Compute R0 Using Next Generation Operator

Baojun Song, Ph.D.

Department of Mathematical Sciences

Montclair State University

June 20, 2016

songb@mail.montclair.edu

B. Song (Montclair State) Compute R0 June 20, 2016 1 / 1

Compute R0 Using Next Generation Operator

Reference: P. van den Driessche and J.

Watmough (2002). Reproduction numbers and
sub-threshold endemic equilibria for
compartmental models of disease transmission,
Mathematical Biosciences, 180:29–48

B. Song (Montclair State) Compute R0 June 20, 2016 2 / 1

Definition and a Theorem

Basic reproductive number (R0 ) is defined as the average number of

secondary infections when a typical infective enters an entirely
susceptible population.

B. Song (Montclair State) Compute R0 June 20, 2016 3 / 1

Definition and a Theorem

Basic reproductive number (R0 ) is defined as the average number of

secondary infections when a typical infective enters an entirely
susceptible population.
Theorem
If R0 < 1, then DFE (disease-free equilibrium) is
locally asymptotically stable(L.A.S.). If R0 > 1, then
DFE is unstable.

B. Song (Montclair State) Compute R0 June 20, 2016 3 / 1

Definition and a Theorem

Basic reproductive number (R0 ) is defined as the average number of

secondary infections when a typical infective enters an entirely
susceptible population.
Theorem
If R0 < 1, then DFE (disease-free equilibrium) is
locally asymptotically stable(L.A.S.). If R0 > 1, then
DFE is unstable.
If you compute R0 , you need not prove the stability of DFE.

B. Song (Montclair State) Compute R0 June 20, 2016 3 / 1

Two Types of Epidemiological Classes

• Let X be vector of infected classes, such as

exposed, infectious, carrier, etc.

B. Song (Montclair State) Compute R0 June 20, 2016 4 / 1

Two Types of Epidemiological Classes

• Let X be vector of infected classes, such as

exposed, infectious, carrier, etc.
• Let Y be vector of uninfected classes, such as
susceptible, recovered, etc.

B. Song (Montclair State) Compute R0 June 20, 2016 4 / 1

Rearrange System of Equations

dX dY
= F(X, Y ) − V(X, Y ); = W(X, Y )
dt dt

B. Song (Montclair State) Compute R0 June 20, 2016 5 / 1

Rearrange System of Equations

dX dY
= F(X, Y ) − V(X, Y ); = W(X, Y )
dt dt
F(X, Y ): Vector of new infection rates (flows from Y to X).

B. Song (Montclair State) Compute R0 June 20, 2016 5 / 1

Rearrange System of Equations

dX dY
= F(X, Y ) − V(X, Y ); = W(X, Y )
dt dt
F(X, Y ): Vector of new infection rates (flows from Y to X).
V(X, Y ): Vector of all other rates (not new infection). These rates
include flows from X to Y (for instance, recovery rates), flows within X
and flows leaving from the system (for instance, death rates). For each
compartment, in-flow in V is negative and out-flow in V is positive.

B. Song (Montclair State) Compute R0 June 20, 2016 5 / 1

Rearrange System of Equations

dX dY
= F(X, Y ) − V(X, Y ); = W(X, Y )
dt dt
F(X, Y ): Vector of new infection rates (flows from Y to X).
V(X, Y ): Vector of all other rates (not new infection). These rates
include flows from X to Y (for instance, recovery rates), flows within X
and flows leaving from the system (for instance, death rates). For each
compartment, in-flow in V is negative and out-flow in V is positive.
DFE is (0, Y ).

B. Song (Montclair State) Compute R0 June 20, 2016 5 / 1

Rearrange System of Equations

dX dY
= F(X, Y ) − V(X, Y ); = W(X, Y )
dt dt
F(X, Y ): Vector of new infection rates (flows from Y to X).
V(X, Y ): Vector of all other rates (not new infection). These rates
include flows from X to Y (for instance, recovery rates), flows within X
and flows leaving from the system (for instance, death rates). For each
compartment, in-flow in V is negative and out-flow in V is positive.
DFE is (0, Y ).
F(0, Y ) = 0 and V(0, Y ) = 0.

B. Song (Montclair State) Compute R0 June 20, 2016 5 / 1

Jacobian around DFE

F(X, Y ) and V(X, Y ) are vector-valued functions of X and Y .

