Tutorials 1: Epidemiological Mathematical Modeling Applications in Homeland Security. Mathematical Modeling of Infectious Diseases: Dynamics and Control (15 Aug - 9 Oct 2005) Jointly organized by Institute for Mathematical Sciences, National University of Singapore and Regional Emerging Diseases Intervention (REDI) Centre, Singapore http://www.ims.nus.edu.sg/Programs/infectiousdiseases/index.htm Singapore, 08-23-2005 Carlos Castillo-Chavez Joaquin Bustoz Jr. Professor Arizona State University ASU/SUMS/MTBI/SFI Bioterrorism The possibility of bioterrorist acts stresses the need for the development of theoretical and practical mathematical frameworks to systemically test our efforts to anticipate, prevent and respond to acts of destabilization in a global community ASU/SUMS/MTBI/SFI From defense threat reduction agency Buildings Ports & Airports Response Attribution Urban Treatment and Consequence Management Detection Interdiction Warning ASU/SUMS/MTBI/SFI From defense reduction agency From defensethreat threat reduction agency Food Safety Medical Surveillance Animal/Plant Health Other Public Health Choke Points Urban Monitoring Characterization Metros Data Mining, Fusion, and Management Emergency Management Tools Toxic Industrials ASU/SUMS/MTBI/SFI State and Local Governments Ricardo Oliva: Research Areas •Biosurveillance; •Agroterrorism; •Bioterror response logistics; •Deliberate release of biological agents; •Impact assessment at all levels; •Causes: spread of fanatic behaviors. ASU/SUMS/MTBI/SFI Modeling Challenges &Mathematical Approaches From a “classical” perspective to a global scale Deterministic Stochastic Computational Agent Based Models ASU/SUMS/MTBI/SFI Some theoretical/modeling challenges •Individual and Agent Based Models--what can they do? •Mean Field or Deterministic Approaches--how do we average? •Space? Physical or sociological? •Classical approaches (PDEs, meta-population models) or network/graph theoretic approaches •Large scale simulations--how much detail? ASU/SUMS/MTBI/SFI Ecological/Epidemiological view point Invasion Persistence Co-existence Evolution Co-evolution Control ASU/SUMS/MTBI/SFI Epidemiological/Control Units Cell Individuals Houses/Farms Generalized households Communities Cities/countries ASU/SUMS/MTBI/SFI Temporal Scales Single outbreaks Long-term dynamics Evolutionary behavior ASU/SUMS/MTBI/SFI Social Complexity Spatial distribution Population structure Social Dynamics Population Mobility Demography--Immigration Social hierarchies Economic systems/structures ASU/SUMS/MTBI/SFI Links/Topology/Networks Local transportation network Global transportation network Migration Topology (social and physical) Geography--borders. ASU/SUMS/MTBI/SFI Control/Economics/Logistics Vaccination/Education Alternative public health approaches Cost, cost & cost Public health infrastructure Response time ASU/SUMS/MTBI/SFI Critical Response Time in FMD epidemics A. L. Rivas, S. Tennenbaum, C. Castillo-Chávez et al. {American Journal of Veterinary Research} (Canadian Journal of Veterinary Research) It is critical to determine the time needed and available to implement a successful intervention. The context--Foot and Mouth Disease A R G E N T .I N A BRAZIL : 1-5 cases (1- 7 days post-onset) 1-5 cases 1 2 3 (8-14 days post-onset) ATLANTIC OCEAN (May 1, 01) Day 30 (May 22, 01) Day 29 (May 21, 01) Day 28 (May 20, 01) Day 27 (May 19, 01) Day 26 (May 18, 01) Day 25 (May 17, 01) Day 24 (May 16, 01) Day 23 (May 15, 01) Day 22 (May 14, 01) Day 21 (May 13, 01) Day 20 (May 12, 01) Day 19 (May 11, 01) Day 18 (May 10, 01) Day 17 (May 9, 01) Day 16 (May 8, 01) Day 15 (May 7, 01) Day 14 (May 6, 01) Region 1 Day 13 (May 5, 01) 40 Day 12 (May 4, 01) Day 11 (May 3, 01) Day 10 (May 2, 01) Day 9 Day 8 (April 30, 01) Day 7 (April 29, 01) Day 6 (April 28, 01) 35 Day 5 (April 27, 01) Day 4 (April 26, 01) Day 3 (April 25, 01) Day 2 (April 24, 01) Day 1 (April 23, 