is what we do here. Starting in December 2019, for each month we show typical influenza activity for that month as well as the number of new fatal SARS-CoV-2 infections that month.
Influenza and SARS-CoV-2 World Map and Timeline
For a more detailed look at these waves of influenza and SARS-CoV-2, you can go to the animated timeline.
The colors — white and three different shades of blue — represent the different levels of flu activity, as explained below. The darker the color the more common significant influenza activity is in that country in that month. Countries for which no influenza activity data was available remain white throughout the animation.
The red dots represent Covid deaths per 1 million population. The bigger the dot the higher the number of deaths.
You can pause, play and change the speed of the animation.
And you can click on countries and the red dots for more information.
How did we measure influenza and SARS-CoV-2 activity, respectively?
Let’s start with influenza.
Measuring Global Influenza Activity
Measuring influenza activity is an imperfect science, even more so when measurement spans the globe. Available data are imperfect and extrapolating from the data is fraught with difficulties. Our goal here was to get a rough sense or estimate of how influenza typically travels across the globe and then to see if SARS-CoV-2 follows a similar pattern. While influenza viruses are always present to some degree there are clear differences in the level of activity over the course of a year. And it is these seasonal differences in the level of activity that we are after.
- look at the past three years (2017–2019) and calculate for each country the average level of influenza activity for a year
- then check for each of the 12 months of the year how many times influenza activity in that month was above the average for that year
- and finally, use this as the basis to give that month a score on a scale from 0–3
For example, if in France in the month of January there was above average activity in all three years, it receives a score of 3 (1 for each year). If there was above average activity in January in only one of those three years, it receives a score of 1. And when in none of the three years there was above average influenza activity that month, it receives a score of 0.
It is important to note that this scale of 0–3 does not express the average intensity of influenza activity in a month. Instead, it expresses how many times in the past 3 years influenza activity was above the annual average. If we had taken the average intensity for a month over that three year period, then if the intensity was extremely high in one year but below average in the other two years, it could have still come out as a high intensity month. But such a number would tell us less about what we are after: how influenza viruses typically travel across the globe, how common it is for there to be significant influenza activity in a given month.
We intentionally did not include data on influenza activity during the SARS-CoV-2 pandemic itself, both because the influenza data is not always available yet and because the SARS-CoV-2 pandemic may have skewed the data: Many healthcare systems were overburdened or had reduced access in anticipation of a surge of SARS-CoV-2 patients. A likely result has been that a meaningfully large number of people with influenza did not access healthcare services in the past few months and hence would now not be included in influenza statistics. It is also quite likely that at least some vulnerable patients with SARS-CoV-2 would have instead been hospitalized or even died from an influenza virus had SARS-CoV-2 not been around this year. So to get a sense of the typical wave of influenza activity it would have been unwise to include data from what may very well have been an outlier year, at least in terms of the data that is or will be available.
Given the difficulties, caveats and choices mentioned above, the numbers we have arrived at should be treated with caution. We welcome criticism, comments and suggestions on how to improve our method of quantitatively measuring and visually expressing the typical wave of influenza activity across the globe. You can contact us at this address.
The data that is the basis for our scoring of global influenza activity is available here (zip). Specifically, we use the data in the ‘ALL_INF’ column as the measurement of influenza activity.
Influenza at the State or Provincial Level
The animated timeline above shows how influenza activity travels across the globe. It shows monthly changes in influenza activity per country, and because it does this for all the countries in the world, it shows the spatial and temporal pattern of influenza activity across the globe.
But the level of detail is limited. Some large countries have complex regional patterns of influenza activity. For example, as influenza activity in the northern part of India peaks in February and starts coming down in March it starts increasing in the southern and then central parts of the country to peak there in July and August. Such level of detail is lost by using only national and not local data.
That is why for some large countries we have also investigated and visualized influenza activity on a state or provincial level.
We did this for
and we may add other large countries in the future.
For a world map of influenza activity that also includes this more detailed view of these four countries, go here.
Please note that for now these more detailed maps at the state or provincial level only cover influenza activity. SARS-CoV-2 data at that level will be added in the near future.
Also note that the data sources and methods we used to estimate influenza activity levels on the state or provincial level are different from those we used for country-level influenza data, and from each other. We list our sources and describe our methods for these local level overviews here.
