The 7:30 p.m. press release has become the new 8 p.m. Every evening, France waits feverishly for the latest figures from the National Public Health Agency on the new confirmed cases of Covid-19 and the number of related deaths. Which will be analyzed, disseminated and dissected throughout the following hours, chased only by a new batch the next day. While a third of the world's population is now confined, it is impossible to imagine not knowing in real time the spread of the epidemic. This concern is less than a century old: during the First World War, it was only in the face of the immensity of the losses that governments began to count deaths precisely.
A hundred years later, statistics have become essential in our societies, and this is even more true in times of uncertainty. The immense success of the real-time map of the epidemic proposed by John Hopkins University attests to this. The need to quantify and evaluate the effects of the Covid-19 seems all the more significant since the enemy is invisible and its diffusion seems uncontrollable. Following the curve of the epidemic is a way of regaining control: statistics have become a safe haven.
The epidemic monitoring map proposed by John Hopkins University on March 25 CR: John Hopkins University
Statistics dependent on screening policy
Like all human activity, they are however fallible. Produced at a given time, according to a specific methodology, they are often used from a perspective that differs from that for which they were originally collected. They strive to bring complex realities into fixed boxes, and can therefore only be fragmented.
Part of the difficulty comes from the virus itself: it does not reach everyone in the same way. Some will develop no symptoms, others some symptoms that will go away naturally, when a significant proportion will develop significant disorders. So how do you qualify an “infected” person?
The figures published every evening by Santé Publique France, here on March 25. CR: Public Health France
The number of Covid-19 cases published each evening by Santé Publique France comes up against this problem. To be counted as having reached Covid-19, it obviously has to have been tested positive. However, if France has indicated that it now wants to carry out mass screening, it has so far reserved these devices for people at risk, caregivers or even patients with serious symptoms.
A very limited fringe of the population, and which does not take into account the mass of "healthy" carriers. The number of people actually infected with Coronavirus can therefore in France be two, three, ten, twenty, fifty times higher than the figures announced. What Public Health France also specifies, on its website: “due to the difficulty of identification and biological confirmation of all the cases of Covid-19, the data presented underestimate the actual number of case. As part of the transition to phase 3, the recommendations now being to carry out a diagnostic sampling only for certain categories of patients, this underestimation becomes more important. "
Comparisons by country, of the number of cases, prove to be ineffective
This crucial question of screening makes any reliable comparison between countries impossible. Each government has its own policy: South Korea tests hard, Germany recommends screening only people with flu-like symptoms, Spain follows patients by phone because they cannot test everyone ... In addition, the figures are not collected, compiled and corrected in the same way. The evaluation of this planetary pandemic comes up against national methods and explains the major differences between the number of cases identified.
The “Coronavirus app” dashboard as of March 25 CR: Coronavirus app
Data aggregation is in itself an extraordinary challenge. The “Coronavirus app” site, which strives to identify the latest figures for each affected country, gives an overview and specifies on its site: “ countries publish their figures at different times of the day, not to mention that they are all or almost all in different time zones; some countries release their statistics at a specific time when others announce them throughout the day; there may also be a substantial delay between when the country stops counting and when it publishes the figures. In France, for example, the figures released at 7 p.m. are those released at 2 p.m. Each site therefore develops its own methodology, which explains the differences between the estimates.
In France, only hospital deaths are currently counted
The number of deaths could appear as a more reliable indicator. However, it is also fragmented: in France, for example, it only lists deaths in hospital of patients already tested positive for Coronavirus. People who died of the virus at home, in a retirement home or before they could be tested with certainty, are therefore excluded. According to the figures communicated by the structures, at least 13 residents of a nursing home in Paris have died since March 11 due to the pandemic, 20 in the Vosges, 7 in Haute-Savoie,… or at least 40 more deaths. To remedy this problem, Jérôme Salomon, the Director General of Health announced Tuesday, March 24 the establishment in the coming days of a daily monitoring of mortality in institutions for the elderly.
INSEE also indicated that it would publish statistics on deaths compiled from civil status data transmitted by the municipalities on a now weekly basis. If the reason for death is not communicated, these figures will however allow an assessment of excess mortality in areas infected with the virus.
Faced with all these difficulties, the temptation to do nothing can be great. However, even incomplete and limited, statistics remain a vector of essential knowledge and understanding. Their use is necessary, if certain conditions are respected: the limits of the figures as well as the methodologies selected to present them must be explained in a transparent manner to the readers.
These statistics should then be used to serve information and provide context, not as absolute barometers. If proposing a reliable and precise count of the number of infected people is not feasible, it is however possible to rely on estimates and relay orders of magnitude to assess the trajectory of countries.
Make choices transparently
This is the objective of the epidemic monitoring dashboard offered by us since March 18. We have chosen to show the number of people infected and deaths on country-specific scales. The goal is not to compare them but to compare their trajectory. The curves make it possible to assess when each country reaches its epidemic peak and to visualize, if necessary, the impact of the containment measures.
Our dashboard to monitor the epidemic worldwide as of March 26 CR: Fig Data
On the other side of the Atlantic, one of the New York Times' data journalism teams also looked at the trajectory of the countries using an interesting methodology. Their graph presents, for each country with more than 25 deaths linked to Covid-19, the number of deaths smoothed over the previous week, according to a logarithmic scale. The color code of the curve allows you to see if the number of deaths doubles in 24 hours (red), or in several days (blue). The time scale is not uniform: the curve for each country starts on the day of the 25th death.
Comparison of country trajectories proposed by the New York Times CR: The Upshot, New York Times
In the text which accompanies the visualization, the journalists justify the choice of this indicator - the number of deaths - and of this unusual temporal representation: " The disease has reached different countries at different times but comparing them in this way makes it possible to see if the disease progresses more or less quickly once declared in a territory. The number of deaths is not a perfect indicator for assessing infection, since the death rate from this disease depends on the age and health status of the population, the accessibility of treatment, and the capacity of health systems themselves. However, it remains one of the best indicators available today for comparing countries. "
Our dashboard of hospitalizations linked to the epidemic, by CR department: Fig Data
We also use the number of deaths and the number of hospitalizations in our epidemic dashboard by department in France. Even incomplete, we believe that these data provide useful information to our readers.