LA-UR-13-26511 Agent-based Modeling Approaches for Simulating Infectious Diseases
Sara Del Valle, PhD Los Alamos National Laboratory Team: Sue Mniszewski, Mac Hyman, Reid Priedhorsky, Geoffrey Fairchild
SAMSI Program on Computational Methods in Social Sciences
August 18, 2013
UNCLASSIFIED
Operated by Los Alamos National Security, LLC for NNSA Motivation
“Previously, scientists had two pillars of understanding: theory and experiment. Now there is a third pillar: simulation”
U.S. Secretary of Energy Steven Chu”
Operated by Los Alamos National Security, LLC for NNSA Agent-based Model (ABM)
A computational approach for simulating the actions of agents that interact within an environment.
Operated by Los Alamos National Security, LLC for NNSA Agent-based Model (ABM)
A computational approach for simulating the actions of agents that interact within an environment.
Operated by Los Alamos National Security, LLC for NNSA ABM History
The concept was originally discovered in the 1940s by Stanislaw Ulam and John von Neumann while working at LANL
Von Neumann Neighborhood ABM History
Due to the computational requirements, ABMs did not become widespread until the 1990s
SimCity Why ABM?
System is too complex to be captured by analytical expressions or experiments
Experiments are too expensive or undesirable
Operated by Los Alamos National Security, LLC for NNSA ABM Characteristics
Heterogeneity
Individual Agent Behavior Specification
Explicit Space
Complex Interactions
Randomness
Emergent Phenomena
Operated by Los Alamos National Security, LLC for NNSA Ecological Systems Ecological Systems
Understand the causes of this behavior
Fish, ants, humans, locus, birds, honeybees, immune system, bacterial infection, tumor proliferation
Locus plagues can contain up to 109 individuals Ecological Systems Buhl et al
Experiments
Vicsek et al.
Vicsek et al. Theory
Simulation
Operated by Los Alamos National Security, LLC for NNSA Transportation Systems Transportation Systems Leonardi S.
Experiments
Theory Wilco Burghout
Static
Dynamic
7:00 10:00 15:00 18:00 Transportation Systems - TRANSIMS Epidemiological Studies 15,000,000 People die every year from infectious diseases Global Disease Burden
UNCLASSIFIED
Operated by Los Alamos National Security, LLC for NNSA MALARIAUNCLASSIFIED
Operated by Los Alamos National Security, LLC for NNSA Source: http://en.wikipedia.org/wiki/Image:Anopheles_stephensi.jpeg {{PD-USGov-HHS-CDC}} DIARRHEALUNCLASSIFIED DISEASES
Operated by Los Alamos National Security, LLC for NNSA RESPIRATORYUNCLASSIFIED INFECTIONS
Operated by Los Alamos National Security, LLC for NNSA RESPIRATORY INFECTIONS Epidemic Modeling
Experiments
Theory
S1 I1 R1
Susceptible Infectious Recovered Agent-based Model (ABM)
Empirical Studies • National Household Transportation Survey • Workgroup size by industry classification • Student-teacher ratio • U.S. Census Data
Theoretical Studies • Gravity algorithms • Monte Carlo methods
Operated by Los Alamos National Security, LLC for NNSA The Demographics – Surveys
Source: C.L. Barrett, S.G. Eubank, J.P. Smith (2005). If Smallpox Strikes Portland… Scientific American 292:54-61.
Stroud P et al. Spatial Dynamics of Pandemic Influenza in a Massive Artificial Society. Journal of Artificial Societies and Social Simulation 2007; 10(4):9. The Locations
Home Home Home
Work Home
School Social
Home Shopping Passenger
College Home Home Home
Stroud P et al. Spatial Dynamics of Pandemic Influenza in a Massive Artificial Society. Journal of Artificial Societies and Social Simulation 2007; 10(4):9. A Typical Family’s Day – Surveys
Work Work Lunch Carpool Carpool
Shopping Home Home
Daycare Car Car
Bus School Bus
time
Stroud P et al. Spatial Dynamics of Pandemic Influenza in a Massive Artificial Society. Journal of Artificial Societies and Social Simulation 2007; 10(4):9. Others Use the Same Locations
Stroud P et al. Spatial Dynamics of Pandemic Influenza in a Massive Artificial Society. Journal of Artificial Societies and Social Simulation 2007; 10(4):9. Bipartite Graph – Theory
People Locations
Operated by Los Alamos National Security, LLC for NNSA Bipartite Graph – Theory
Operated by Los Alamos National Security, LLC for NNSA Explicitly Time Dependent – Theory
[t1, t2] [t2 , t3] [t86399, t86400]
Operated by Los Alamos National Security, LLC for NNSA People-people Graph – Theory
[t1, t2] [t2 , t3] [t86399, t86400]
[t1,t2] [t2,t3] [t86399,t86400]
Operated by Los Alamos National Security, LLC for NNSA Social Contact Network Emerges