Measuring SARS-CoV-2 Activity
Now that we have explained how we arrived at our overviews of influenza activity across the globe, let’s return to SARS-CoV-2. How did we measure SARS-CoV-2 activity for each month?
The animation uses deaths data as a proxy for SARS-CoV-2 activity. Why deaths instead of infections? Data for the latter depend to a large extent on how much testing is being done. With big increases in testing there can be big increases in the recorded number of infections without there being corresponding increases in the number of actual infections. It’s just that many infections that would have gone undetected in the absence of widespread testing are now recorded and included in the official statistics.
To be sure, that doesn’t mean there are no problems with using deaths data as a reliable proxy for SARS-CoV-2 activity. For example, there are measurement problems: Not all countries or regions use the same criteria to determine which deaths count as SARS-CoV-2 deaths. Some countries count everybody who tests positive for Covid and subsequently dies as a Covid death, no matter whether Covid was the primary cause or even meaningfully contributed to the death. Other countries use much stricter criteria. That makes comparisons between countries more difficult. And when a country changes the way it counts Covid deaths, it can even make it difficult to compare different months for the same country.
In addition, different countries have different demographics. Given that for example older people and obese people are more likely to die from Covid, and given that some countries have more old and/or obese people than other countries do, differences in death rates between those countries need not signify corresponding differences in virus activity.
Moreover, in the past six months societies have taken all sorts of measures to combat the virus and the damage it can cause, and so reductions in the number of deaths may be the result of those measures rather than of subsiding virus activity. For example, if vulnerable groups are better protected now, deaths may go down even though virus activity may be very similar.
Lastly, while it is of course a positive development that treatment has been improving over time, it also creates a methodological problem: With improved treatment a decrease in the number of daily deaths does not necessarily mean a decrease in SARS-CoV-2 activity.
Despite all these caveats, however, it is still clear that compared to infection rates, death rates are a more reliable indicator of the level of underlying SARS-CoV-2 activity.
But deaths obviously lag behind infections. So if we use deaths as a proxy for virus activity we have to account for this lag. How did we do this? Because the time between infection and death is typically roughly three weeks, we took the number of deaths from the period between the 22nd of that month and the 21st of the next month, to estimate the number of infections that occurred in that month and that ultimately turned out to be fatal. For example, for the month of January the number of fatal infections that occurred that month is calculated by taking the number of Covid deaths that occurred between January 22 and February 21.
SARS-CoV-2 has only been known to the world since last November. And while we are learning more and more about the virus and how it is typically transmitted, and while we are finding that the virus does seem to spread more easily in settings that are more common during flu season, when it comes to the empirical question how the virus travels across the globe we have meaningful data for only the past six months or so.
Therefore, it is quite possible that the correlation so far between influenza and SARS-CoV-2 activity may not be due to any meaningful and persistent similarities between the two types of viruses and the ways in which they typically interact with the changing environments in which they occur.
For example, we cannot exclude the possibility that the fact that SARS-CoV-2 emerged in China at the start of flu season rather than in some other time of the year was relatively meaningless and coincidental. Maybe it could have just as well emerged in May or July instead and started its spread across the globe several months earlier and/or in different ways than it has now. And if that had happened, there might not have been much of an overlap with influenza activity patterns.
So we caution against drawing simple causal conclusions merely on the basis of the data presented here.
With more time will come more data, which is why we will continue to update this article and the animations as the year goes on.
For comments, questions, suggestions and criticism send us a message.
More Articles by Koen Swinkels
- Covid-19 Superspreading Events Database: 1,100 Superspreading Events From Around the World
- How to Reopen the Economy: 58 Suggestions for Reopening the Economy and Public Life while Continuing to Limit the Damage COVID-19 Is Doing
- The Non-Covid Casualties of the Covid Crisis: A Continually Updated Collection of Stories About The Devastating Effects of the Lockdown — (part II, part III)
- 18 Simple Suggestions for Dealing with Future Respiratory Virus Pandemics: Without a need for vaccines, contact tracing, lockdowns or any other far-reaching measures
- Covidonomics: What Will the Covid-19 Crisis Do to Our Political Economy?
- Pandemic Threat Inflation: 14 Ways in Which We May Get An Inflated Sense of the Threat Posed by Covid-19