lecture 5: deterministic compartmental epidemiological
TRANSCRIPT
Modelling EpidemicsLecture 5: Deterministic compartmental
epidemiological models in homogeneous populations
Andrea Doeschl-Wilson
Overview
β’ Key characteristics of epidemicsβ’ What is an epidemic and what do we need to know about them?
β’ The basic reproductive number R0
β’ Modelling epidemics: basic compartmental modelsβ’ Deterministic model formulation
β’ SIR model without demography
β’ SIR model with demography
β’ Adding complexityβ’ Loss of immunity
β’ Inclusion of chronic carriers
What is an epidemic?
Definition: Epidemic
A widespread occurrence of an infectious diseasein a community at a particular time
Questions for modelling epidemics
β’ What is the risk of an epidemic to occur?
β’ How severe is the epidemic? β’ What proportion of the population will become infected?
β’ What proportion will die?
β’ How long will it last?
β’ Are all individuals at risk of becoming infected?
β’ How far will it spread?
β’ What impact does a particular intervention have on the risk, severity and duration of the epidemic?
The basic reproductive ratio R0
β’ R0 is a key epidemiological measure for how βinfectiousβ a disease is
β’ R0 = 1 is a threshold between epidemic / no epidemic
β’ R0 > 1: Disease can invade
β’ R0 < 1: Disease will die out
Definition: Basic reproductive ratio R0
The average number of people an infectious person will infect, assuming that the rest of the population is susceptible
The basic reproductive ratio R0
β’ R0 is a key epidemiological measure for how βinfectiousβ a disease is
β’ E.g. R0 = 2 (Contagion)
1 billionRepeat 30 times
1 infected
2 infected
4 infected
8 infected
16infected
In Contagion, Dr. Erin Mears (Kate Winslet) explains R0
Examples for R0
BSE0 β€ π 0 β€ 14
Scrapie 1.6 β€ π 0 β€ 3.9
Foot & Mouth Disease
1.6 β€ π 0 β€ 4.6
Modelling Epidemics: The SIR model
β’ X = nr of susceptibles, Y = nr of infectives, Z = nr of recoveredβ’ Describes acute infections transmitted by infected individuals;β’ Pathogen causes illness for a period of time followed by death or life-long
immunity
Susceptibles Infectives Recovered/ Removed
S I R
Modelling Epidemics: The SIR model
We must determine:β’ The rate at which susceptible individuals get infected (SI) β’ The rate at which infected individuals recover (or die) (IR)This gives rise to 2 model parameters: β’ The transmission term Ξ²β’ The recovery rate Ξ³
Susceptibles Infectives Recovered/ Removed
S I RΞ² Ξ³
Transmission rate SI
β’ Depends on the prevalence of infectives, the contact rate and the probability of transmission given contact
β’ New infectives are produced at a rate Ξ» x X, where X = nr of susceptibles
Definition: Transmission coefficient Ξ² = contact rate x transmission probability
Definition Force of infection Ξ»: Per capita rate at which susceptible individuals contract the infection
Frequency versus density dependent transmission
β’ Force of infection Ξ» depends on the number of infectious individuals (Y(t))
β’ Frequency dependent transmission: Ξ»(t) = π½xπ(π‘)/π(π‘)β’ Force of infection depends on frequency of infectives
β’ Assumption holds for most human diseases where contact is determined by social constraints rather than population size
β’ Density dependent transmission: Ξ»(t) = π½xπ(π‘)β’ Force of infection increases with population size (e.g. individuals crowded in a small
space)
β’ Assumption appropriate for plant and animal diseases
β’ The distinction only matters if the population size varies.β’ Otherwise 1/N can be absorbed into the parameter Ξ²
The SIR model without demography
β’ Consider a closed population of constant size N
β’ Model VariablesS(t) = proportion of susceptibles (X(t)/N)
I(t) = proportion of infectives (Y(t)/N)
R(t) = proportion of recovered (Z(t)/N)
β’ Model Parametersβ’ Ξ²: transmission coefficient
β’ Ξ³: recovery rate ( the inverse of average infectious period)
S I R
The SIR model without demography
β’ Model equations
S I R
β’ Equations describe the rate at which the proportions of susceptible, infectious and recovered individuals change over time
β’ The model cannot be solved explicitly, i.e. no analytical expression for S(t), I(t), R(t)!β’ Need computer programme
β’ Constant population size implies S(t) + I(t) + R(t) = 1 for all times t
π πΊ
π π= βπ· πΊ π°
π π°
π π= π· πΊ π° β πΈπ°
π πΉ
π π= πΈπ°
With initial conditions
S(t=0) = S(0)
I(t=0) = I(0)
R(t=0) = R(0)
The Threshold Phenomenon
Imagine a scenario where I0 infectives are introduced into a susceptible population.
Will there be an epidemic?