Degree
1 – 6 7 – 12 13 – 18 19 – 24 25 – 30 31 – 36 37+ Degree Distribution
5 10
4 10
3
contacts 10 k
2 10
1 10
0 10 Number of people with
ï1 10 10 20 30 40 50 60 70 80 90 100 Number of contacts, k Del Valle et al. Estimating Mixing Patterns for the United States for Modeling Infectious Diseases (in preparation) Demographic and Activity Differentiation FLORIDA NEW YORK UTAH WYOMING FL ï HOME NY ï HOME UT ï HOME WY ï HOME 70 70 70 70
60 60 60 60 50 50 50 50 HOME 40 40 40 40
30 30 30 30
20 20 20 20
10 10 10 10
10 20 30 40 50 60 70 10 20 30 40 50 60 70 10 20 30 40 50 60 70 10 20 30 40 50 60 70
FL ï WORK NY ï WORK UT ï WORK WY ï WORK 80 80 80 80
70 70 70 70 60 60 60 60 WORK 50 50 50 50
40 40 40 40
30 30 30 30
20 20 20 20 20 40 60 80 20 40 60 80 20 40 60 80 20 40 60 80
FL ï SCHOOL NY ï SCHOOL UT ï SCHOOL WY ï SCHOOL 20 20 20 20 18 18 18 18 16 16 16 16 14 14 14 14 12 12 12 12 10 10 10 10 SCHOOL 8 8 8 8 6 6 6 6 4 4 4 4 2 2 2 2 5 10 15 20 5 10 15 20 5 10 15 20 5 10 15 20
FL ï OTHER NY ï OTHER UT ï OTHER WY ï OTHER 90 90 90 90 80 80 80 80 70 70 70 70 60 60 60 60 50 50 50 50 OTHER 40 40 40 40 30 30 30 30 20 20 20 20 10 10 10 10 Del Valle et al.20 Estimating40 60 80 Mixing Patterns20 40 for 60the 80United States20 for 40Modeling60 80 Infectious 20Diseases40 60 (in80 preparation) Population Age Distribution
20% Florida FloridaNew York Utah 18% New YorkWyoming U.S.
16% Utah Wyoming 14% U.S.
12%
10%
8%
6%
4% Percentage Distribution of the Total Population
2%
0% 0 ï 4 5 ï 14 15 ï 24 25 ï 34 35 ï 44 45 ï 54 55 ï 64 65 ï 74 75 ï 84 85+ Age Group Del Valle et al. Estimating Mixing Patterns for the United States for Modeling Infectious Diseases (in preparation) Comparison with Empirical Data
30 MOSSONG SIMULATEDPOLYMOD SIMULATED 25
20
15
Number of Contacts 10
5
0 0ï4 5ï9 10ï14 15ï19 20ï29 30ï39 40ï49 50ï59 60ï69 70+ Age Groups
Mossong et al. Social Contacts and Mixing Patterns Relevant to the Spread of Infectious Diseases. PLoS Med 2008. Disease Model in EpiSimS
Susceptible Asymptomatic Recovered Incubating, Incubating, Symptomatic- not infectious circulating infectious
1 Half-day Histogram 0.9 Longini Symptomatic- 0.8 0.7 stay home 0.6 0.5 0.4 Convalescence 0.3
Fraction still incubating 0.2 0.1 0 Dead 0 1 2 3 4 5 Hospitalized Incubation Stage Duration (days) Incubation Symptomatic
Stroud P, Del Valle S, Sydoriak S, Riese J, Mniszewski S. Spatial Dynamics of Pandemic Influenza in a Massive Artificial Society. (2007) Journal of Artificial Societies and Social Simulation, 10(9). Probability of Infection
−∑TS j I i tij i Pj (t) =1− e All infected people Susceptible co-located with j Person j
Stroud P, Del Valle S, Sydoriak S, Riese J, Mniszewski S. Spatial Dynamics of Pandemic Influenza in a Massive Artificial Society. (2007) Journal of Artificial Societies and Social € Simulation, 10(9).
Operated by Los Alamos National Security, LLC for NNSA Attack rate red: >35% amber: 29% to 35% grey: 17% to 29% green: <17%
Stroud P et al. Spatial Dynamics of Pandemic Influenza in a Massive Artificial Society. Journal of Artificial Societies and Social Simulation 2007; 10(4):9. Attack Rate Correlates with Demographics
Average Household Size Per Capita Income
Population Aged 5 – 18 Population Density Heterogeneous Spread Demographic Differentiation
Operated by Los Alamos National Security, LLC for NNSA Impact of School Closures
Relaxing the closures can generate second wave of infection
Operated by Los Alamos National Security, LLC for NNSA Pharmaceutical Interventions
Vaccine effectiveness and production rates
Operated by Los Alamos National Security, LLC for NNSA Social Media & Behavior Demographic factors Age Gender Ethnicity Educational level
“Who” is likely to change behavior and “How” is implemented
Operated by Los Alamos National Security, LLC for NNSA Facemask Study Summary
ABMs provide a test-bed for: • Testing hypothesis • Forecasting population demographics, infectious disease transmission, etc. • Developing complex social networks
Operated by Los Alamos National Security, LLC for NNSA Challenges and Opportunities
Statistical Inference
Model Fitting, Validation, and Verification
Uncertainty Quantification for Agent-based Simulations
Prediction & Forecasting
Emulators
Emergent Behavior
Social Media
Operated by Los Alamos National Security, LLC for NNSA Thank You! J Epidemiol Community Health. 2006 January; 60(1): 6.– 7.
Operated by Los Alamos National Security, LLC for NNSA