Mathematical models of infectious diseases D. Gurarie Lecture outline • Goals • Methodology • Basic SIR and SEIR – BRN: its meaning and implications – Control strategies: treatment, vaccination/culling, quarantine – Multiple-hosts: zoonotics and vector-born diseases 1. Math modeling: issues, problems Spread of diseases in populations •Biological factors (host-parasite interactions) •Environmental – behavioral factors (‘transmission environment’) Public health assessment (morbidity, mortality) Intervention and control •Drug treatment (symptomatic, prophylactic) •Vaccines •Transmission prevention Modeling Goals Develop mathematical/computer techniques, tools, methodology to i) Predict outcomes ii) Analyze, develop control strategies 2. Early history of math. modeling (XVIII century smallpox) • Known facts: – Short duration (10 days), high mortality (up to 75%) – Life-long immunity for survivors – Possible prevention: inoculation by cow-pox • Q: could life expectancy be increased by preventive inoculation? • Approach: age-structured model of transmission + ‘analysis’ => • Answer: gain of 2.5 years Daniel Bernoulli 1700-1782 3. Transmission patterns 1. Direct: host – to-host (flu, smallpox, STD,…) 2. Vector –borne diseases Macro-parasites: schistosome life cycle schisto malaria This diagram is provided by Center for Disease Control and Prevention (CDC). 3. Epizootic: WNV, Marburg, … 4. Infection patterns: typical flu outbreak Data (British Medical Journal, March 4 1978, p. 587) Influenza Epidemic Infectives 300 250 200 150 100 50 0 Day 0 2 4 6 8 10 12 14 Explain outbreak pattern ? Predict (peak, duration, cumulative incidence) ? Control (drug, vaccine, quarantine) ? 5. SIR –methodology: host ‘disease states’ and history S – Susceptible E – Exposed I – Infectious R - Removed /immune I R E S Latency Infective stage Immune stage … 6. SIR transmission in randomly mixing community • Community of N hosts, meet in random groups of c (or less) = contact rate • Host states and transitions: S I R • Probability of infection/infectious contact = 1-p • Recovery rate = 1-r (=> mean duration of I-state T=1/(1-r)) • Life long immunity Groups Day 1 2 3 … 1 2 3 4 5 6 7 8 9 10 11 12 13 1 2 3 4 5 6 7 8 9 10 11 12 13 1 2 3 4 5 6 7 8 9 10 11 12 13 1 2 3 4 5 6 7 8 9 10 11 12 13 1 2 3 4 5 6 7 8 9 10 11 12 13 1 2 3 4 5 6 7 8 9 10 11 12 13 1 2 3 4 5 6 7 8 9 10 11 12 13 {1},{2,9,13},{4,5,10},{6,7,8} Simulated pattern: infection outbreak Questions •Outbreak duration, peak -? •Cumulative incidence (other health statistics)-? • Dependence on c (contact), p (transmissibility), r (recovery) - ? • Control, prevention ?? •Drug treatment • Vaccine • Quarantine II. SIR methodology: diagrams SI S SIR SEIR S I Birth I R SEIR S Death E I R V S E I Loss R recruitment V Variables: S, E, I, R, V (vaccinated) – host states, or populations /fractions Models: •Continuous (DE) for {S(t),… }- functions of time t •Discrete {S(t),… } (t=0,1,2,…) •Community level (populations) • Individual level (agent based) Discrete SIR: Reed-Frost S+I+R=N (or S+E+I+R=N) - populations, or prevalences: S+E+I+R=1 Parameters: c - contact rate (‘average # contacts’/host/day) p – probability to ‘survive infectious contact’ (1-p = susceptibility) l(p,c) – force of infection q – probability to stay latent => latency duration =1/(1-q) - ?? r –probability to stay infected => infectious period=1/(1-r) s –probability to stay immune => immune duration=1/(1-s) l S 1-s S I 1-r R l E 1-q I 1-r 1-s R Reed-Frost map (discrete time step) “current state” “next state” (S,I,R) (S’,I’,R’) (S,E,I,R) (S’,E’,I’,R’) (S=S(t),… ) (S’= S(t+1),…) Equations: S p c I / N S 1 s R E 1 p c I / N S qE I (1 q) E rI R 1 r I sR S p c I / N S 1 s R c I / N S rI I 1 p R 1 r I sR Parameter Probability Time (duration) p survive infectious contact q stay latent (E-state) 1/(1-q) r stay infected (I-state) 1/(1-r) s stay immune (R-state) 1/(1-s) c contact rate/day c(I/N) number of ‘infectious contacts’ No analytic solution! SIR Smallpox Duration (days) Probability = 1-1/T latent (E-state) 3 q=.66 infected (I-state) 7 r=.86 immune (R-state) All life s=1 10 r=.9 All life s=1 Flu Duration (days) Probability = 1-1/T latent (E-state) 2 q=.5 infected (I-state) 5 r=.8 immune (R-state) 150 s=.993 7 r=.86 150 s=.993 Numeric simulations Smallpox Flu SIR Infect. period Epidemic peak Day Cumulative incidence BRN 10. 0.495439 17 0.984751 5.26803 Infect. p eriod 7. 1.0 1.0 0.8 0.8 0.6 0.6 Ep idemic p eak 0.442523 Day 15 Endemic levels 0.23, 0.03, 0.74 BRN 4.42514 0.4 0.4 0.2 0.2 0 20 40 60 80 Infection period Epidemic peak Day Cumulative incidence BRN 10. 0.192664 34 0.931552 3.16082 SEIR 50 100 150 200 100 Infect. p eriod 7. 