Reproduction numbers and sub-threshold endemic

equilibria for compartmental models of disease transmission

P. van den Driessche
a,1
, James Watmough
b,
*
,2
a
Department of Mathematics and Statistics, University of Victoria, Victoria, BC, Canada V8W 3P4
b
Department of Mathematics and Statistics, University of New Brunswick, Fredericton, NB, Canada E3B 5A3
Received 26 April 2001; received in revised form 27 June 2001; accepted 27 June 2001
Dedicated to the memory of John Jacquez
Abstract
A precise denition of the basic reproduction number, R
0
, is presented for a general compartmental
disease transmission model based on a system of ordinary dierential equations. It is shown that, if R
0
< 1,
then the disease free equilibrium is locally asymptotically stable; whereas if R
0
> 1, then it is unstable.
Thus, R
0
is a threshold parameter for the model. An analysis of the local centre manifold yields a simple
criterion for the existence and stability of super- and sub-threshold endemic equilibria for R
0
near one. This
criterion, together with the denition of R
0
, is illustrated by treatment, multigroup, staged progression,
multistrain and vectorhost models and can be applied to more complex models. The results are signicant
for disease control.
2002 Elsevier Science Inc. All rights reserved.
Keywords: Basic reproduction number; Sub-threshold equilibrium; Disease transmission model; Disease control
1. Introduction
One of the most important concerns about any infectious disease is its ability to invade a
population. Many epidemiological models have a disease free equilibrium (DFE) at which the
Mathematical Biosciences 180 (2002) 2948
www.elsevier.com/locate/mbs
*
Corresponding author. Tel.: +1-506 458 7323; fax: +1-506 453 4705.
E-mail addresses: [email protected] (P. van den Driessche), [email protected] (J. Watmough).
URL: http://www.math.unb.ca/$watmough.
1
Research supported in part by an NSERC Research Grant, the University of Victoria Committee on faculty
research and travel and MITACS.
2
Research supported by an NSERC Postdoctoral Fellowship tenured at the University of Victoria.
0025-5564/02/$ - see front matter 2002 Elsevier Science Inc. All rights reserved.
PII: S0025- 5564( 02) 00108- 6
population remains in the absence of disease. These models usually have a threshold parameter,
known as the basic reproduction number, R
0
, such that if R
0
< 1, then the DFE is locally as-
ymptotically stable, and the disease cannot invade the population, but if R
0
> 1, then the DFE is
unstable and invasion is always possible (see the survey paper by Hethcote [1]). Diekmann et al.
[2] dene R
0
as the spectral radius of the next generation matrix. We write down in detail a general
compartmental disease transmission model suited to heterogeneous populations that can be
modelled by a system of ordinary dierential equations. We derive an expression for the next
generation matrix for this model and examine the threshold R
0
1 in detail.
The model is suited to a heterogeneous population in which the vital and epidemiological
parameters for an individual may depend on such factors as the stage of the disease, spatial
position, age or behaviour. However, we assume that the population can be broken into homo-
geneous subpopulations, or compartments, such that individuals in a given compartment are
indistinguishable from one another. That is, the parameters may vary from compartment to
compartment, but are identical for all individuals within a given compartment. We also assume
that the parameters do not depend on the length of time an individual has spent in a compart-
ment. The model is based on a system of ordinary equations describing the evolution of the
number of individuals in each compartment.
In addition to showing that R
0
is a threshold parameter for the local stability of the DFE,
we apply centre manifold theory to determine the existence and stability of endemic equilib-
ria near the threshold. We show that some models may have unstable endemic equilibria near
the DFE for R
0
< 1. This suggests that even though the DFE is locally stable, the disease may
persist.
The model is developed in Section 2. The basic reproduction number is dened and shown to be
a threshold parameter in Section 3, and the denition is illustrated by several examples in Section
4. The analysis of the centre manifold is presented in Section 5. The epidemiological ramications
of the results are presented in Section 6.
2. A general compartmental epidemic model for a heterogeneous population
Consider a heterogeneous population whose individuals are distinguishable by age, behaviour,
spatial position and/or stage of disease, but can be grouped into n homogeneous compartments. A
general epidemic model for such a population is developed in this section. Let x x
1
; . . . ; x
n

t
,
with each x
i
P0, be the number of individuals in each compartment. For clarity we sort the
compartments so that the rst m compartments correspond to infected individuals. The distinc-
tion between infected and uninfected compartments must be determined from the epidemiological
interpretation of the model and cannot be deduced from the structure of the equations alone, as
we shall discuss below. It is plausible that more than one interpretation is possible for some
models. A simple epidemic model illustrating this is given in Section 4.1. The basic reproduction
number can not be determined from the structure of the mathematical model alone, but depends
on the denition of infected and uninfected compartments. We dene X
s
to be the set of all disease
free states. That is
X
s
fx P0jx
i
0; i 1; . . . ; mg:
30 P. van den Driessche, J. Watmough / Mathematical Biosciences 180 (2002) 2948
In order to compute R
0
, it is important to distinguish new infections from all other changes in
population. Let F
i
x be the rate of appearance of new infections in compartment i, V

i
x be the
rate of transfer of individuals into compartment i by all other means, and V

i
x be the rate of
transfer of individuals out of compartment i. It is assumed that each function is continuously
dierentiable at least twice in each variable. The disease transmission model consists of non-
negative initial conditions together with the following system of equations:
_ xx
i
f
i
x F
i
x V
i
x; i 1; . . . ; n; 1
where V
i
V

i
V

i
and the functions satisfy assumptions (A1)(A5) described below. Since
each function represents a directed transfer of individuals, they are all non-negative. Thus,
(A1) if x P0, then F
i
; V

i
; V

i
P0 for i 1; . . . ; n.
If a compartment is empty, then there can be no transfer of individuals out of the compartment
by death, infection, nor any other means. Thus,
(A2) if x
i
0 then V

i
0. In particular, if x 2 X
s
then V

i
0 for i 1; . . . ; m.
Consider the disease transmission model given by (1) with f
i
x, i 1; . . . ; n, satisfying con-
ditions (A1) and (A2). If x
i
0, then f
i
x P0 and hence, the non-negative cone (x
i
P0,
i 1; . . . ; n) is forward invariant. By Theorems 1.1.8 and 1.1.9 of Wiggins [3, p. 37] for each non-
negative initial condition there is a unique, non-negative solution.
The next condition arises from the simple fact that the incidence of infection for uninfected
compartments is zero.
(A3) F
i
0 if i > m.
To ensure that the disease free subspace is invariant, we assume that if the population is free of
disease then the population will remain free of disease. That is, there is no (density independent)
immigration of infectives. This condition is stated as follows:
(A4) if x 2 X
s
then F
i
x 0 and V