B. Song (Montclair State) Compute R0 June 20, 2016 6 / 1

Jacobian around DFE

F(X, Y ) and V(X, Y ) are vector-valued functions of X and Y .

h i
 ∂F ∂F
 ∂F
DF|(0,Y ) = ∂X ∂Y (0,Y ) = ∂X (0,Y ) 0

B. Song (Montclair State) Compute R0 June 20, 2016 6 / 1

Jacobian around DFE

F(X, Y ) and V(X, Y ) are vector-valued functions of X and Y .

h i
 ∂F ∂F
 ∂F
DF|(0,Y ) = ∂X ∂Y (0,Y ) = ∂X (0,Y ) 0

h i
 ∂V ∂V
 ∂V
DV|(0,Y ) = ∂X ∂Y (0,Y ) = ∂X (0,Y ) 0

B. Song (Montclair State) Compute R0 June 20, 2016 6 / 1

Formula for R0

   
∂F ∂V
F = , V =
∂X (0,Y ) ∂X (0,Y )

B. Song (Montclair State) Compute R0 June 20, 2016 7 / 1

Formula for R0

   
∂F ∂V
F = , V =
∂X (0,Y ) ∂X (0,Y )

F V −1 is called the next generation matrix.

B. Song (Montclair State) Compute R0 June 20, 2016 7 / 1

Formula for R0

   
∂F ∂V
F = , V =
∂X (0,Y ) ∂X (0,Y )

F V −1 is called the next generation matrix.

The spectral radius of F V −1 is equal to R0 .

B. Song (Montclair State) Compute R0 June 20, 2016 7 / 1

Formula for R0

   
∂F ∂V
F = , V =
∂X (0,Y ) ∂X (0,Y )

F V −1 is called the next generation matrix.

The spectral radius of F V −1 is equal to R0 . The spectral radius of

F V −1 is equal to the dominant eigenvalue of F V −1 that is the
maximum eigenvalue of F V −1 .

B. Song (Montclair State) Compute R0 June 20, 2016 7 / 1

Formula for R0

   
∂F ∂V
F = , V =
∂X (0,Y ) ∂X (0,Y )

F V −1 is called the next generation matrix.

The spectral radius of F V −1 is equal to R0 . The spectral radius of

F V −1 is equal to the dominant eigenvalue of F V −1 that is the
maximum eigenvalue of F V −1 .

R0 = the maximum eigenvalue of F V −1

B. Song (Montclair State) Compute R0 June 20, 2016 7 / 1

Example 1: An SIR Model

Model Equations

dS I
= Λ − βS − µS,
dt N
dI I
= βS − (µ + γ)I,
dt N
dR
= γI − µR,
dt
N = S + I + R.

B. Song (Montclair State) Compute R0 June 20, 2016 8 / 1

Example 1: An SIR Model

Model Equations Rearrange

dS I X = I, Y = [S, R]T
= Λ − βS − µS,
dt N
I
dI I F(I) = βS ,
= βS − (µ + γ)I, N
dt N
dR V(I, S, R) = (µ + γ)I
= γI − µR, dI
dt = F(I) − V(I, S, R)
N = S + I + R. dt
DFE: (0, Λ/µ, 0)

B. Song (Montclair State) Compute R0 June 20, 2016 8 / 1

Example 1: An SIR Model

Model Equations Rearrange

dS I X = I, Y = [S, R]T
= Λ − βS − µS,
dt N
I
dI I F(I) = βS ,
= βS − (µ + γ)I, N
dt N
dR V(I, S, R) = (µ + γ)I
= γI − µR, dI
dt = F(I) − V(I, S, R)
N = S + I + R. dt
DFE: (0, Λ/µ, 0)
   
∂F ∂V
F = = β, V = =µ+γ
∂I (0,Λ/µ,0) ∂I (0,Λ/µ,0)

B. Song (Montclair State) Compute R0 June 20, 2016 8 / 1

Example 1: An SIR Model

   
∂F ∂V
F = = β, V = =µ+γ
∂I (0,Λ/µ,0) ∂I (0,Λ/µ,0)

B. Song (Montclair State) Compute R0 June 20, 2016 9 / 1

Example 1: An SIR Model

   
∂F ∂V
F = = β, V = =µ+γ
∂I (0,Λ/µ,0) ∂I (0,Λ/µ,0)

1 β
V −1 = , F V −1 =
µ+γ µ+γ

B. Song (Montclair State) Compute R0 June 20, 2016 9 / 1

Example 1: An SIR Model

   
∂F ∂V
F = = β, V = =µ+γ
∂I (0,Λ/µ,0) ∂I (0,Λ/µ,0)

1 β
V −1 = , F V −1 =
µ+γ µ+γ

β
R0 =
µ+γ

B. Song (Montclair State) Compute R0 June 20, 2016 9 / 1

Example 2: An SEIR Model

Model Equations

dS I
= Λ − βS − µS,
dt N
dE I
= βS − (µ + k + r1 )E,
dt N
dI
= kE − (γ + µ)I,
dt
dR
= r1 E + γI − µR.
dt