1.0 1.0 0.8 0.8 0.6 0.6 Ep idemic p eak 0.188307 Day 27 Endemic levels 0.38, 0.01, 0.02, 0.59 BRN 2.63401 0.4 0.4 0.2 0.2 0 20 40 60 80 100 50 100 150 200 250 300 Analysis of outbreaks and endemic equilibria 3 basic parameters i) Susceptibility :1-p (‘resistance to infection’ = p) ii) Contact rate: c iii) recovery rate: r Questions: 1. How (p,c,r) would determine infection pattern: outbreak, endemic equilibria levels et al? 1. Control intervention -? Key index: BRN 1 c R0 ln 1 r p R0 > 1 – stable endemic infection (flu); outbreak of increased strength (smallpox) R0 < 1 – stable eradication (flu); no outbreaks (smallpox) Control, prevention •Drug treatment r (“prophylactic MDT“-> p) • Vaccine ‘S- fraction’, p • Quarantine c 1. Effect of vaccine 0<f<1 – cover fraction e>1 – efficacy (enhanced resistance): (normal) p(vaccinated) pV = p1/e 1. Perfect vaccine (1/e = 0 – full resistance) Vaccination = Effective reduction of contact rate: c (1-f)c Reduced BRN V 0 R S E f I V c 1 f 1 ln 1 f R0 1(?) 1 r p If R0 is known (?), cover fraction f=1-1/R0 needed to eradicate infection. R 2. Imperfect vaccine (1/e>0) 0<f<1 – cover fraction e>1 - efficacy Effects of vaccine: • reduce risk of infection under identical ‘infected contacts’: p pV = p1/e > p • enhance recovery: r rV = re <r l 1 p lV 1 pV cI 1-f S l E I R f cI Effect (f,e) - ? BRN: R0(f,e) -? Can BRN be brought <1 ? S’ lV E’ I’ III. Continuous (DE) models r r lbI r S I R S lbI dS dS Differential equations b dt dI dt dR dt b I N I N S r R ... S r I 1 d rI r R dt dE dt dI dt dR dt Parameters b b r a /d r a E b I I b N I N r R S ... S a E aE r I 1 d rI ... transmission coefficient = “contact rate” x “prob. infection/contact” recovery rate = 1/”mean duration”) 1/”latency period” Natural/disease mortality immune loss rate 2. Smallpox SIR (immune loss r=0) S b SI ... Solution ... I b SI rI Phase-plane S Time series 1.0 1.0 0.8 0.8 0.6 Cumulative incidence 0.6 I 0.4 0.4 0.2 0.2 0 0.0 0.0 0.2 0.4 0.6 0.8 2 4 6 8 1.0 1/R0 BRN: R0 b r or b r Transmission = Re cov ery (+ loss) 10 3. SIR with immune loss (flu) dS Prevalence DE dt dI dt dR dt b IS r R dI b IS r d I r – recovery d –disease mortality r – immune loss R0 S * 1/ R0 ; b I * 1 1/ R0 rd Jacobian matrix for (S,I) R0 1 R0 1 Eradication: (1,0,0) r / b 0 ... 1 1/ R0 S , I , R I r / b 1 r / b Endemic: * * * * I* Saddle/ unstable Sink /stable Sink /stable Saddle/ unstable r ; rr r R* 1 1/ R0 ; rr rI r R Equilibrium Analysis: Endemic Equilibrium BRN 0 4. BRN: meaning, implications • (SIR with life-long immunity): R0 determines whether outbreak occurs (R0 >1), or infection dies out (R0 <1) R 1 rt • BRN is related to initial infection growth : I t I 0e As e R 1 R0 , R0 approximately measures “# secondary cases/per single infected” over “time range ” r t=1 • BRN (R0 >1) determines infection peak and timing, depending on initial state I0 • For SIR with immune loss sets apart: (i) endemic equilibrium state (R0 >1), or waning of infection (R0 <1) 0 0 5. Control intervention • Vaccination (herd immunity): – vaccinating fraction f of susceptibles decreases R0 (1f)R0. So f>1-1/R0 prevents outbreak – culling of infected animals has the same effect I(1-f)I • Demographics: – increased population density N drives R0 = bN/r up (enhanced outbreaks, higher endemicity) • Transmission prevention: – Lower transmission rate b decreases R0 IV. Vector mediated transmission Viral: RVF, Dengue, Yellow fever, Plasmodia: Malaria, toxoplasma Parasitic worms: schistosomiasis, Filariasis 2. Coupled SIR-SEIR diagrams Host: X Y Vector: w X Z v u Z w R R u 3. Macro-parasites: schistosome life cycle This diagram is provided by Center for Disease Control and Prevention (CDC). 4. Macdonald model: mean intensity-burden (host) + prevalence (vector) Infection intensity (burden) is important for macro-parasites w=mean worm burden of H population; y=prevalence of shedding snail; Premises: •Steady snail population and environment •Homogeneous human population, and transmission patterns (contact /contamination rates, worm establishment ets) b 2 HN BRN: R0 AB => equilibria, analysis and control (??) Summary (math modeling) • Models either ‘physical’ (mice) or ‘virtual’ (math) allow one to recreate ‘reality’ (or part of it) for analysis, prediction, control experiment s • Methodology: • Models need not reproduce a real system (particularly, complex biological ones) in full detail. • The ‘model system’ is made of ‘most essential’ (in our view) components and processes • For multi-component systems we start with diagrams, then produce more detailed description (functions, equations, procedures) • Math models have typically many unknown/uncertain parameters that need to be calibrated (estimated) and validated with real data •Simple math. models can be studied by analytic means (pen and paper) to draw conclusions •Any serious modeling nowadays involves computation.