i
x 0 for i 1; . . . ; m.
The remaining condition is based on the derivatives of f near a DFE. For our purposes, we
dene a DFE of (1) to be a (locally asymptotically) stable equilibrium solution of the disease free
model, i.e., (1) restricted to X
s
. Note that we need not assume that the model has a unique DFE.
Consider a population near the DFE x
0
. If the population remains near the DFE (i.e., if the
introduction of a few infective individuals does not result in an epidemic) then the population will
return to the DFE according to the linearized system
_ xx Df x
0
x x
0
; 2
where Df x
0
is the derivative of
i
=ox
j
evaluated at the DFE, x
0
(i.e., the Jacobian matrix). Here,
and in what follows, some derivatives are one sided, since x
0
is on the domain boundary. We restrict
our attention to systems in which the DFE is stable in the absence of new infection. That is,
(A5) If Fx is set to zero, then all eigenvalues of Df x
0
have negative real parts.
P. van den Driessche, J. Watmough / Mathematical Biosciences 180 (2002) 2948 31
The conditions listed above allow us to partition the matrix Df x
0
as shown by the following
lemma.
Lemma 1. If x
0
is a DFE of (1) and f
i
x satises (A1)(A5), then the derivatives DFx
0
and
DVx
0
are partitioned as
DFx
0

F 0
0 0
_ _
; DVx
0

V 0
J
3
J
4
_ _
;
where F and V are the m m matrices dened by
F
oF
i
ox
j
x
0

_ _
and V
oV
i
ox
j
x
0

_ _
with 1 6i; j 6m:
Further, F is non-negative, V is a non-singular M-matrix and all eigenvalues of J
4
have positive real
part.
Proof. Let x
0
2 X
s
be a DFE. By (A3) and (A4), oF
i
=ox
j
x
0
0 if either i > m or j > m.
Similarly, by (A2) and (A4), if x 2 X
s
then V
i
x 0 for i 6m. Hence, oV
i
=ox
j
x
0
0 for i 6m
and j > m. This shows the stated partition and zero blocks. The non-negativity of F follows from
(A1) and (A4).
Let fe
j
g be the Euclidean basis vectors. That is, e
j
is the jth column of the n n identity matrix.
Then, for j 1; . . . ; m,
oV
i
ox
j
_ _
x
0
lim
h!0

