Ten at the table? Calculate the risk of a guest having coronavirus thanks to this site

In Washington, it’s 18%. In Walsh County, North Dakota, it’s over 99%. In Prague it is 58%. The math is simple, based on real-time data from the Covid-19 pandemic, but presented legibly by a site that has become increasingly popular since it was published in July, created by researchers at Georgia Tech University and whose methodology it was validated on Monday with a publication in one of the prestigious Nature journals.

The address is https://covid19risk.biosci.gatech.edu/ (in English), but on Thursday it fell victim to its success, the site sometimes posting error messages in the face of an influx of connections.

Researchers calculate risk based on the official number of new cases reported each day in a given location (by county in the United States or by department in France).

The model also takes into account that the actual number of contaminations is five to 10 times greater than the number of positive tests and the user can calculate the risk based on these two assumptions, five or 10. In the United States, the director of the Centers for Disease Control (CDC) said this summer that tests have likely only detected one in 10.

The numbers at the beginning of this article are based on a 10-fold underestimate. If we estimate that there are only five times more infections than the official number, the risk of having a positive person in 10 of a Parisian dinner drops to 18%, 10% in Washington and 94% in Walsh, North Dakota.

Then you can select the size of the event you plan to attend: 10, 25, 50, 100 or up to 5,000 people. But going up to 5,000 makes no sense. Currently, in many places in the United States and Europe, the probability of being in the presence of an infected person of 50 people is over 50%.

With 50 people – at a wedding or in crowded bars, if they were open – the risk of at least one being contaminated would be 86% in Paris and 99% in the Rhone.

The site takes a conservative approach – it assumes a person stays positive for 10 days, its creator Joshua Weitz told AFP. In fact, researchers estimate that a person is highly contagious for less time, on the order of five to six days, and that the rest of the time is or is not, despite the residual presence of the virus.

The model also fails to take into account that an infected person is more likely to stay home after symptoms appear. But it notes that half of the infections come from people who have no or few symptoms or are unaware that they are infected, according to studies.