01) Number of daily cases Daily cases in the first month of the epidemic “exponential”growth Region 2 30 Region 3 25 20 15 10 5 0 The Basic Reproductive Number R0 R0 is the average number of secondary cases generated by an infectious unit when it is introduced into a susceptible population (at demographic steady state) of the same units. If R0 >1 then an epidemic is expected to occur--number of infected units increases If R0 < 1 then the number of secondary infections is not enough to sustain an apidemic. The goal of public health interventions is to reduce R0 to a number below 1. However, timing is an issue! How fast do we need to respond? ASU/SUMS/MTBI/SFI Estimated CRTs for implementing intervention(s) resulting in R_o <= 1 (successful intervention) 1.4 days 2.6 days 3.0 days Epidemic Models ASU/SUMS/MTBI/SFI Basic Epidemiological Models: SIR Susceptible - Infected - Recovered ASU/SUMS/MTBI/SFI contacts probability of transmission time contact I B(S,I) S N S(t): susceptible at time t I(t): infected assumed infectious at time t R(t): recovered, permanently immune N: Total population size (S+I+R) N S I I S I R R ASU/SUMS/MTBI/SFI SIR - Equations dS I N S S dt N dI I S I dt N dR I R dt Parameters (1) (2) (3) Per-capita death (or birth) rate Transmission coefficient Per-capita recovery rate N SIR contacts probability of transmission (4) dN d S I R 0 dt dt (5) unit time ASU/SUMS/MTBI/SFI contact SIR - Model (Invasion) dS I N S S dt N dI I S I dt N SN dI I I I dt or I(t) I(t) I(0)e t R0 ASU/SUMS/MTBI/SFI 1 Ro “Number of secondary infections generated by a “typical” infectious individual in a population of mostly susceptibles at a demographic steady state Ro<1 No epidemic Ro>1 Epidemic ASU/SUMS/MTBI/SFI Establishment of a Critical Mass of Infectives! Ro >1 implies growth while Ro<1 extinction. ASU/SUMS/MTBI/SFI Phase Portraits ASU/SUMS/MTBI/SFI SIR Transcritical Bifurcation I* I * (R0 ) R0 unstable ASU/SUMS/MTBI/SFI Deliberate Release of Biological Agents ASU/SUMS/MTBI/SFI Effects of Behavioral Changes in a Smallpox Attack Model Impact of behavioral changes on response logistics and public policy (appeared in Mathematical Biosciences, 05) Sara Del Valle1,2 Herbert Hethcote2, Carlos Castillo-Chavez1,3, Mac Hyman1 1Los Alamos National Laboratory 2University of Iowa 3Cornell University ASU/SUMS/MTBI/SFI MODEL •All individuals are susceptible •The population is divided into two groups: normally active and less active •No vital dynamics included (single outbreak) •Disease progression: Exposed (latent) and Infectious •News of a smallpox outbreak leads to the implementation of the following interventions: –Quarantine –Isolation –Vaccination (ring and mass vaccination) –Behavioral changes (3 levels: high, medium & low) ASU/SUMS/MTBI/SFI The Model Sn V En In Q S Sl R W E El I Il D The subscript refers to normally active (n) or less active (l): Susceptibles (S), Exposed (E), Infectious (I), Vaccinated (V), Quarantined (Q), Isolated (W), ASU/SUMS/MTBI/SFI Recovered (R), Dead (D) The Model The behavioral change rates are modeled by a non-negative, bounded, monotone increasing function i (for i = S, E, I) given by i ai (In I ) 1 1 bi (In I ) day with ASU/SUMS/MTBI/SFI S E I Numerical Simulations ASU/SUMS/MTBI/SFI Numerical Simulations ASU/SUMS/MTBI/SFI Conclusions •Behavioral changes play a key role. • Integrated control policies are most effective: behavioral changes and vaccination have a huge impact. •Delays are bad. ASU/SUMS/MTBI/SFI Mass Transportation and Epidemics "An Epidemic Model with Virtual Mass Transportation" ASU/SUMS/MTBI/SFI Mass Transportation Systems/HUBS Baojun Song Juan Zhang Carlos Castillo-Chavez ASU/SUMS/MTBI/SFI Subway Transportation Model NSU NSU SU SU Subway SU SU NSU NSU ASU/SUMS/MTBI/SFI Vaccination Strategies • Vaccinate civilian health-care and public health workers • Ring vaccination (Trace vaccination) • Mass vaccination • Mass vaccination