V
i
x
0
he
j
V
i
x
0

h
_ _
:
To show that V is a non-singular M-matrix, note that if x
0
is a DFE, then by (A2) and (A4),
V
i
x
0
0 for i 1; . . . ; m, and if i 6 j, then the ith component of x
0
he
j
0 and
V
i
x
0
he
j
60, by (A1) and (A2). Hence, oV
i
=ox
j
0 for i m and j 6 i and V has the Z sign
pattern (see Appendix A). Additionally, by (A5), all eigenvalues of V have positive real parts.
These two conditions imply that V is a non-singular M-matrix [4, p. 135 (G
20
)]. Condition
(A5) also implies that the eigenvalues of J
4
have positive real part.
3. The basic reproduction number
The basic reproduction number, denoted R
0
, is the expected number of secondary cases
produced, in a completely susceptible population, by a typical infective individual [2]; see also [5,
p. 17]. If R
0
< 1, then on average an infected individual produces less than one new infected
individual over the course of its infectious period, and the infection cannot grow. Conversely, if
R
0
> 1, then each infected individual produces, on average, more than one new infection, and the
disease can invade the population. For the case of a single infected compartment, R
0
is simply
the product of the infection rate and the mean duration of the infection. However, for more
complicated models with several infected compartments this simple heuristic denition of R
0
is
32 P. van den Driessche, J. Watmough / Mathematical Biosciences 180 (2002) 2948
insucient. A more general basic reproduction number can be dened as the number of new
infections produced by a typical infective individual in a population at a DFE.
To determine the fate of a typical infective individual introduced into the population, we
consider the dynamics of the linearized system (2) with reinfection turned o. That is, the system
_ xx DVx
0
x x
0
: 3
By (A5), the DFE is locally asymptotically stable in this system. Thus, (3) can be used to de-
termine the fate of a small number of infected individuals introduced to a disease free population.
Let w
i
0 be the number of infected individuals initially in compartment i and let
wt w
1
t; . . . ; w
m
t
t
be the number of these initially infected individuals remaining in the
infected compartments after t time units. That is the vector w is the rst m components of x. The
partitioning of DVx
0
implies that wt satises w
0
t V wt, which has the unique solution
wt e
Vt
w0. By Lemma 1, V is a non-singular M-matrix and is, therefore, invertible and all of
its eigenvalues have positive real parts. Thus, integrating F wt from zero to innity gives the
expected number of new infections produced by the initially infected individuals as the vector
FV
1
w0. Since F is non-negative and V is a non-singular M-matrix, V
1
is non-negative [4, p. 137
(N
38
)], as is FV
1
.
To interpret the entries of FV
1
and develop a meaningful denition of R
0
, consider the fate of
an infected individual introduced into compartment k of a disease free population. The (j; k) entry
of V
1
is the average length of time this individual spends in compartment j during its lifetime,
assuming that the population remains near the DFE and barring reinfection. The (i; j) entry of F is
the rate at which infected individuals in compartment j produce new infections in compartment i.
Hence, the (i; k) entry of the product FV
1
is the expected number of new infections in com-
partment i produced by the infected individual originally introduced into compartment k. Fol-
lowing Diekmann et al. [2], we call FV
1
the next generation matrix for the model and set
R
0
qFV
1
; 4
where qA denotes the spectral radius of a matrix A.
The DFE, x
0
, is locally asymptotically stable if all the eigenvalues of the matrix Df x
0
have
negative real parts and unstable if any eigenvalue of Df x
0
has a positive real part. By Lemma 1,
the eigenvalues of Df x
0
can be partitioned into two sets corresponding to the infected and
uninfected compartments. These two sets are the eigenvalues of F V and those of J
4
. Again by
Lemma 1, the eigenvalues of J
4
all have negative real part, thus the stability of the DFE is
determined by the eigenvalues of F V . The following theorem states that R
0
is a threshold
parameter for the stability of the DFE.
Theorem 2. Consider the disease transmission model given by (1) with f x satisfying conditions
(A1)(A5). If x
0
is a DFE of the model, then x
0
is locally asymptotically stable if R
0
< 1, but un-
stable if R
0
> 1, where R
0
is dened by (4).
Proof. Let J
1
F V . Since V is a non-singular M-matrix and F is non-negative, J
1
V F
has the Z sign pattern (see Appendix A). Thus,
sJ
1
< 0 () J
1
is a non-singular M-matrix;
P. van den Driessche, J. Watmough / Mathematical Biosciences 180 (2002) 2948 33
where sJ
1
denotes the maximum real part of all the eigenvalues of the matrix J
1
(the spectral
abscissa of J
1
). Since FV
1
is non-negative, J
1
V
1
I FV
1
also has the Z sign pattern. Ap-
plying Lemma 5 of Appendix A, with H V and B J
1
V F , we have
J
1
is a non-singular M-matrix () I FV
1
is a non-singular M-matrix:
Finally, since FV
1
is non-negative, all eigenvalues of FV
1
have magnitude less than or equal to
qFV
1
. Thus,
I FV
1
is a non-singular M-matrix; () qFV
1
< 1:
Hence, sJ
1
< 0 if and only if R
0
< 1.
Similarly, it follows that
sJ
1
0 ()J
1
is a singular M-matrix;
() I FV
1
is a singular M-matrix;
() qFV
1
1:
The second equivalence follows from Lemma 6 of Appendix A, with H V and K F . The
remainder of the equivalences follow as with the non-singular case. Hence, sJ
1
0 if and only
if R
0
1. It follows that sJ
1
> 0 if and only if R
0
> 1.
A similar result can be found in the recent book by Diekmann and Heesterbeek [6, Theorem
6.13]. This result is known for the special case in which J
1
is irreducible and V is a positive di-
agonal matrix [710]. The special case in which V has positive diagonal and negative subdiagonal
elements is proven in Hyman et al. [11, Appendix B]; however, our approach is much simpler (see
Section 4.3).
4. Examples
4.1. Treatment model
The decomposition of f x into the components F and V is illustrated using a simple treat-
ment model. The model is based on the tuberculosis model of Castillo-Chavez and Feng [12, Eq.
(1.1)], but also includes treatment failure used in their more elaborate two-strain model [12, Eq.
(2.1)]. A similar tuberculosis model with two treated compartments is proposed by Blower et al.
[13]. The population is divided into four compartments, namely, individuals susceptible to tu-
berculosis (S), exposed individuals (E), infectious individuals (I) and treated individuals (T). The
dynamics are illustrated in Fig. 1. Susceptible and treated individuals enter the exposed com-
partment at rates b
1
I=N and b
2
I=N, respectively, where N E I S T. Exposed individuals
progress to the infectious compartment at the rate m. All newborns are susceptible, and all indi-
viduals die at the rate d > 0. Thus, the core of the model is an SEI model using standard inci-
dence. The treatment rates are r
1
for exposed individuals and r
2
for infectious individuals.
However, only a fraction q of the treatments of infectious individuals are successful. Unsuc-
cessfully treated infectious individuals re-enter the exposed compartment (p 1 q). The disease
34 P. van den Driessche, J. Watmough / Mathematical Biosciences 180 (2002) 2948
transmission model consists of the following dierential equations together with non-negative
initial conditions:
_
EE b
1
SI=N b
2
TI=N d m r
1
E pr
2
I; 5a
_
II mE d r
2
I; 5b
_
SS bN dS b
1
SI=N; 5c
_
TT dT r
1
E qr
2
I b
2
TI=N: 5d
Progression from E to I and failure of treatment are not considered to be new infections, but
rather the progression of an infected individual through the various compartments. Hence,
F
b
1
SI=N b
2
TI=N
0
0
0
_
_
_
_
_
_
_
_
and V
d m r
1
E pr
2
I
mE d r
2
I
bN dS b
1
SI=N
dT r
1
E qr
2
I b
2
TI=N
_
_
_
_
_
_
_
_
: 6
The infected compartments are E and I, giving m 2. An equilibrium solution with E I 0 has
the form x
0
0; 0; S
0
; 0
t
, where S
0
is any positive solution of bS
0
dS
0
. This will be a DFE
if and only if b
0
S
0
< d. Without loss of generality, assume S
0
1 is a DFE. Then,
F
0 b
1
0 0
_ _
; V
d m r
1
pr
2
m d r
2
_ _
;
giving
V
1

1
d m r
1
d r
2
mpr
2
d r
2
pr
2
m d m r
1
_ _
and R
0
b
1
m=d m r
1
d r
2
mpr
2
. A heuristic derivation of the (2; 1) entry of V
1
and
R
0
are as follows: a fraction h
1
m=d m r
1
of exposed individuals progress to compartment
I, a fraction h
2
pr
2
=d r
2
of infectious individuals re-enter compartment E. Hence, a fraction
h
1
of exposed individuals pass through compartment I at least once, a fraction h
2
1
h
2
pass through
Fig. 1. Progression of infection from susceptible (S) individuals through the exposed (E), infected (I), and treated (T)
compartments for the treatment model of (5a)(5d).
P. van den Driessche, J. Watmough / Mathematical Biosciences 180 (2002) 2948 35
at least twice, and a fraction h
k
1
h
k1
2
pass through at least k times, spending an average of s
1=d r
2
time units in compartment I on each pass. Thus, an individual introduced into com-
partment E spends, on average, sh
1
h
2
1
h
2
sh
1
=1 h
1
h
2
m=d m r
1
d r
2