β’ Epidemic will occur if the proportion of infectives increases with time: ππΌ
ππ‘> 0
β’ From the 2nd equation of the SIR model:ππΌ
ππ‘= π½ππΌ β πΎπΌ = πΌ π½π β πΎ > 0 only if S > Ξ³/Ξ²
β’ Thus, the infection will only invade if the initial proportion of susceptibles
S0 > Ξ³/Ξ²
What does this result imply for vaccination or other prevention strategies?
S I RΞ² Ξ³
Imagine a scenario where I0 infectives are introduced into a susceptible population.Will there be an epidemic?β’ The infection will only invade if the initial proportion of susceptibles π0 >
πΎπ½
β’ An average infected individual β’ is infectious for a period of 1/Ξ³ daysβ’ infects Ξ² susceptible individuals per dayβ’ will thus generate Ξ² x 1/Ξ³ new infections over its lifetime
π 0 = π½
πΎ
β’ Infection can only invade if π0 > 1 π 0
The Threshold Phenomenon & R0
S I RΞ² Ξ³
Epidemic burnout
Imagine a scenario where I0 infectives are introduced into a susceptible population with π0 >
πΎπ½
What happens in the long-term?
What proportion of the population will contract the infection?
S I RΞ² Ξ³
ππ
ππ‘= βπ½ π πΌ
ππΌ
ππ‘= π½ π πΌ β πΎπΌ
ππ
ππ‘= πΎπΌ
ππ
ππ = β
π½π
πΎ= β π 0π
π π‘ = π 0 πβπ π‘ π 0
Given R(t) β€1, this implies that the proportion of susceptibles remains always positive with π(π‘) > πβπ 0 for any time t What will stop the epidemic?
Solve:
Epidemic burnoutWhat happens in the long-term?
What proportion of the population will contract the infection?
S I RΞ² Ξ³
π π‘ = π 0 πβπ π‘ π 0From above:
At end of epidemic: I = 0. Given S+I+R=1, we can use this equation to calculatethe final size πΊ β of the epidemic (R β = 1 β π β ):
π β = π 0 π(π β β1)π 0
β’ The above equation can only be solved numerically for π β . It produces the graph on the right.
β’ This equation if often used to estimate R0.For R0=2, what proportion of the population will eventually get infected if everybody is initially susceptible?
Basic reproductive ratio, R0
Fra
ctio
n in
fect
ed
NoEpidemic
Dynamic behaviour
β’ The exact time profiles depend on the model parameters and on the initial conditions S(0), I(0), R(0)
β’ See Tutorial 1 for investigating the impact of these on prevalence profiles
S I RΞ² Ξ³
ππ
ππ‘= βπ½ π πΌ
ππΌ
ππ‘= π½ π πΌ β πΎπΌ
ππ
ππ‘= πΎπΌ
The SIR model with demography
β’ Assume the epidemic progresses at a slower time scale so that the assumption of a closed population is no longer valid
β’ Assume a natural host lifespan of 1/ΞΌ years and that the birth rate is similar to the mortality rate ΞΌ (i.e. population size is constant)
Generalized SIR model equations:
π πΊ
π π= π β π· πΊ π° β ππΊ
π π°
π π= π· πΊ π° β πΈπ° β ππ°
π πΉ
π π= πΈπ° β ππΉ
With initial conditions:S(t=0) = S(0)I(t=0) = I(0)R(t=0) = R(0)
S I RΞ² Ξ³ΞΌΞΌ ΞΌ ΞΌ
ππ
ππ‘+ππΌ
ππ‘+ππ
ππ‘= 0
The SIR model with demography
β’ Transmission rate per infective is Ξ²
β’ Each infective individual spends on average 1
πΎ+πtime units in class I
β’ π 0 is always smaller than for a closed population
π πΊ
π π= π β π· πΊ π° β ππΊ
π π°
π π= π· πΊ π° β πΈπ° β ππ°
π πΉ
π π= πΈπ° β ππΉ
S I RΞ² Ξ³ΞΌΞΌ ΞΌ ΞΌ
πΉπ =π·
πΈ + π
Equilibrium state
What is the final outcome of the infection?
β’ The disease will eventually settle into an equilibrium state ( S*, I*, R*) where
β’ Setting the SIR model equations to zero leads to 2 equilibria (outcomes):
( S*, I*, R*) = (1, 0, 0) Disease free equilibirum
( S*, I*, R*) = 1
π 0,π
π½π 0 β 1 , 1 β
1
π 0β
π
π½π 0 β 1 Endemic equilibrium
ππ
ππ‘=ππΌ
ππ‘=ππ
ππ‘= 0
π πΊ
π π= π β π· πΊ π° β ππΊ
π π°
π π= π· πΊ π° β πΈπ° β ππ°
π πΉ
π π= πΈπ° β ππΉ
Which outcome will be achieved?