B. Song (Montclair State) Compute R0 June 20, 2016 10 / 1

Example 2: An SEIR Model

Model Equations
Rearrange Equations

dS I    
= Λ − βS − µS, E S
dt N X= , Y =
dE I I R
= βS − (µ + k + r1 )E,
βS NI
 
dt N F=
dI 0
= kE − (γ + µ)I,
dt
 
(µ + k + r1 )E
dR V=
= r1 E + γI − µR. −kE + (γ + µ)I
dt

B. Song (Montclair State) Compute R0 June 20, 2016 10 / 1

Example 2: An SEIR Model

Model Equations
Rearrange Equations

dS I    
= Λ − βS − µS, E S
dt N X= , Y =
dE I I R
= βS − (µ + k + r1 )E,
βS NI
 
dt N F=
dI 0
= kE − (γ + µ)I,
dt
 
(µ + k + r1 )E
dR V=
= r1 E + γI − µR. −kE + (γ + µ)I
dt

DFE: (0, 0, Λ/µ, 0)

B. Song (Montclair State) Compute R0 June 20, 2016 10 / 1

Example 2: An SEIR Model

  
0 β µ + k + r1 0
F = , V =
0 0 −k γ+µ

B. Song (Montclair State) Compute R0 June 20, 2016 11 / 1

Example 2: An SEIR Model

  
0 β µ + k + r1 0
F = , V =
0 0 −k γ+µ
 
1 γ+µ 0
V −1 =
(µ + k + r1 )(γ + µ) k µ + k + r1

B. Song (Montclair State) Compute R0 June 20, 2016 11 / 1

Example 2: An SEIR Model

  
0 β µ + k + r1 0
F = , V =
0 0 −k γ+µ
 
−1 1 γ+µ 0
V =
(µ + k + r1 )(γ + µ) k µ + k + r1
 
−1 1 kβ β(µ + k + r1 )
FV =
(µ + k + r1 )(γ + µ) 0 0

B. Song (Montclair State) Compute R0 June 20, 2016 11 / 1

Example 2: An SEIR Model

  
0 β µ + k + r1 0
F = , V =
0 0 −k γ+µ
 
−1 1 γ+µ 0
V =
(µ + k + r1 )(γ + µ) k µ + k + r1
 
−1 1 kβ β(µ + k + r1 )
FV =
(µ + k + r1 )(γ + µ) 0 0

  
k β
R0 =
µ + k + r1 γ+µ

B. Song (Montclair State) Compute R0 June 20, 2016 11 / 1

Example 3: An SIS Model with Two Strains

dS
= µ − β1 I1 S − β2 I2 S − µS + γ1 I1 + γ2 I2 ,
dt
dI1
= β1 I1 S − (µ + γ1 )I1 + νI1 I2 ,
dt
dI2
= β2 I2 S − (µ + γ2 )I2 − νI1 I2 .
dt

B. Song (Montclair State) Compute R0 June 20, 2016 12 / 1

Example 3: An SIS Model with Two Strains

dS
= µ − β1 I1 S − β2 I2 S − µS + γ1 I1 + γ2 I2 ,
dt
dI1
= β1 I1 S − (µ + γ1 )I1 + νI1 I2 ,
dt
dI2
= β2 I2 S − (µ + γ2 )I2 − νI1 I2 .
dt

     
I1 β1 I1 S (µ + γ1 )I1 − νI1 I2
X= , Y = S, F= , V=
I2 β2 I2 S (µ + γ2 )I2 + νI1 I2

B. Song (Montclair State) Compute R0 June 20, 2016 12 / 1

Example 3: An SIS Model with Two Strains

dS
= µ − β1 I1 S − β2 I2 S − µS + γ1 I1 + γ2 I2 ,
dt
dI1
= β1 I1 S − (µ + γ1 )I1 + νI1 I2 ,
dt
dI2
= β2 I2 S − (µ + γ2 )I2 − νI1 I2 .
dt

     
I1 β1 I1 S (µ + γ1 )I1 − νI1 I2
X= , Y = S, F= , V=
I2 β2 I2 S (µ + γ2 )I2 + νI1 I2

DFE: (0, 0, 1)

B. Song (Montclair State) Compute R0 June 20, 2016 12 / 1

Example 3: An SIS Model with Two Strains

       
f1 β 1 I1 S v1 (µ + γ1 )I1 − νI1 I2
F= = , V= =
f2 β 2 I2 S v2 (µ + γ2 )I2 + νI1 I2

" #
∂f1 ∂f1  
∂I1 ∂I2 β1 0
F = ∂f2 ∂f2 =
∂I1 ∂I2
0 β2
(0,0,1)
" # " #
∂v1 ∂v1   1
µ + γ1 0 −1 0
V = ∂I1
∂v2
∂I2
∂v2 = ,V = µ+γ1 1
∂I1 ∂I2
0 µ + γ2 0 µ+γ2
(0,0,1)