mpr
2
time units in compartment I over its expected lifetime. Multiplying this by b
1
gives R
0
.
The model without treatment (r
1
r
2
0) is an SEI model with R
0
b
1
m=dd m. The
interpretation of R
0
for this case is simpler. Only a fraction m=d m of exposed individuals
progress from compartment E to compartment I, and individuals entering compartment I spend,
on average, 1=d time units there.
Although conditions (A1)(A5) do not restrict the decomposition of f
i
x to a single choice for
F
i
, only one such choice is epidemiologically correct. Dierent choices for the function Flead to
dierent values for the spectral radius of FV
1
, as shown in Table 1. In column (a), treatment
failure is considered to be a new infection and in column (b), both treatment failure and pro-
gression to infectiousness are considered new infections. In each case the condition qFV
1
< 1
yields the same portion of parameter space. Thus, qFV
1
is a threshold parameter in both cases.
The dierence between the numbers lies in the epidemiological interpretation rather than the
mathematical analysis. For example, in column (a), the infection rate is b
1
pr
2
and an exposed
individual is expected to spend m=d m r
1
d r
2
time units in compartment I. However,
this reasoning is biologically awed since treatment failure does not give rise to a newly infected
individual.
Table 1
Decomposition of f leading to alternative thresholds
(a) (b)
F b
1
SI=N b
2
TI=N pr
2
I
0
0
0
_
_
_
_
_
_
_
_
b
1
SI=N b
2
TI=N pr
2
I
mE
0
0
_
_
_
_
_
_
_
_
V d m r
1
E
mE d r
2
I
bN dS b
1
SI=N
dT r
1
E qr
2
I b
2
TI=N
_
_
_
_
_
_
_
_
d m r
1
E
d r
2
I
bN dS b
1
SI=N
dT r
1
E qr
2
I b
2
TI=N
_
_
_
_
_
_
_
_
F
0 b
1
pr
2
0 0
_ _
0 b
1
pr
2
m 0
_ _
V
d m r
1
0
m d r
2
_ _
d m r
1
0
0 d r
2
_ _
q(FV
1
) b
1
m pr
2
m
d m r
1
d r
2

b
1
m pr
2
m
d m r
1
d r
2

36 P. van den Driessche, J. Watmough / Mathematical Biosciences 180 (2002) 2948

4.2. Multigroup model
In the epidemiological literature, the term multigroup usually refers to the division of a het-
erogeneous population into several homogeneous groups based on individual behaviour (e.g.,
[14]). Each group is then subdivided into epidemiological compartments. The majority of mul-
tigroup models in the literature are used for sexually transmitted diseases, such as HIV/AIDS or
gonorrhea, where behaviour is an important factor in the probability of contracting the disease
[7,8,14,15]. As an example, we use an m-group SIRS-vaccination model of Hethcote [7,14] with a
generalized incidence term. The sample model includes several SI multigroup models of HIV/
AIDS as special cases [8,15]. The model equations are as follows:
_
I
i
I
i

m
j1
b
ij
xS
i
I
j
d
i
c
i

i
I
i
; 7a
_
S
i
S
i
1 p
i
b
i
d
i
h
i
S
i
r
i
R
i

m
j1
b
ij
xS
i
I
j
; 7b
_
R
i
R
i
p
i
b
i
c
i
I
i
h
i
S
i
d
i
r
i
R
i
; 7c
for i 1; . . . ; m, where x I
1
; . . . ; I
m
; S
1
; . . . ; S
m
; R
1
; . . . ; R
m

t
. Susceptible and removed individu-
als die at the rate d
i
> 0, whereas infected individuals die at the faster rate d
i

i
. Infected in-
dividuals recover with temporary immunity from re-infection at the rate c
i
, and immunity lasts an
expected 1=r
i
time units. All newborns are susceptible, and a constant fraction b
i
are born into
each group. A fraction p
i
of newborns are vaccinated at birth. Thereafter, susceptible individuals
are vaccinated at the rate h
i
. The incidence, b
ij
x depends on individual behaviour, which
determines the amount of mixing between the dierent groups (see, e.g., Jacquez et al. [16]).
The DFE for this model is
x
0
0; . . . ; 0; S
0
1
; . . . ; S
0
m
; R
0
1
; . . . ; R
0
m

t
;
where
S
0
i

b
i
d
i
1 p
i
r
i

d
i
d
i
h
i
r
i

;
R
0
i

b
i
h
i
d
i
p
i

d
i
d
i
h
i
r
i

:
Linearizing (7a) about x x
0
gives
F S
0
i
b
ij
x
0

_
and
V d
i
c
i

i
d
ij
;
where d
ij
is one if i j, but zero otherwise. Thus,
FV
1
S
0
i
b
ij
x
0
=d
i
_
c
i

i

:
P. van den Driessche, J. Watmough / Mathematical Biosciences 180 (2002) 2948 37
For the special case with b
ij
separable, that is, b
ij
x a
i
xk
j
x, F has rank one, and the basic
reproduction number is
R
0

m
i1
S
0
i
a
i
x
0
k
i
x
0

d
i
c
i

i
: 8
That is, the basic reproduction number of the disease is the sum of the reproduction numbers for
each group.
4.3. Staged progression model
The staged progression model [11, Section 3 and Appendix B] has a single uninfected com-
partment, and infected individuals progress through several stages of the disease with changing
infectivity. The model is applicable to many diseases, particularly HIV/AIDS, where transmission
probabilities vary as the viral load in an infected individual changes. The model equations are as
follows (see Fig. 2):
_
I
1
I
1

m1
k1
b
k
SI
k
=N m
1
d
1
I
1
; 9a
_
I
i
I
i
m
i1
I
i1
m
i
d
i
I
i
; i 2; . . . ; m 1; 9b
_
I
m
I
m
m
m1
I
m1
d
m
I
m
; 9c
_
SS b bS

m1
k1
b
k
SI
k
=N: 9d
The model assumes standard incidence, death rates d
i
> 0 in each infectious stage, and the nal
stage has a zero infectivity due to morbidity. Infected individuals spend, on average, 1=m
i
time
units in stage i. The unique DFE has I
i
0, i 1; . . . ; m and S 1. For simplicity, dene m
m
0.
Then F F
ij
and V V
ij
, where
F
ij

b
j
i 1; j 6m 1;
0 otherwise;
_
10
V
ij

m
i
d
i
j i;
m
j
i 1 j;
0 otherwise:
_
_
_
11
Fig. 2. Progression diagram for the staged progression model of (9a)(9d).
38 P. van den Driessche, J. Watmough / Mathematical Biosciences 180 (2002) 2948
Let a
ij
be the (i; j) entry of V
1
. Then
a
ij