if ring vaccination fails •Integrated approaches likely to be most effective Assumptions 1.The population is divided into N neighborhoods; 2.Epidemiologically each individual is in one of four status: susceptible, exposed, infectious, and recovered; 3.A person is either a subway user or not 4.A ``vaccinated” class is included-everybody who is successfully vaccinated is sent to the recovered class Proportionate mixing K subpopulations with densities N1(t), N2(t), …, Nk(t) at time t. cl : the average number of contacts per individual, per unit time among members of the lth subgroup. Pij : the probability that an i-group individual has a contact with a j-group individual given that it had a contact with somebody. Proportionate mixing (Mixing Axioms) k>0 Pij P 1 ij j1 (1) (2) (3) ci Ni Pij = cj Ncj PN ji Pij Pj Then j j K cl N l l 1 is the only separable solution satisfying (1) , (2), and (3). Definitions Pai ai the mixing probability between non-subway users from neighborhood i given that they mixed. Pa b the mixing probability of non-subway and subway users from neighborhood i, given that they mixed. Pb a the mixing probability of subway and non-subway users from neighborhood i, given that they mixed. Pb b the mixing probability between subway users from neighborhood i, given that they mixed. Pb b the mixing probability between subway users from neighborhoods i and j, given that they mixed. Paia j the mixing probability between non-subway users from neighborhoods i and j, given that they mixed. Pa b the mixing probability between non-subway user from neighborhood i and subway users from neighborhood j, given that they mixed. i i i i i i i j i j Formulae of Mixing Probabilities (depends on activity level and allocated time) Identities of Mixing Probabilities State Variables i index of neighborhood Wi number of individuals of susceptibles of SU in neighborhood i Xi number of individuals of exposed of SU in neighborhood i Yi number of individuals of infectious of SU in neighborhood i Zi number of individuals of recovered of SU in neighborhood i Si number of individuals of susceptibles of NSU in neighborhood i Ei number of individuals of exposed of NSU in neighborhood i Ii number of individuals of infectious of NSU in neighborhood i Ri number of individuals of recovered of NSU in neighborhood i Smallpox Model for NSU in neighborhood i Ai Si Bi (t ) Ei S i Ei Ei Ii ( d ) I i ql 2 Ei I i ql1 S i Ri Ri Model Equations for neighborhood i Nonsubway users dS i Ai Bi (t ) ( S i ql1 S i ) dt dEi Bi (t ) ( Ei Ei ql 2 Ei ) dt dI i Ei ( d ) I i dt dRi I i Ri ql1 S ql 2 E dt Qi (t ) S i (t ) Ei (t ) I i (t ) Ri (t ) Subway users dWi i Vi (t ) ( Wi ql1Wi ) dt dX i Vi (t ) ( X i X i ql 2 X i ) dt dYi X i ( d )Yi dt dZ i Yi Z i ql1Wi ql 2 X i dt Ti (t ) Wi (t ) X i (t ) Yi (t ) Z i (t ) Infection Rates Rate of infection for NSU Ii Bi (t) i ai Si P˜a i P˜bi i Qi Ti i i i Yi i i i Ti Qi i i Rate of infection for SU Vi (t) ibiW i Pa i i j Y Yi j N Ii i i j j Pbi Pb j i i j1 j Ti Ti T j Qi Qi i i i i j j R0 for Two Neighborhoods (a special case) q 0, i 0, i 1 R0 max{ R 0,1 , R0, 2 } 1 a i (Ai / ) Ai / R0, i i a i d (a i Ai bi i ) / (Ai i ) / 1 bi ( i / ) i / i b i d (a i Ai bi i ) / (Ai i ) / Two neighborhood simulations (NYC type city) 1. There are 8 million long-term and 0.2 million shortterm (tourists) residents in NYC. 2. Time span of simulation is 30 days +. 3. Control parameters in the model are: q1 and q2 (vaccination rates) 4. We use two ``neighborhoods”, one for NYC residents and the second for tourists. Curve R0 (q1, q2) =1 Plot R0 (q1, q2) vs q1 and q2 Cumulative deaths: One day delay (q1 = q2=0.5) Cases: One day delay (q1 = q2=0.5) Cumulative deaths: One day delay (q1 = q2=0.8) Cases: One day delay (q1 = q2=0.5) Conclusions •Integrated control policies are most effective: behavioral changes and vaccination have a huge impact. •Delays are bad. ASU/SUMS/MTBI/SFI