Mathematical Biosciences 201 (2006) 58–71

www.elsevier.com/locate/mbs

Backward bifurcation of an epidemic model with treatmentq

Wendi Wang
Department of Mathematics, Southwest Normal University, Chongqing 400715, PR China

Received 12 July 2004; received in revised form 4 February 2005; accepted 3 December 2005
Available online 8 February 2006

Abstract

An epidemic model with a limited resource for treatment is proposed to understand the effect of the
capacity for treatment. It is assumed that treatment rate is proportional to the number of infectives below
the capacity and is a constant when the number of infectives is greater than the capacity. It is found that a
backward bifurcation occurs if the capacity is small. It is also found that there exist bistable endemic equi-
libria if the capacity is low.
Ó 2005 Elsevier Inc. All rights reserved.

Keywords: Epidemic; Bifurcation; Treatment; Bistable; Cycle

1. Introduction

The asymptotic behavior of epidemic models has been studied by many researchers (see [2–
5,9,13,15,17,18] and the references cited therein). For classical epidemic models, it is common that
a basic reproduction number is a threshold in a sense that a disease is persistent if the basic repro-
duction number is greater than 1, and dies out if it is below 1. In this case, the bifurcation leading
from a disease free equilibrium to an endemic equilibrium is forward. In recent years, papers
[1,6,8,11,12,16,14] found backward bifurcations due to social groups with different susceptibilities,

q
Research supported by the National Natural Science Fund of China (No. 10271096).
E-mail address: [email protected]

0025-5564/$ - see front matter Ó 2005 Elsevier Inc. All rights reserved.
doi:10.1016/j.mbs.2005.12.022
 W. Wang / Mathematical Biosciences 201 (2006) 58–71 59

pair formation, macroparasite infection, nonlinear incidences, and age structures in epidemic
models. In this case, the basic reproduction number does not describe the necessary elimination
effort; rather the effort is described by the value of the critical parameter at the turning point.
Thus, it is important to identify backward bifurcations to obtain thresholds for the control of
diseases.
The treatment is an important method to decrease the spread of diseases such as measles, tuber-
culosis and flu (Feng and Thieme [7], Wu and Feng [20], Hyman and Li [10]). In classical epidemic
models, the treatment rate of infectives is assumed to be proportional to the number of the infec-
tives. This is unsatisfactory because the resources for treatment should be quite large. In fact, every
community should have a suitable capacity for treatment. If it is too large, the community pays for
unnecessary cost. If it is too small, the community has the risk of the outbreak of a disease. Thus, it
is important to determine a suitable capacity for the treatment of a disease. In paper [19], we adopt
a constant treatment, which simulates a limited capacity for treatment. Note that a constant treat-
ment is suitable when the number of infectives is large. In this paper, we modify it into

rI; if 0 6 I 6 I 0 ;
T ðIÞ ¼ ð1:1Þ
k; if I > I 0 ;
where k = rI0. This means that the treatment rate is proportional to the number of the infectives
when the capacity of treatment is not reached, and otherwise, takes the maximal capacity. This
renders for example the situation where patients have to be hospitalized: the number of hospital
beds is limited. This is true also for the case where medicines are not sufficient. Evidently, this
improves the classical proportional treatment and the constant treatment in [19].
We will consider a population that is divided into three types: susceptible, infective and recov-
ered. Let S, I and R denote the numbers of susceptible, infective, recovered individuals, respec-
tively. Then the model to be studied takes the following form:
dS
¼ A  dS  kSI;
dt
dI
¼ kSI  ðd þ c þ ÞI  T ðIÞ; ð1:2Þ
dt
dR
¼ cI þ T ðIÞ  dR;
dt
where A is the recruitment rate of the population, d the natural death rate of the population, c the
natural recovery rate of infective individuals,  > 0 the disease-related death rate, k the infection
coefficient. It is assumed that all the parameters are positive constants. Clearly, R3þ is positively
invariant for system (1.2). Since the first two equations in (1.2) are independent of the variable
R, it suffices to consider the following reduced model:
dS
¼ A  dS  kSI;
dt ð1:3Þ
dI
¼ kSI  ðd þ c þ ÞI  T ðIÞ.
dt
The purpose of this paper is to show that (1.3) has a backward bifurcation if the capacity for
treatment is small. We will prove that (1.3) has bistable endemic equilibria if the capacity is low.
60 W. Wang / Mathematical Biosciences 201 (2006) 58–71

The organization of this paper is as follows. In the next section, we study the bifurcations of
(1.3). In Section 3 we present a global analysis and simulations of the model.