0 i < j;
1=m
i
d
i
i j;

i1
kj
m
k

i
kj
m
k
d
k

j < i:
_

_
12
Thus,
R
0

b
1
m
1
d
1

b
2
m
1
m
1
d
1
m
2
d
2

b
3
m
1
m
2
m
1
d
1
m
2
d
2
m
3
d
3

b
m1
m
1
. . . m
m2
m
1
d
1
. . . m
m1
d
m1

: 13
The ith term in R
0
represents the number of new infections produced by a typical individual
during the time it spends in the ith infectious stage. More specically, m
i1
=m
i1
d
i1
is the
fraction of individuals reaching stage i 1 that progress to stage i, and 1=m
i
d
i
is the average
time an individual entering stage i spends in stage i. Hence, the ith term in R
0
is the product of the
infectivity of individuals in stage i, the fraction of initially infected individuals surviving at least to
stage i, and the average infectious period of an individual in stage i.
4.4. Multistrain model
The recent emergence of resistant viral and bacterial strains, and the eect of treatment on their
proliferation is becoming increasingly important [12,13]. One framework for studying such sys-
tems is the multistrain model shown in Fig. 3, which is a caricature of the more detailed treatment
model of Castillo-Chavez and Feng [12, Section 2] for tuberculosis and the coupled two-strain
vectorhost model of Feng and Velasco-Hernaandez [17] for Dengue fever. The model has only a
single susceptible compartment, but has two infectious compartments corresponding to the
two infectious agents. Each strain is modelled as a simple SIS system. However, strain one may
super-infect an individual infected with strain two, giving rise to a new infection in compartment
Fig. 3. Progression diagram for the multistrain model of (14a)(14c).
P. van den Driessche, J. Watmough / Mathematical Biosciences 180 (2002) 2948 39
I
1
. The parameter m > 0 is the contact rate for the super-infection. The model equations are as
follows:
_
I
1
I
1
b
1
I
1
S b c
1
I
1
mI
1
I
2
; 14a
_
I
2
I
2
b
2
I
2
S b c
2
I
2
mI
1
I
2
; 14b
_
SS b bS c
1
I
1
c
2
I
2
b
1
I
1
b
2
I
2
S: 14c
For simplicity we have scaled the birth and death rates to b > 0. Hence, the DFE is x
0
0; 0; 1
t
,
and
F
b
1
0
0 b
2
_ _
; V
b c
1
0
0 b c
2
_ _
; 15
with V non-singular as required. The next generation matrix, FV
1
, has the two eigenvalues
R
i

b
i
b c
i
; i 1; 2: 16
In this example, J
1
F V is reducible and (14a) and (14b) decouple near the DFE. The two
eigenvalues correspond to the reproduction numbers for each strain. The basic reproduction
number for the system is the maximum of the two. That is,
R
0
max
i2f1;2g
R
i
: 17
An alternate interpretation of this model is that I
1
is the sole infected compartment and that I
2
is
an uninfected compartment. The strain two equilibrium is 0; 1 b c
2
=b
2
; b c
2
=b
2
. Lin-
earizing about this equilibrium gives F b
1
b c
2
=b
2
m1 b c
2
=b
2
, and V b c
1
.
Thus,
R
12

R
1
R
2

m
b c
1
1
_

1
R
2
_
18
is the reproduction number for strain one near the strain two equilibrium. The interesting case is,
of course, if R
2
> 1 > R
1
, but R
12
> 1. That is, strain two can invade the DFE, but strain one
cannot, and yet strain one can invade the strain two equilibrium. This can occur if m is suciently
large.
4.5. Vectorhost model
The general framework developed in Section 2 includes vectorhost models. As an example,
consider the following simplication of the two-strain, vectorhost model proposed by Feng and
Velasco-Hernaandez [17] for Dengue fever. The model couples a simple SIS model for the hosts
with an SI model for the vectors. The four compartments correspond to infected hosts (I), infected
vectors (V ), susceptible hosts (S) and susceptible vectors (M). Hosts are infected by contacts with
infected vectors, and vectors are in turn infected by contacts with infected hosts. These infection
rates are given by the two terms b
s
SV and b
m
MI. The model is written as follows (see Fig. 4):
40 P. van den Driessche, J. Watmough / Mathematical Biosciences 180 (2002) 2948
_
II b
s
SV b cI; 19a
_
VV b
m
MI cV ; 19b
_
SS b bS cI b
s
SV ; 19c
_
MM c cM b
m
MI: 19d
The birth and death rates have been scaled to b > 0 for the host and c > 0 for the vector. Thus,
the DFE is x
0
0; 0; 1; 1
t
,
F
0 b
s
b
m
0
_ _
; V
b c 0
0 c
_ _
; 20
with V non-singular, and the basic reproduction number is
R
0

b
s
b
m
cb c

: 21
Near the DFE, each infected host produces b
m
=c new infected vectors over its expected infectious
period, and each infected vector produces b
s
=b c new infected hosts over its expected infec-
tious period. The square root arises from the two generations required for an infected vector or
host to reproduce itself.
5. The existence of sub-threshold equilibria
5.1. Analysis of the centre manifold near x x
0
, R
0
1
In this section we consider the nature of the equilibrium solutions of the disease transmission
model near the bifurcation point x x
0
, R
0
1. Since R
0
is often inconvenient to use directly as a
Fig. 4. Progression diagram for the vectorhost model of (19a)(19d).
P. van den Driessche, J. Watmough / Mathematical Biosciences 180 (2002) 2948 41
bifurcation parameter, we introduce a bifurcation parameter l. Let l be a bifurcation parameter
such that R
0
< 1 for l < 0 and R
0
> 1 for l > 0 and such that x
0
is a DFE for all values of l.
Consider the system
_ xx f x; l; 22
where f is as described in Section 2, with the further restriction that f is continuously dierentiable
at least twice in both x and l. The DFE is the line x
0
; l and the local stability of the DFE
changes at the point x
0
; 0. We use results of centre manifold theory (see e.g., [3]) to show that
there are non-trivial (endemic) equilibria near the bifurcation point x
0
; 0. Before stating these
results we introduce some notation and collect a few facts.
We use the notation D
x
f x
0
; 0 for the partial derivative of f with respect to x evaluated at the
point x x
0
, l 0. Assume that the zero eigenvalue of D
x
f x
0
; 0 is simple and let v and w be the
corresponding left and right nullvectors chosen such that vw 1. By Lemma 1 and Theorem 2, all
other eigenvalues of D
x
f x
0
; 0 have negative real parts. Let
a
v
2
D
xx
f x
0
; 0w
2

