URL:
https://hal.archives-ouvertes.fr/hal-02592264v1

Type d’article :
Preprint

Thème : Stratégies de contrôle Autre : modélisation, épidémiologie

Que retenir de cet article, en 1-2 phrases ? :
Focus on post-lockdown dynamics of the epidemic based on safety measures (wearing masks, tracking infections). It uses a discrete probability model that accounts for 5 age groups and several epidemiological categories. The work uses available data in 5 regions of France and age-dependent data.

Objectifs de l’étude / Questions abordées :
Describe the dynamics of the epidemic in France before lockdown to validate a mathematical model. Predict the epidemic dynamics after lockdown lift (after May 11th). Estimate parameter values.

Méthode :
The authors develop a discrete dynamical probabilistic model accounting for 5 different age groups (group 1=0-24, group 2=25-49, group 3=50-59, group 4=60-69, and group 5=70+) and 8 different infection related states , including number of individuals infected, hospitalized, in intensive care units (ICU), dead and recovered from hospitalization. The model accounts for age-dependent interactions between population groups and predicts consequences for 8 various infection categories. The authors use online health care data for the 5 most infected regions in France to calibrate the model. Data are available and everything is well referenced.

Résultats principaux :
The authors compute R0 values for each age-class before lockdown and motivate the use of an age-dependent model. At day of lockdown lift (May 11th), they find that 13% (around 4.8M) of the population is infected in the 5 most affected regions of France, and compute post-lockdown R0 values in agreement with other studies. The model predicts the evolution of the epidemic without restriction measures after the end of the lockdown, and investigate the influence of safety measures post-lockdown: if the reproduction number R0 is reduced by at least a factor 2.5-3 for all age groups after lockdown, which could be achieved by wearing masks and social distancing, a significant second peak can be prevented. However, if the reduction in R0 for the age group 0-25 is less than 2 (due to school openings for instance), a second peak with ICU saturation is unavoidable. The authors argue that testing should be focused on children, but without tracing it will nevertheless have only a very limited impact on reducing the spread. This last part is less clear (methodologically) than the previous ones.

Commentaire/brève évaluation :
Very efficient use of available data. The authors focus on a detailed description of the epidemic dynamics among 5 age groups in 5 regions in France, which is highly relevant. The description of the model is clear, as is the whole manuscript. Addressed questions are of crucial importance for the post-lockdown period.

There are some minor limitations:

  • the model overestimates the number of recovered individuals for the group 70+ starting around day 20 after lockdown, and the authors mention that the “exact reason is unclear, a possible explanation could be that the number of interactions for this group further decreased during lockdown”.
  • In order to describe contacts after lockdown very few data have been used so values used are arbitrary
  • References do not appear in the preprint.