2. Equilibria

In this section, we first consider the equilibria of (1.3) and their local stability. E0 = (A/d, 0) is a
disease free equilibrium. An endemic equilibrium of (1.3) satisfies

A  dS  kSI ¼ 0;
ð2:1Þ
kSI  ðd þ c þ ÞI  T ðIÞ ¼ 0.
When 0 < I 6 I0, system (2.1) becomes

A  dS  kSI ¼ 0;
ð2:2Þ
kSI  ðd þ c þ  þ rÞI ¼ 0.
When I > I0, system (2.1) becomes

A  dS  kSI ¼ 0;
ð2:3Þ
kSI  ðd þ c þ ÞI  k ¼ 0.
Let
kA
R0 ¼ .
dðd þ c þ  þ rÞ
Then R0 is a basic reproduction number of (1.3). If R0 > 1, (2.2) admits a unique positive solution
E* = (S*, I*) where
d þcþþr
S ¼ ;
k
dðR0  1Þ
I ¼ .
k
Clearly, E* is an endemic equilibrium of (1.3) if and only if
k
1 < R0 6 1 þ I 0 . ð2:4Þ
d
In order to obtain positive solutions of system (2.3), we solve S from the first equation of (2.3)
to obtain S = A/(d + kI). Substituting it into the second equation, we have
kðd þ c þ ÞI 2 þ bI þ dk ¼ 0; ð2:5Þ
where
b ¼ d 2 þ dc þ d  Ak þ kk.
If b P 0, it is clear that (2.5) does not have a positive solution. Let us consider the case where
b < 0. If D = b2  4k(d + c + )dk, it is easy to obtain
D ¼ ðR0 dðd þ c þ  þ rÞ þ d 2 þ dc þ d þ kkÞ2  4kðd þ c þ Þkd.
 W. Wang / Mathematical Biosciences 201 (2006) 58–71 61

It follows that D P 0 is equivalent to

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
kk  dr kðd þ c þ Þkd
R0 P 1 þ þ2 , p0 ; ð2:6Þ
dðd þ c þ  þ rÞ ðd þ c þ  þ rÞd
or
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
kk  dr kðd þ c þ Þkd
R0 6 1 þ 2 . ð2:7Þ
dðd þ c þ  þ rÞ ðd þ c þ  þ rÞd
Note that b < 0 is equivalent to
kk  dr
R0 > 1 þ . ð2:8Þ
dðd þ c þ  þ rÞ
It follows that b < 0 and D P 0 if and only if (2.6) holds. Let us suppose that (2.6) holds. Then
(2.5) has two positive solutions I1 and I2 where
 pffiffiffiffi  pffiffiffiffi
b  D b þ D
I1 ¼ ; I2 ¼ . ð2:9Þ
ð2kðd þ c þ ÞÞ ð2kðd þ c þ ÞÞ
Set Si = A/(d + kIi) and Ei = (Si, Ii) for i = 1, 2. Then Ei is an endemic equilibrium of (1.3) if
Ii P I0. Let us consider the conditions under which I1 > I0. By the definitions, we see that it is
equivalent to
pffiffiffiffi
b  D > 2kðd þ c þ ÞI 0 . ð2:10Þ
This implies that
b þ 2kðd þ c þ ÞI 0 < 0. ð2:11Þ
It follows from the definition of b that
kk  dr kðd þ c þ ÞI 0
R0 > 1 þ þ2 , p1 . ð2:12Þ
dðd þ c þ  þ rÞ ðd þ c þ  þ rÞd
Further, (2.10) implies that
2
ðb þ 2kðd þ c þ ÞI 0 Þ > D. ð2:13Þ
By direct calculations, we see that (2.13) is equivalent to
kI 0
R0 < 1 þ , p2 . ð2:14Þ
d
Hence, I1 > I0 holds if and only if (2.12) and (2.14) are valid. Moreover, if R0 6 p1 or R0 P p2 , we
have I1 6 I0.
By similar arguments as above, we see that I0 < I2 if (2.12) holds or
p 2 < R0 6 p 1 . ð2:15Þ
Furthermore, I2 6 I0 if
R0 6 minfp1 ; p2 g. ð2:16Þ
Summarizing the discussions above, we have the following conclusions.
62 W. Wang / Mathematical Biosciences 201 (2006) 58–71

