Download Disease Dynamics in a Dynamic Social Network

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Kawasaki disease wikipedia , lookup

Behçet's disease wikipedia , lookup

Hospital-acquired infection wikipedia , lookup

Herd immunity wikipedia , lookup

Transmission (medicine) wikipedia , lookup

Infection wikipedia , lookup

Childhood immunizations in the United States wikipedia , lookup

Eradication of infectious diseases wikipedia , lookup

Schistosomiasis wikipedia , lookup

Infection control wikipedia , lookup

Sociality and disease transmission wikipedia , lookup

Germ theory of disease wikipedia , lookup

Globalization and disease wikipedia , lookup

Transcript
Disease Dynamics in a Dynamic
Social Network
Claire Christensen1, István Albert3,
Bryan Grenfell2, and Réka Albert1,2
1Department
of Physics and
for Infectious Disease Dynamics,
3The Huck Institutes for the Life Sciences
The Pennsylvania State University, University Park, PA 16802, USA
2Center
Social Structure and Disease Dynamics
What is it about the demographics and connectivity of an underlying contact
network that creates the most prominent features of the landscape in which a
disease travels?
Below threshold:
•chaotic epidemics
•intervening fadeouts
Above threshold:
• regular epidemics
•few fadeouts
The Classic
Example:
Population Size
Grenfell, B.T., et. al, Ecological Monographs 72 (2002): 185-202
How do changes to that landscape affect the dynamic behavior of the
disease?
Classical Mathematical Modeling of Disease
Dynamics
Basic model assumes full-mixing of susceptibles (S), infecteds (I), recovereds (R),
such that rate of change of each group can be described by system of coupled
differential equations:
Contacttransmission
parameter
dS
I
  S
dt
N
dI
I
 S
 I
dt
N
Recovery rate
dR
 I
dt
Model works well in large populations.
Variants of model account for some features
of small populations, but not all.
In-silico Simulation of Contact Networks and
Disease Propagation: an Overview
Why:
-Probe link between (changing) topology of contact network and
disease dynamics;
-Incorporate statistical (social) data;
-Small size not a problem
How (first steps):
-Simulate social network (families+workplaces+schools+individuals);
-Statistical growth and change;
-Disease propagates (WAIFW) along edges of network
Dominant Social Processes/Features
(For more information, please visit the poster exhibit)
Family Graphs:
•~54% adults married;
•Fully-connected;
•Average size is 3-4
Work Graphs:
•Ages 18-65;
•8% unemployment;
•Number=1% net. size;
•Hub-and-spoke;
random attachment to
min. 3 others
thereafter
Basic
Demographics/Social
Processes:
•Age distribution;
•Fertility distribution;
•Death rate;
•Population growth;
•Immigration
School Graphs:
•Ages 6-18;
•Classes fullyconnected; max.
size=40;
•“Elementary” kids
randomly
interconnected to other
classes with between .25
and .5 times class size
interlinks (.25; .75 for
High School);
•Term forcing scheme
**All statistics adapted from:
1)Vital Statistic of New York State (1990-2002), in ‘Information for a Healthy New York’, Health, N.Y.S.D. o,Ed.
2)‘U.S. Department of Labor Bureau of Labor Statistics’’
3) Scardamalia, R., The Face of New York– the Numbers (2001)
Dominant Disease Processes/Features
(For more information, please visit the poster exhibit)
Immunity and Infection
•Initial immunity-by-age
profile;
•Newborns lose maternallyacquired immunity;
•Disease spreads from infected
to susceptible via edges, and
with rates from WAIFW matrix
** All statistics adapted from:
Edmunds, W.J. et al., Epidemiology and Infection
125 (2000): 635-650;
Scardamalia, R., 'The Face of New York-- The
Numbers' (2001);
Grenfell, B.T. and Anderson, R.M., Proceedings of
the Royal Society of London.
Series B,
Biological Sciences 236 (1989): 213-252.
Maintaining Epidemics
•Sometimes necessary to “spark”
population (S →I randomly);
•Sparking function of form:
Pspark(t) ~ln(N(t)), where N(t) is total
population size
Susceptible: Age=6
5,6
Infected: Age=5
Prevaccine Measles
Comparison of infection profiles, interepidemic periods, and epidemic lengths
Data Set
<I(t)>
<Tinter >
<τ>
Data Set
<I(t)>
<Tinter >
<τ>
Teignmouth
~81
1-4 years
38
weeks
Blackburn
~500
1-2 years
41
weeks
10,700
Network
75±5
2.13±1.15
years
40±2
weeks
107,000
Network
450±10
1.47±.6
years
40±2
weeks
Grenfell, B.T., et. al, Ecological Monographs 72 (2002): 185-202
Basic Reproduction Number (R0), Force of Infection, and the
Influence of Contact Network Topology and Dynamics
Force of Infection by Age Cohort
R0 vs Population Size
0.5
14
Force of Infection
12
R0
10
8
6
4
2
0.4
0.3
Population=107,000
0.2
Population=10,700
0.1
0
0
4
16
32
64
107
0
Population Size (Thousands)
Past explanation:
Topology of schools– contact network of most
influential age cohort– is essentially the same,
regardless of population size.
However...
3
6
9
12
15
Age (years)
Force of Infection
n
i 

j 1
i, j
Ij
N
School Network Topology is NOT Independent of
Population Size
Class interlinking and abundance of full classes differ with population size.
Cumulative School Degree Distribution (Max. School Size=40)
1.2
1
P(k>K)
Why?
Pop= 105
0.8
0.6
0.4
Pop=104
0.2
0
20
30
40
50
60
70
80
K
Different dynamic processes; same global
characteristics!
Summary
•
In-silico simulation of contact networks and disease propagation has benefits of:
A) Greater realism;
B) Overcoming size/connectivity barriers;
C) Window into interrelationship between (changing)
network topology and disease dynamics at multiple scales of
complexity and time
•
Using measles as a test bed, we have:
1) Captured dominant features of observed measles dynamics
in large and small populations;
2) Provided some insight into epidemiological trends by clearly
synthesizing topological and disease dynamic features.
•
We will extend the algorithm to other, less well-studied diseases and/or wellstudied diseases in novel social environments.
Thank You
Thank you to:
Dr. Réka Albert
Dr. Bryan Grenfell
Dr. István Albert
Dr. Anshuman Gupta
Dr. Jeff Nucciarone
NIH MIDAS Cluster (Dr. Ganapathi Laxminarayana)
The organizers and sponsors of NetSci 2007
Extras
Total Degree Distributions
Cumulative Total Degree Distribution
1
P(k>K)
0.8
0.6
0.4
0.2
0
0
10
20
30
40
50
60
70
80
K
Cumulative degree distribution for populations of size 105 (blue)
and 104 (pink).
School Degree Distributions (non-cumulative)
Non-cumulative School Degree Distribution (Population=10,700)
120
100
60
40
20
0
0
10
20
30
40
50
60
70
N(k)
Non-cumulative School Degree Distribution (Population=107,000)
N(k)
k
80
9000
8000
7000
6000
5000
4000
3000
2000
1000
0
0
10
20
30
40
50
k
60
70
80
90
Seasonal Network Dynamics and Disease Dynamics
Recent work suggests:
Early-peaking epidemic at
yeart
Seasonally-curtailed
epidemic at yeart+n
No epidemic at yeart+n
Stone, L. et al., Nature 446 (2007): 532-536
Infection Profile (Population ~107,000)
Infected Individuals
500
400
End of school
term
Late January
300
Late June
200
100
0
125
130
135
140
145
150
Time (Months)
Evidence of this in our networks, as well.
155