S I RΞ² Ξ³ΞΌΞΌ
Equilibrium state
Which outcome will be achieved?
Disease free or endemic equilibrium?
This can be answered with mathematical stability analysis:β’ Determines for which parameter values a specific equilibrium is
stable to small perturbations
β’ It can be shown that
β’ The disease free equilibrium is stable if R0 < 1
β’ The endemic equilibrium is stable if R0 > 1
If an infection can invade (i.e. if R0 > 1) , then the topping up of the susceptible pool causes the disease to persist
S I RΞ² Ξ³ΞΌΞΌ ΞΌ ΞΌ
Dynamic behaviour
β’ The SIR model with demography generates damped oscillations:
Time
Fact
ion
infe
cted
, I Period T
β’ With some algebra* it can be shown that:
β’ The period T ~ ππππ πππ ππ ππππππ‘πππ β ππππ ππ’πππ‘πππ ππ ππππππ‘πππ’π ππππππ
β’ Mean age of infection π΄ βπΏ
π 0 β1where L = 1/ΞΌ is the average life expectancy
β’ Measures of A and L lead to estimates for R0
Equilibrium
*See e.g. Keeling & Rohani 2008
S I RΞ² Ξ³ΞΌΞΌ ΞΌ ΞΌ
Adding complexity: The SIRS model
β’ The SIR model assumes lifelong immunity
β’ What if this is not the case, i.e. assume immunity is lost at a rate Ξ±
S I RΞ² Ξ³ΞΌ
ΞΌ ΞΌ ΞΌ
Ξ±
π πΊ
π π= π + πΆπΉ β π· πΊ π° β ππΊ
π π°
π π= π· πΊ π° β πΈπ° β ππ°
π πΉ
π π= πΈπ° β πΆπΉ β ππΉ
πΉπ =π·
πΈ + π
β’ Loss of immunity does not affect the onset of the disease (same R0!)
β’ What effect does loss of immunity have on the disease dynamics & equilibrium? See tutorial
Adding complexity: Infections with a carrier state
β’ Assume a proportion p of infected individuals become chronic carriers
β’ These carriers transmit the infection at a reduced rate ππ½, with Ξ΅<1
S IR
π πΊ
π π= π β π·π° + πΊπ·πͺ πΊ β ππΊ
π π°
π π= π·π° + πΊπ·πͺ πΊ β πΈπ° β ππ°
π πΉ
π π= πΈ(π β π)π° β ππΉ
C
Acutely infecteds
Chronic Carriers
π πͺ
π π= πΈππ° β ππͺ
p
1-p
πΉπ =π·
πΈ + π+ππΈ
π
πΊπ·
(πΈ + π)
Adding complexity: Infections with a carrier state
β’ Assume a proportion p of infected individuals become chronic carriers
β’ These carriers transmit the infection at a reduced rate ππ½, with Ξ΅<1
S IR
πΉπ =π·
πΈ + π+ππΈ
π
πΊπ·
(πΈ + π)
C
Acutely infecteds
Chronic Carriers
p
1-p
β’ Asymptomatic chronic carriers can cause underestimation of R0
β’ What effect does the presence of chronic carriers have on the disease dynamics & equilibrium? See tutorial
Summary
β’ Epidemics can be represented by compartmental ODE models
β’ Even the simplest epidemiological models require computer algorithms to estimate prevalence profilesβ’ But criteria for invasion and for equilibrium conditions can be derived analytically
β’ The basic reproductive ratio R0 is a key epidemiological measure affecting criteria for invasion extinction and size of the epidemic
β’ An infection experiences deterministic extinction if R0 <1
β’ In the absence of demography, strongly immunizing infections will always go extinct eventually and not all individuals will have become infected
β’ In the SIR model with demography, the endemic equilibrium is feasible if R0>1. Prevalence curves approach this equilibrium through damped oscillations.
References
β’ Keeling, Matt J., and Pejman Rohani. Modeling infectious diseases in humans and animals. Princeton University Press, 2008.
β’ Anderson, Roy M., Robert M. May, and B. Anderson. Infectious diseases of humans: dynamics and control. Vol. 28. Oxford: Oxford university press, 1992.
β’ Heesterbeek, J. A. Mathematical epidemiology of infectious diseases: model building, analysis and interpretation. Vol. 5. John Wiley & Sons, 2000.
β’ Roberts, M., & Heesterbeek, H. (1993). Bluff your way in epidemic models.Trends in microbiology, 1(9), 343-348.
β’ Hesterbeek J.A., A brief history of R0 and a recipe for its calculation. Acta Biotheor. 2002, 50(4):375-6.
β’ Heffernan, J. M., R. J. Smith, and L. M. Wahl. Perspectives on the basic reproductive ratio. Journal of the Royal Society Interface 2.4 (2005): 281-293.