Download Modeling the global spread of infectious diseases in the age of data

Data-driven computational
epidemiology
Modeling the global spread of infectious diseases
in the age of data
Michele Tizzoni
Computational Epidemiology Lab
ISI Foundation
Data and epidemiology
Data and epidemiology
Epidemic forecasts?
Can we use data on
human behavior
to model and predict
the spread of
infectious diseases?
Epidemic forecasts?
GLobal Epidemic And Mobility model
GLobal Epidemic and Mobility model
A matter of fact
People travel
and diseases travel
with them
Epidemics and human mobility
14th century - Black death
Epidemics and human mobility
Increasingly efficient mobility infrastructures:
vehicle of rapid worldwide spreading of infections
2009 H1N1 pandemic
GLEAM: the approach
Mobility Data
+
Mathematical
models
Global
Epidemic and
Mobility Model
Image by B. Gonçalves
GLEAM: the data layers
Population layer
Human mobility layers
Geographic scale
GLEAM: population data
Data source: Gridded Population of the World (GPW)
by NASA Socioeconomic Data and Application Center
http://sedac.ciesin.columbia.edu
• population estimates from worldwide census data
• cell resolution of 15x15 minutes of arc
• 823,680 cells cover the whole planet
• 250,206 cells are populated
GLEAM: population data
GLEAM: commuting data
Data source: multiple national bureaus of statistics
GLEAM: commuting data
Data source: multiple national bureaus of statistics
The worldwide airport network
Data source: IATA
3,362 airports and 16,846 connections
1.4 billion passengers a year
Integrating data into GLEAM
Two steps process:
1. Define a new geographic resolution
2. Map all the data to the newly defined
resolution
Integrating data into GLEAM
1. Define a new geographic resolution
airport
(transportation hub)
census cell
1/4° x 1/4°
census cell
population
geographical
census area
(from
tessellation)
Des Moines
Detroit
Cleveland
Pittsburgh
Chicago
St. Louis
Indianapolis
Cincinnati
Louisville
Nashville
Memphis
Charlotte
Integrating data into GLEAM
2. Map data to the new resolution
GLEAM: commuting data
What if data is missing?
Use modeling and statistics!
GLEAM: commuting data
What if data is missing?
Use modeling and statistics!
Gravity law:
wij =
α β
Ni Nj
C
γ
dij
Create synthetic commuting
networks for all the countries
where we miss data
GLEAM: commuting data
Global commuting networks
GLEAM: the full model
susceptible
latent
+
symptomatic
infectious (no travel)
symptomatic
infectious (travel)
recovered
asymptomatic
infectious
Final step: epidemic layer
Compartments
R0 = disease transmissibility
susceptible
R0 > 1
latent
symptomatic
infectious (no travel)
symptomatic
infectious (travel)
recovered
asymptomatic
infectious
infection
progresses
A case study: 2009 H1N1 pandemic
April 2009
first confirmed cases
in the US
May 2009
pandemic is
officially declared
June 2009
confirmed cases
are reported
in 112 countries
??
A case study: 2009 H1N1 pandemic
April 2009
first confirmed cases
in the US
May 2009
pandemic is
officially declared
Parameter
estimation
June 2009
confirmed cases
are reported
in 112 countries
??
Predictions
New secondary cases
Parameter estimation
Arrival times of
countries seeded by
Mexico
R0 = 1.75
Timing of the epidemic
USA
Canada
30 32 34 36 38 40 42 44 46 48 50 52 2 4 6 8 10
Aug
Sep
Oct
2009
Nov
Dec
Jan
Feb
2010
Timing of the epidemic
Observed peaktimes
USA
Canada
30 32 34 36 38 40 42 44 46 48 50 52 2 4 6 8 10
Aug
Sep
Oct
2009
Nov
Dec
Jan
Feb
2010
Timing of the epidemic
Team
Vittoria Colizza
Alessandro Vespignani
Paolo Bajardi
Duygu Balcan
Corrado Gioannini
Chiara Poletto
Marco Quaggiotto
Michele Roncaglione
Michele Tizzoni
Wouter Van den Broeck
Fabio Ciulla
Marcelo Gomes
Bruno Gonçalves
Ana Pastore y Piontti
Nicola Perra
Luca Rossi
Qian Zhang
GLEAMviz Simulator
GLEAMviz Simulator
Thank you!