1
2

n
i;j;k1
v
i
w
j
w
k
o
2
f
i
ox
j
ox
k
x
0
; 0; 23
b vD
xl
f x
0
; 0w

n
i;j1
v
i
w
j
o
2
f
i
ox
j
ol
x
0
; 0: 24
We show below that the sign of a determines the nature of the endemic equilibria near the bi-
furcation point. First, however, we note that the expression for a can be written in a dierent form
using results of the previous sections.
Lemma 3. If f x; l is continuously dierentiable at least twice in both x and l and conditions (A1)
(A5) are satised, and 0 is a simple eigenvalue of D
x
f x
0
; 0, then in the nullvectors of D
x
f x
0
; 0,
v
i
P0 and w
i
P0 for i 1; . . . ; m, v
i
0 for i m 1; . . . ; n, and
a

m
i;j;k1
v
i
w
j
w
k
1
2
o
2
f
i
ox
j
ox
k
x
0
; 0
_

n
lm1
a
lk
o
2
f
i
ox
j
ox
l
x
0
; 0
_
; 25
with a
lk
, l m 1; . . . ; n, k 1; . . . ; m, denoting the l m; k entry of J
1
4
J
3
where J
3
and J
4
are
the lower blocks of D
x
f x
0
; 0 DFx
0
Vx
0
j
R
0
1
dened in Lemma 1.
Proof. By Lemma 1 and Theorem 2, the rst m components of v and w are the left and right null
vectors of J
1
. Since J
1
is essentially non-negative (i.e., J
1
has the Z sign pattern), v and w can be
chosen such that v
i
P0 and w
i
P0 for i 1; . . . ; m [4]. Further, since the eigenvalues of J
4
all have
positive real parts, J
1
4
exists and the remaining components of v must be zero. Hence, from the
denition of a
lk
,
w
l

m
k1
a
lk
w
k
; l m 1; . . . ; n: 26
With these facts, (23) leads to (25) as follows:
42 P. van den Driessche, J. Watmough / Mathematical Biosciences 180 (2002) 2948
a
1
2

m
i1
v
i

n
j;k1
w
j
w
k
o
2
f
i
ox
j
ox
k
x
0
; 0

1
2

m
i1
v
i

m
j;k1
w
j
w
k
o
2
f
i
ox
j
ox
k
_
2

m
j1

n
lm1
w
j
w
l
o
2
f
i
ox
j
ox
l
_

x
0
;0

1
2

m
i1
v
i

m
j;k1
w
j
w
k
o
2
f
i
ox
j
ox
k
_
2

m
j1

n
lm1
w
j

m
k1
a
lk
w
k
o
2
f
i
ox
j
ox
l
_

x
0
;0

m
i;j;k1
v
i
w
j
w
k
1
2
o
2
f
i
ox
j
ox
k
_

n
lm1
a
lk
o
2
f
i
ox
j
ox
l
_

x
0
;0
:
For the second step, the second partial derivatives with respect to the uninfected compartments
are zero by (A2)(A4) (the details are similar to those in the proof of Lemma 1):
Since the rst m components of v and w are non-negative, the sign of a is determined by the
signs of the partial derivatives and of a
lk
. In many applications, the rst set of partial derivatives
are negative. Hence the sign of a is determined by the mixed partial derivatives involving both
infected and uninfected compartments and a
lk
.
Theorem 4. Consider the disease transmission model dened by (22) with the function f x; l
satisfying the conditions (A1)(A5) of Section 2 and the parameter l as described above. Assume
that the zero eigenvalue of D
x
f x
0
; 0 is simple. Let a and b be as dened by (23) and (24) and assume
that b 6 0. Then, there exists d > 0 such that
(i) if a < 0, then there are locally asymptotically stable endemic equilibria near x
0
for 0 < l < d and
(ii) if a > 0, then there are unstable endemic equilibria near x
0
for d < l < 0.
Proof. Centre manifold theory [3, Theorem 2.1.1] states that there exists a local centre manifold
parameterized by u and l of the form
W
c
fx; ljx x
0
uw zu; lg; 27
where zu; l is orthogonal to w and is second order in both u and l. Further, the centre manifold,
W
c
, is invariant under (22). That is,
_ uuw
dz
dt
f x
0
uw zu; l; l: 28
Premultiplying (28) by v leads to the equation
_ uu vf x
0
uw zu; l; l; 29
since vz 0 for all u; l. Centre manifold theory further states [3, Theorem 2.1.2] that the be-
haviour of solutions of (22) near the bifurcation point x
0
; 0 is governed by (29). The right-hand
side of (29) can be expanded in a Taylor series as follows:
P. van den Driessche, J. Watmough / Mathematical Biosciences 180 (2002) 2948 43
_ uu vf x
0
; 0 vD
l
f x
0
; 0l vD
x
f x
0
; 0uw z
v
2
D
ll
f x
0
; 0l
2
vD
xl
f x
0
; 0luw z
v
2
D
xx
f x
0
; 0uw z
2
O3: 30
The notation O3 is used to denote terms of third order and higher in u and l. Since f x
0
; l 0
for all l, the rst, second and fourth terms in the expansion are zero, and since v is a left null
vector of D
x
f x
0
; 0, the third term vanishes. Hence, all remaining terms involving z are higher
order, and
_ uu au
2
bul O3; 31
where a and b are dened by (23) and (24).
For d > 0 suciently small, there are non-zero, steady state solutions of (31) near the line
u bl=a for jlj < d. Since we have chosen l so that the DFE is stable for l < 0, a local stability
analysis of (31) shows that b must be positive. Further these non-zero solutions are stable if a < 0
and unstable if a > 0. Since the rst m components of w are non-negative, it follows that the
endemic solutions of (22) corresponding to these non-zero solutions of (31) are feasible (i.e., the
components of x are non-negative) only if either l > 0 and a < 0 or if l < 0 and a > 0.
In summary, the nature of the bifurcation at R
0
1 is given by the sign of a. If either a or b are
zero, then higher order terms in the Taylor series must be considered. If a is negative, then a
branch of super-threshold endemic equilibria exists, and the bifurcation is supercritical. If a > 0,
then there are unstable sub-threshold endemic equilibria, and the bifurcation is subcritical. These
cases are often referred to as a forward bifurcation and a backward bifurcation respectively.
5.2. Examples
5.2.1. Treatment models
The result of Theorem 4 can be applied to the tuberculosis example of Section 4.1, since J
1
has a
simple zero eigenvalue when R
0
1. All second derivatives of f
i
in (23) are zero at the DFE
except the following:
o
2
f
1
oEoI
b
1
;
o
2
f
1
oI
2
2b
1
;
o
2
f
1
oIoT
b
2
b
1
:
Hence,
a b
1
v
1
w
2
w
1
w
2
1 b
2
=b
1
w
4
:
Computation shows that the eigenvectors v and w can be chosen so that each component of w is
positive and v
1
is also positive. Since biologically b
2
< b
1
, it follows that a < 0. Hence, by The-
orem 4 the DFE is locally asymptotically stable if R
0
is slightly less than one (i.e., l < 0), and if
R
0
is slightly greater than one then the DFE is unstable and there is a locally asymptotically stable
positive equilibrium near the DFE. The positivity of the endemic equilibrium follows from the
positivity of infected components (w
1
and w
2
) of the right null vector. This vector gives the
direction of the invasion when the DFE is unstable.
44 P. van den Driessche, J. Watmough / Mathematical Biosciences 180 (2002) 2948
Castillo-Chavez et al. [18] propose the addition of a second infection term, b
3
EI=N, to f
2
and
the negative of that term to f
1
. Thus, progression from the exposed to the infected compartments
is not linear, but is increased by exogenous re-infection. This change does not alter the DFE or
R
0
. However, with this term
a b
1
v
1
w
2
w
1
w
2
1 b
2
=b
1
w
4
b
3
w
1
w
2
v
2
v
1
:
Calculation shows that v
2
v
1
> 0. Hence, the direction of the bifurcation changes if b
3
is suf-
ciently large. If b
3
is such that a > 0, then there exists an unstable sub-threshold endemic
equilibrium near the DFE. The signicance of this unstable equilibrium is not trivial. It implies
that, although the DFE is locally stable, perturbations above a small threshold can grow. Further,
if R
0
> 1, then the analysis of the centre manifold tells us not only that the DFE is unstable, but
that there is no non-zero stable equilibrium near the DFE, and thus a small invasion will grow
rapidly and to signicant proportions even for R
0
near one. The importance of this backward
bifurcation for disease control is discussed in Section 6.
5.2.2. Multigroup model
Next, consider the multigroup model of Section 4.2. The 3m 3m Jacobian matrix D
x
f x
0
; 0
can be partitioned into blocks corresponding to I, S and R compartments as follows:
D
x
f x
0
; 0
S
0
i
b
ij
x
0
d
i
c
i