B. Song (Montclair State) Compute R0 June 20, 2016 13 / 1

Example 3: An SIS Model with Two Strains

" #
 1
β 0 0
F = 1 , V −1
= µ+γ1
1
0 β2 0 µ+γ2

B. Song (Montclair State) Compute R0 June 20, 2016 14 / 1

Example 3: An SIS Model with Two Strains

" #
 1
β 0 0
F = 1 , V −1
= µ+γ1
1
0 β2 0 µ+γ2

 " 1
# "
β1
#
β 0 0 0
FV −1
= 1 µ+γ1
1 = µ+γ1
β2
0 β2 0 µ+γ2 0 µ+γ2

B. Song (Montclair State) Compute R0 June 20, 2016 14 / 1

Example 3: An SIS Model with Two Strains

" #
 1
β 0 0
F = 1 , V −1
= µ+γ1
1
0 β2 0 µ+γ2

 " 1
# "
β1
#
β 0 0 0
FV −1
= 1 µ+γ1
1 = µ+γ1
β2
0 β2 0 µ+γ2 0 µ+γ2

 
β1 β2
R0 = max ,
µ + γ1 µ + γ2

B. Song (Montclair State) Compute R0 June 20, 2016 14 / 1

A TB Model with Exogenous Reinfections

dS I
= Λ − β1 S − µS + r1 L + r2 I,
dt N
dE I I
= β1 S − β2 E − (µ + k + r1 )E,
dt N N
dI I
= β2 E + kE − (µ + d + r2 )I,
dt N
N =S+E+I

B. Song (Montclair State) Compute R0 June 20, 2016 15 / 1

A TB Model with Exogenous Reinfections

dS I
= Λ − β1 S − µS + r1 L + r2 I,
dt N
dE I I
= β1 S − β2 E − (µ + k + r1 )E,
dt N N
dI I
= β2 E + kE − (µ + d + r2 )I,
dt N
N=S+E+I
 
E
X= , Y = [S]
I
β1 S NI (µ + k + r1 )E + β2 E NI
   
F= , V=
0 −kE + (µ + d + r2 )I − β2 E NI

B. Song (Montclair State) Compute R0 June 20, 2016 15 / 1

A TB Model with Exogenous Reinfections

 
E
X= , Y = [S]
I
β SI (µ + k + r1 )E + β2 E NI
   
F= 1 N , V=
0 −kE + (µ + d + r2 )I − β2 E NI

DFE=(0, 0, Λ/µ)

B. Song (Montclair State) Compute R0 June 20, 2016 16 / 1

A TB Model with Exogenous Reinfections

 
E
X= , Y = [S]
I
β SI (µ + k + r1 )E + β2 E NI
   
F= 1 N , V=
0 −kE + (µ + d + r2 )I − β2 E NI

DFE=(0, 0, Λ/µ)
   
0 β1 µ + k + r1 0
F = , V =
0 0 −k µ + d + r2

B. Song (Montclair State) Compute R0 June 20, 2016 16 / 1

A TB Model with Exogenous Reinfections
   
0 β1 µ + k + r1 0
F = , V =
0 0 −k µ + d + r2

B. Song (Montclair State) Compute R0 June 20, 2016 17 / 1

A TB Model with Exogenous Reinfections
   
0 β1 µ + k + r1 0
F = , V =
0 0 −k µ + d + r2

 
−1 1 µ + d + r2 0
V =
(µ + k + r1 )(µ + d + r2 ) k µ + k + r1

B. Song (Montclair State) Compute R0 June 20, 2016 17 / 1

A TB Model with Exogenous Reinfections
   
0 β1 µ + k + r1 0
F = , V =
0 0 −k µ + d + r2

 
−1 1 µ + d + r2 0
V =
(µ + k + r1 )(µ + d + r2 ) k µ + k + r1

 
−1 1 kβ1 β1 (µ + k + r1 )
FV =
(µ + k + r1 )(µ + d + r2 ) 0 0

B. Song (Montclair State) Compute R0 June 20, 2016 17 / 1

A TB Model with Exogenous Reinfections
   
0 β1 µ + k + r1 0
F = , V =
0 0 −k µ + d + r2

 
−1 1 µ + d + r2 0
V =
(µ + k + r1 )(µ + d + r2 ) k µ + k + r1

 
−1 1 kβ1 β1 (µ + k + r1 )
FV =
(µ + k + r1 )(µ + d + r2 ) 0 0

  
β1 k
R0 =
µ + d + r2 µ + k + r1

B. Song (Montclair State) Compute R0 June 20, 2016 17 / 1