1. Computational and Systems Biology
  2. Epidemiology and Global Health

Emerging dynamics from high-resolution spatial numerical epidemics

  1. Olivier Thomine  Is a corresponding author
  2. Samuel Alizon
  3. Corentin Boennec
  4. Marc Barthelemy
  5. Mircea Sofonea
  1. LIS UMR 7020 CNRS, Aix Marseille University, France
  2. MIVEGEC, Université de Montpellier, CNRS, IRD, France
  3. Institut de Physique Théorique, CEA, France
https://doi.org/10.7554/eLife.71417
Share this article
Cite this article
  1. Olivier Thomine
  2. Samuel Alizon
  3. Corentin Boennec
  4. Marc Barthelemy
  5. Mircea Sofonea
(2021)
Emerging dynamics from high-resolution spatial numerical epidemics
eLife 10:e71417.
https://doi.org/10.7554/eLife.71417
  2. Download BibTeX
  3. Download .RIS
5 figures, 3 videos, 3 tables and 1 additional file

Figures

Figure 1
Download asset Open asset
Outline of the Epidemap simulation framework.

(A) The 66 millions inhabitants of metropolitan France are explicitly mapped to housing buildings following cartographic and demographic data. (B) At each time point of the simulation, the number of infected individuals in each building of the country is recorded, as well as the time past since each got infected (the panel shows the Paris area). (C) Every day, individuals can randomly move from their home to other buildings according to a mobility kernel and meet other people. If an infected individual encounters a susceptible host, a transmission event can occur. (D) The contagiousness of a infected individual varies depending on the time since infection. (E) A small fraction θ of infected individuals develop a critical form of the disease that requires their hospitalisation to the nearest facility. Clinical dynamics can be assumed not to affect transmission dynamics because more than 95% of the secondary transmission events occur before hospital admission.

Figure 2 with 1 supplement
Download asset Open asset
Epidemiological dynamics at the national (a) and regional (b) level.

(a) The daily prevalence (the number of infected individuals) is in red, the temporal reproduction number (Rt) in blue, and the cumulative number of recovered individuals in green. Shaded areas show the 95% sample quantiles of 100 stochastic simulations. The dashed line shows the median Rt calculated on the case incidence data. (b) Each colour shows the prevalence in a French region.

Figure 2—figure supplement 1
Download asset Open asset
Mean field epidemiological dynamics.

The transmission model is identical to that in Figure 2 of the main text and the R0 is set to 3. The dynamics of the number of currently infected individuals (in millions in pink) does not exhibit a two-peaks dynamics. See Figure 2 for additional details.

Figure 3
Download asset Open asset
Epidemic arrival date and size at the district level.

(a) Effect of the distance from the epicentre on the number of days until the epidemic begins in a district. The colour indicates the population density in the district (number of inhabitants divided by the district’s surface). (b) District final epidemic size as a function of the characteristic distance between two individuals normalised by the average dispersal distance. The latter is computed as 2⁢E⁢[X]⁢density, where X is the log-normal distribution of daily individual covered distance. Both panels show the value for 35,234 French districts and 100 stochastic simulations.

Figure 4
Download asset Open asset
Individual reproduction number dynamics.

(a) Distribution obtained over the whole population on 4 different days post outbreak. (b) Daily variation of the mean (in blue) and dispersion (in orange) of the distribution of individual reproduction numbers, which is assumed to follow a Negative Binomial distribution (Lloyd-Smith et al., 2005). Shaded areas show the 95% CI.

Figure 5
Download asset Open asset
Infection model flow chart.

Susceptible individuals (yellow figurine), are exposed to viral transmission from contagious individuals (pink figurine). Once infected, a host is more or less contagious depending on the time since contamination according to distribution ζ called the generation time (and usually parameterised using the empiric serial interval). A fraction θa, the value of which depends on the age of the host a, will develop a critical infection and be admitted to a hospital (purple figurine) according to the complication delay distribution η. The complementary fraction 1-θa is assumed to recover with perfect and long-lasting immunity (green figurine). This compartment is also reachable after hospitalisation with the age-dependent probability 1-μa, after the discharge delay distribution υ. The complementary fraction μa eventually dies from COVID-19. See Sofonea et al., 2021 for additional details.

Videos

Video 1
Download asset
Density of residents in ICU throughout an uncontrolled epidemic.

The top panel shows the value at a fine geographical scale (French ‘canton’) and the bottom panel shows the total density at the national level along with the corresponding temporal reproduction number (Rt).

Video 2
Download asset
Density of infected people / 100 k residents for an uncontrolled epidemic.

The top panel shows the value at a fine geographical scale (French ‘canton’) and the bottom panel shows the total density at the national level along with the corresponding temporalreproduction number (R(t)).

Video 3
Download asset
Complete transmission chains colored by chain length of an uncontrolled epidemic.

Tables

Table 1
Lognormal randomly choosen values and INSEE statistic of the distance between residential and distant building.

Original datas (INSEE, 2016) and fitting.

P⁢D⁢F⁢(0.5⁢l)=𝐥𝐨𝐠𝐧𝐨𝐫𝐦⁢(2,0.88)INSEE statistic
lt10 km32.9%33.7%
10 km <X<20 km30.6%30.5%
20 km <X<30 km15.5%16.0%
30 km <X<50 km12.7%12.5%
50 km <X<100 km6.8%5.8%
gt100 km1.5%1.5%
Table 2
Parameters used for the transmission model in the simulations.

LN stands for log-normal, We for Weibull.

NameDescriptionValueReference
NNumber of agents (population size)6.6E7INSEE, 2019
HNumber of buildings visited during the day (+ home)3Schneider et al., 2013
lIndividual dispersal kernel distanceP⁢D⁢F⁢(0.5⁢l)=LN⁢(2,0.88)fit of INSEE, 2016
N1Maximum daily number of agent met (distant)17user-defined
N2Maximum daily number of agent met (at home)5user-defined
ζ⁢(t)Contagiousness t days after infectionP⁢D⁢F⁢(ζ)=We⁢(2.24,5.42)Nishiura et al., 2020
b⁢(t)Transmission probability per contact5%⁢ζ⁢(t)Variant Technical group, 2021
Table 3
Parameters used for the clinical progression of the infections in the simulations.

These parameters do not affect virus spread. LN stands for log-normal, We for Weibull.

NameDescriptionValueReference
µHospital mortality rate56%Santé Publique France, 2020
η⁢(t)ICU admission daily probability for a severely infected patientP⁢D⁢F⁢(η)=We⁢(1.77,6.52)Linton et al., 2020
υ⁢(t)ICU departure daily probabilityP⁢D⁢F⁢(υ)=We⁢(2,10)Linton et al., 2020
θ⁢(a)Probability for a severely infected patient of age a to dieIFR⁢(a)/μVerity et al., 2020

Additional files

Transparent reporting form
https://cdn.elifesciences.org/articles/71417/elife-71417-transrepform-v2.pdf

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Olivier Thomine
  2. Samuel Alizon
  3. Corentin Boennec
  4. Marc Barthelemy
  5. Mircea Sofonea
(2021)
Emerging dynamics from high-resolution spatial numerical epidemics
eLife 10:e71417.
https://doi.org/10.7554/eLife.71417