i
d
ij
_
0 0
S
0
i
b
ij
x
0

_
d
i
h
i
d
ij
_
r
i
d
ij
_
c
i
d
ij
_
h
i
d
ij
_
d
i
r
i
d
ij
_
_
_
_
_
_
_:
The upper left block is J
1
evaluated at R
0
1, and the four lower right blocks comprise J
4
. Note
that J
4
is a non-singular M-matrix, and therefore sJ
4
< 0. Let w
S
i
w
mi
and w
R
i
w
2mi
, for
i 1; . . . ; m. Then,
w
S
i

d
i
r
i
d
i

i
d
i
c
i
d
i
d
i
h
i
r
i

w
i
; 32
w
R
i

h
i
d
i

i
d
i
c
i
d
i
d
i
h
i
r
i

w
i
: 33
Applying (25) with the second partial derivatives for this model leads to
a a
0

m
i;j;k1
v
i
w
j
S
0
i
w
k
ob
ij
oI
k
x
0

_
w
S
k
ob
ij
oS
k
x
0
w
R
k
ob
ij
oR
k
x
0

_
; 34
where
a
0

m
k1
v
k
w
2
k
d
k
c
k

k
d
k
c
k
d
k
r
k
d
k

k

b
k
d
k
1 p
k
r
k

< 0: 35
Consider now two cases of interest. First, if b
ij
is constant, then a a
0
< 0 and the bifurcation is
in the forward direction. Second, if b
ij
x b
ij
N
1
; . . . ; N
m
where N
i
I
i
S
i
R
i
, then w
k

w
S
k
w
R
k

k
w
k
=d
k
and (34) leads to
P. van den Driessche, J. Watmough / Mathematical Biosciences 180 (2002) 2948 45
a a
0

m
i;j;k1
v
i
w
j
w
k

k
S
0
i
d
k
ob
ij
oN
k
x
0
: 36
The results of several models [7,8] can be generalized using
b
ij
x
k
ij
N
i

