The PredictEst tool gathers a lot of data then modeled in the form of diagram maps etc.
The Grand-Est region has developed, with the help of institutional, academic and industrial partners, Predict'Est.
This modeling tool should make it possible, in particular, to predict the evolution of the Covid-19 epidemic.
How is it possible ?
With public data from INSEE and Public Health France.
"And thanks to the use of a mass of scientific publications combined with these existing public data, we have the possibility of predicting the evolution of the epidemic according to given scenarios", details Professor Benoît Gallix, director of the Strasbourg University Hospital Institute (IHU).
What will be the number of hospitalizations due to Covid-19 in two weeks in Reims or Strasbourg?
The number of patients who will be infected with the virus in Alsace in six to eight weeks?
This kind of forecast, PredictEst can and has already achieved.
Without being too wrong.
“On arrival, we had an error rate of 5-8%,” recalls Professor Benoît Gallix for the first case.
The director of the Institut hospitalo-universitaire (IHU) in Strasbourg contributed to the development of this famous modeling tool quite similar to that of the Institut Pasteur but specific to the Grand-Est region.
The region was precisely the driving force behind this project, helped by other institutional, academic and also private partners (Dassault Systemes and CapGemini).
The goal ?
"Getting a little ahead of the epidemic to better adapt our decisions at the local level", explains the president of Grand-Est and doctor by training, Jean Rottner, particularly satisfied with the result.
A huge database
PredictEst is first and foremost a huge database made up of public data, from INSEE or Public Health France.
Up to a very fine scale, like that of the district in Strasbourg.
For analysts, it is thus possible to follow the evolution of the incidence rate almost everywhere in the great region.
And many other data: the rate of reproduction of the virus, its active circulation, hospital pressure, the number of PCR tests carried out, but also the rate of social housing in the area, the non-graduate population, etc.
Figures all anonymized.
The creators assure it: the "highest level of protection" surrounds this information.
PredcitEst also allows you to see the data at a very fine scale.
As here in the districts of Strasbourg where it is possible to superimpose the incidence rate of Covid-19 with other data.
Like the unemployment rate, the number of social housing etc.
Everything does not lend itself to analysis.
On the screen, the modeling tool draws curves, diagrams, tables, maps.
"And thanks to the use of a mass of scientific publications combined with these existing public data, we have the possibility of predicting the evolution of the epidemic according to given scenarios", continues Benoît Gallix.
“In the event of a first confinement, we know, for example, that the number of hospitalizations will decrease in proportions that will be different than in the event of a curfew or a second confinement.
In short, with the conclusions of past episodes, PredictEst foresees the future.
"The machine learns as it goes"
With a limit: he does not know all the reactions of the population.
What about the Christmas holidays?
This scenario has not yet been experienced and therefore cannot generate any prospective.
“The machine learns as it records data.
It is improving every day, ”summarizes Jean Rottner, who already imagines using it for specific cases and above all for prevention.
"As if to help our next vaccine strategy because we know that such and such places have been more affected than others," blows the prefect of the Grand-Est, Josiane Chevalier.
Our file on the coronavirus
PredictEst could also be deployed in other regions.
Baden-Württemberg (Germany) and Luxembourg have already requested it.
But will it be accessible to everyone tomorrow?
“It's not a website that gives forecasts,” replies Benoît Gallix.
It is too early for wide dissemination.
The tool is prospective but must be analyzed by people who are used to it.
It is intended for decision-makers.
"It should really help us to anticipate and better explain our choices to the public," concludes the representative of the State.
Coronavirus: Five professionals call on the government to "act quickly" on mental health
Coronavirus: How messenger RNA vaccination works, the technique used by Pfizer and Moderna