Estimating the burden of SARS-CoV-2 in France

Epidemiology Transversal
Salje H et al

Main result

In France, the Northeast (Nord-Est) and the Paris region are the most affected regions with more than 1,000 deaths each, while the western regions experienced less than 300 deaths each.
In early April, France reached the plateau of hospitalizations and deaths and started declining in mid-April.
The older age groups remain broadly the most affected, with almost 12,000 men hospitalized and 10,000 women over age 80. This number drops to around 8,000 men and 8,000 women in the 70-80-year-old group, between 6,000 and 7,000 in the 60-70-year-old group, between 6,000 and 5,000 in the 50-60-year-old group and less than 3,000 below.
Regarding mortality, men are 1.5 times more affected on average with for example 4000 deaths (men) versus 3000 (women) in those over age 80. Mortality is exceptional below 50 years. However, regarding hospitalizations in intensive care (ICU), the most represented age group is the 60-70-year-old, with a Gaussian distribution; therefore cases of 30-50-year-old in the ICU are not rare.
Thus, the probability of being hospitalized if you are infected ranges from less than 2% for those under 50 to more than 10% for those aged 70-80, and between 20% (women) and 30% (men) for over age 80. The probability of death for hospitalized patients is between 20% and 30% for the 70-80-year-old and exceeds 40% for men over age 80. Despite everything, it remains below 5% among those under 50.
Modeling enables a prediction of a return to normalcy (situation on March 1) in terms of availability of ICU beds when the containment measures will be lifted, but hospital beds must still count around 2000 COVID cases (compared to nearly 7000 at the peak in early April). Lockdown made it possible to drastically reduce the number of new infections with a diminution from March 17 and undoubtedly less than 1000 infections per day when the containment measures will be lifted.
Regarding collective immunity, it seems that 5.7% (between 2.3 and 6.7%) of the French population will have been infected when the containment measures will be over on May 11 with significant variations depending on the region. Indeed, the most affected regions will have collective immunity higher than 10%, the rest of the eastern regions between 5% and 10%, and probably less than 5% in the western regions.


Using models applied to data on hospitalizations and deaths in France, the authors try to estimate the impact of containment and the current collective immunity.
2.6% of the infected patients are hospitalized and 0.53% die, ranging from 0.001% in the < 20-year-old groups to 8.3% in the > 80-year-old roup. Men at all ages are more likely to be hospitalized, and be transferred to ICU and die than women.
Containment reduced the reproduction rate of the virus (R0) from 3.3 to 0.5 (84% reduction). By May 11, when the containment measures will be lifted, we anticipate that 3.7 million (CI: 2.3-6.7), or 5.7% of the population, will have been infected, with significant variability depending on the French region, the most affected being the most immune (up to almost 15% against less than 5%). Public immunity seems insufficient to avoid a second wave if all of the control measures are lifted by the end of the containment.

Strength of evidence Moderate

- Largest epidemiological modeling study for France
- Reliable administrative data
- Models can ben be criticised but relatively robust
- Not taking into account retired populations living in nursing homes
- No consideration of the impact of chronic diseases or co-morbidities on mortality (based on data from the Diamond Princess)
- Partly based on Chinese contagiousness data


Study the impact of containment in France on the COVID-19 epidemic (hospitalizations, deaths) and on the development of collective immunity


Real-time data obtained by the various French administrations with only biologically confirmed cases, conventional hospitalizations, or those in ICU and deaths by geographic area.
Simple mathematical modeling to estimate the number of hospitalized cases or the number of deaths (negative binomial distribution). Modeling the delay before death for each of the 8 age groups (less than 20 years old, every 10 years, more than 80 years old) via a normal log distribution. The probability of death was dependant on age and gender only. Exclusion of populations living in nursing homes or shelters in France.
In order to dissociate the probability of underlying infection from the probability of hospitalization and death, the authors used the results of an active surveillance campaign in a different population (Data from the Diamond Princess) where almost all the patients were tested, and therefore the probability of detection is independent of serious illnesses requiring hospitalization. For the counting of hospitalizations and deaths (passive hospital surveillance), the authors used a classical Poisson probability.
They used aa deterministic compartmentalized model stratified by age to describe the transmission of SARS-CoV-2 in the French population. During infection, susceptible individuals will enter a latent compartment, in which they will stay for an average of 4 days. During this period, they are not infectious. They will then pass into a second exposed compartment where they will stay on average 1.0 day. Upon entering this compartment, infected individuals become infectious. Then, they move to a compartment where they will stay for an average of 3 days, where all individuals are infectious and a subset will develops symptoms. This parametrization gives an average incubation period of 5 days and allows a day of pre-symptomatic transmission, in line with several estimates from Chinese data.
Complex modeling of the impact of confinement on March 17 in France with adjustment of the contact matrix.

bibliovid.org and its content are bibliovid property.

Legal Notice