K
ij

m
l1
r
l
N
l
: 37
For this model,
a

m
k1
d
k
v
k
w
2
k
d
k
c
k

k

k
p
k
c
k
d
k
r
k

b
k
d
k
1 p
k
r
k

m
i;k1

m
j1
K
kj
w
j
_ _
S
0
k
v
k
r
i

i
w
i
b
k

k
w
k
b
i

b
k
d
i

m
j1
r
j
b
j
=d
j
_ _
2
: 38
In the case studied by Hethcote and Van Ark [7], b
ij
x k
ij
=N
i
. That is, K
ij
0 in (37), and, by
(38), a < 0 and the bifurcation is always in the forward direction. Huang et al. [8] used this model
with p
i
h
i
c
i
r
i
0, k
ij
diagonal and K
ij
irreducible and found that backward bifurcation
is possible. Our results remove these restrictions.
6. Discussion
The analysis presented herein can be applied to a large class of compartmental epidemic models
that possess a DFE. The basic reproduction number, R
0
(given by (4)), is a threshold parameter
for these models. Moreover, the local analysis of the centre manifold yields a second parameter, a
(given by (23) or (25)), whose sign indicates the existence and stability of a branch of endemic
equilibria near the threshold R
0
1. The stability of these equilibria is important for disease
control, as there are large dierences in the solutions of the system between the two cases a < 0
and a > 0. For the forward bifurcation (a < 0), there are stable super-threshold endemic equi-
libria near the DFE. Thus, reducing R
0
through one lowers the incidence of the disease until it is
eliminated as R
0
passes below one. For a backward bifurcation (a > 0), there are unstable sub-
threshold endemic equilibria near the DFE. The unstable sub-threshold endemic equilibria in-
dicate that the DFE is stable only to very small perturbations, and that even a small perturbation
can result in an epidemic. Further, as R
0
increases through the threshold, there is a catastrophic
increase in disease incidence. The lack of a local super-threshold endemic equilibrium suggests the
existence of a non-local endemic equilibrium with a relatively large fraction of infected individ-
uals, or a periodic solution. Backward bifurcations have been studied in models for HIV/AIDS
[8,15], tuberculosis [18] and for BRSV [19].
Throughout the analysis, we have assumed that a well dened DFE exists. However, some
models may be cast in terms of fractions so that there is an equilibrium distribution of individuals
over the compartments even though the total population size is not constant. In this case the
analysis can be applied to the fractions of individuals in each compartment to yield a threshold
parameter (see, e.g., [20,21]). This threshold is not the basic reproduction number, since it is a
46 P. van den Driessche, J. Watmough / Mathematical Biosciences 180 (2002) 2948
threshold for the fraction rather than the number of infected individuals, but the analysis for both
the threshold condition and the direction of the bifurcation is similar.
Acknowledgements
The authors thank Dr. Herbert W. Hethcote for suggestions and for a preprint of [1] and
Dr. Charles R. Johnson for helpful comments on the material in Appendix A.
Appendix A
Let sA be the maximum real part of the eigenvalues of A (the spectral abscissa), and let qA
be the maximum modulus of the eigenvalues of A (the spectral radius). In Section 3, we make use
of several results from the theory of M-matrices. A matrix B b
ij
has the Z sign pattern if b
ij
60
for all i 6 j. If B sI P, where I is the identity matrix, P is non-negative (P P0 entrywise), and
s > qP, then B is a non-singular M-matrix; if s qP, then B is a singular M-matrix. There are
many denitions of M-matrices equivalent to the above. For example, if a matrix B has the Z sign
pattern and sB > 0, then B is a non-singular M-matrix [4, p. 135 (G
20
)].
Lemma 5. Let H be a non-singular M-matrix and suppose B and BH
1
have the Z sign pattern.
Then B is a non-singular M-matrix if and only if BH
1
is a non-singular M-matrix.
The forward implication is stated in a slightly dierent form as Exercise 6b of Horn and
Johnson [22, p. 127] and the reverse implication is stated in Berman and Plemmons [4, p. 159
(5.2)].
In general, this lemma does not hold if B a singular M-matrix. It can be shown to hold if B is
singular and irreducible. However, this is not sucient for our needs. More specically, our proof
of Theorem 4 makes use of the following lemma.
Lemma 6. Let H be a non-singular M-matrix and suppose K P0. Then,
(i) H K is a non-singular M-matrix if and only if H KH
1
is a non-singular M-matrix.
(ii) H K is a singular M-matrix if and only if H KH
1
is a singular M-matrix.
Proof. Let B H K. Then both B and BH
1
I KH
1
have the Z sign pattern. (Recall that
H
1
P0 since H is a non-singular M-matrix.) Hence, Lemma 5 implies statement (i). A separate
continuity argument can be constructed for each implication in the singular case.
References
[1] H.W. Hethcote, The mathematics of infectious diseases, SIAM Rev. 42 (2000) 599.
[2] O. Diekmann, J.A.P. Heesterbeek, J.A.J. Metz, On the denition and the computation of the basic reproduction
ratio R
0
in models for infectious diseases in heterogeneous populations, J. Math. Biol. 28 (1990) 365.
P. van den Driessche, J. Watmough / Mathematical Biosciences 180 (2002) 2948 47
[3] S. Wiggins, Introduction to Applied Nonlinear Dynamical Systems and Chaos, Springer, Berlin, 1990.
[4] A. Berman, R.J. Plemmons, Nonnegative Matrices in the Mathematical Sciences, Academic Press, New York,
1970.
[5] R.M. Anderson, R.M. May, Infectious Diseases of Humans, Oxford University, Oxford, 1991.
[6] O. Diekmann, J.A.P. Heesterbeek, Mathematical Epidemiology of Infectious Diseases: Model Building, Analyis
and Interpretation, Wiley, New York, 1999.
[7] H.W. Hethcote, J.W.V. Ark, Epidemiological models for heterogeneous populations: proportionate mixing,
parameter estimation, and immunization programs, Math. Biosci. 84 (1987) 85.
[8] W. Huang, K.L. Cooke, C. Castillo-Chavez, Stability and bifurcation for a multiple-group model for the dynamics
of HIV/AIDS transmission, SIAM J. Appl. Math. 52 (3) (1992) 835.
[9] C. Castillo-Chavez, Z. Feng, W. Huang, On the computation of R
0
and its role in global stability, in: C. Castillo-
Chavez, with S. Blower, P. van den Driessche, D. Kirschner, A.-A. Yakubu (Eds.), Mathematical Approaches for
Emerging and Reemerging Infectious Diseases: An Introduction, Springer, 2002, p. 229.
[10] A. Nold, Heterogeneity in disease-transmission modeling, Math. Biosci. 52 (1980) 227.
[11] J.M. Hyman, J. Li, E.A. Stanley, The dierential infectivity and staged progression models for the transmission of
HIV, Math. Biosci. 155 (1999) 77.
[12] C. Castillo-Chavez, Z. Feng, To treat of not to treat: the case of tuberculosis, J. Math. Biol. 35 (1997) 629.
[13] S.M. Blower, P.M. Small, P.C. Hopewell, Control strategies for tuberculosis epidemics: new models for old
problems, Science 273 (1996) 497.
[14] H.W. Hethcote, An immunization model for a heterogeneous population, Theor. Populat. Biol. 14 (1978) 338.
[15] J. Dusho, W. Huang, C. Castillo-Chavez, Backwards bifurcations and catastrophe in simple models of fatal
diseases, J. Math. Biol. 36 (1998) 227.
[16] J.A. Jacquez, C.P. Simon, J. Koopman, L. Sattenspiel, T. Perry, Modelling and analyzing HIV transmission: the
eect of contact patterns, Math. Biosci. 92 (1988) 119.
[17] Z. Feng, J.X. Velasco-Hernaandez, Competitive exclusion in a vectorhost model for the Dengue fever, J. Math.
Biol. 35 (1997) 523.
[18] C. Castillo-Chavez, Z. Feng, A.F. Capurro, A model for TB with exogenous reinfection, Theor. Populat. Biol. 57
(2000) 235.
[19] D. Greenhalgh, O. Diekmann, M.C.M. de Jong, Subcritical endemic steady states in mathematical models for
animal infections with incomplete immunity, Math. Biosci. 165 (2000) 1.
[20] S. Busenberg, P. van den Driessche, Disease transmission in multigroup populations of variable size, in: O. Arino,
D. Axelrod, M. Kimmel, M. Langlais (Eds.), Mathematical Population Dynamics: Analysis of Heterogeneity,
Theory of Epidemics, vol. 1, Wuerz, 1995, p. 15.
[21] X. Lin, H.W. Hethcote, P. van den Driessche, An epidemiological model for HIV/AIDS with proportional
recruitment, Math. Biosci. 118 (1993) 181.
[22] R.A. Horn, C.R. Johnson, Topics in Matrix Analysis, Cambridge University, Cambridge, 1991.
48 P. van den Driessche, J. Watmough / Mathematical Biosciences 180 (2002) 2948