Theorem 2.1. E* = (S*, I*) is an endemic equilibrium of (1.3) if and only if 1 < R0 6 p2 .
Furthermore, E* is the unique endemic equilibrium of (1.3) if 1 < R0 6 p2 and one of the following
conditions is satisfied:
(i) R0 < p0 .
(ii) p0 6 R0 < p1 .
Note that dr > k(d + c + )I0 is equivalent to that p1 < p2. We also have
Theorem 2.2. Endemic equilibria E1 and E2 do not exist if R0 < p0 . Further, if R0 P p0 , we have
the following:
(i) If dr > k(d + c + )I0, then both E1 = (S1, I1) and E2 = (S2, I2) exist when p1 < R0 < p2 .
(ii) If dr > k(d + c + )I0, then E1 does not exist but E2 exists if R0 P p2 .
(iii) Let dr 6 k(d + c + )I0. Then E1 does not exist. Further, E2 exists when p2 < R0 , and E2 does
not exist when R0 6 p2 .
We consider p0 > 1. If dr P k(d + c + )I0, a typical bifurcation diagram is illustrated in Fig. 1,
where the bifurcation from the disease free equilibrium at R0 ¼ 1 is forward and there is a back-
ward bifurcation from an endemic equilibrium at R0 ¼ 1:5, which gives rise to the existence of

250

200

150

100

50

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
R0

Fig. 1. The figure of infective sizes at equilibria versus R0 when I0 = 40, k = 0.01, c = 0.01,  = 0.01, d = 0.8, r = 1.5,
where (i) of Theorem 2.2 holds. The bifurcation from the disease free equilibrium at R0 ¼ 1 is forward and there is a
backward bifurcation from an endemic equilibrium at R0 ¼ 1:5, which leads to the existence of multiple endemic
equilibria.
 W. Wang / Mathematical Biosciences 201 (2006) 58–71 63

multiple endemic equilibria. Further, if dr < k(d + c + )I0, a typical bifurcation diagram is illus-
trated in Fig. 2, where the bifurcation at R0 ¼ 1 is forward and (1.3) has one unique endemic equi-
librium for all R0 > 1.
Note that a backward bifurcation with endemic equilibria when R0 < 1 is very interesting in
applications. We present the following corollary to give conditions for such a backward bifurca-
tion to occur.

Corollary 2.3. (1.3) has a backward bifurcation with endemic equilibria when R0 < 1 if
dr > k(d + c + )I0 and p0 < 1.

Proof. This corollary is a simple consequence of (i) of Theorem 2.2. h

Example 2.4. Fix I0 = 20, k = 0.01, c = 0.1,  = 0.1, d = 1 and r = 6. Then p1  0.4, p0  0.6667
and dr  k(d + c + )I0 = 5.76. Thus, (1.3) has a backward bifurcation with endemic equilibria
when R0 < 1 in this case (see Fig. 3).
As I0 (the capacity of treatment resources) increases, by the definition we see that p0 increases.
When I0 is so large that p0 > 1, it follows from Theorem 2.2 that there is no backward bifurcation
with endemic equilibria when R0 < 1. If we increase I0 to R0 < p0 , (1.3) does not have a backward
bifurcation because endemic equilibria E1 and E2 do not exist. This means that an insufficient
capacity for treatment is a source of the backward bifurcation.

100

90

80

70

60

I 50

40

30

20

10

0
0 2 4 6 8 10 12 14 16 18 20
R0

Fig. 2. The diagram of I*, I2 versus R0 when I0 = 40, k = 0.2, c = 0.1,  = 0.1, d = 0.8, r = 0.6, where (iii) of Theorem
2.2 holds. The bifurcation at R0 ¼ 1 is forward and (1.3) has a unique endemic equilibrium for R0 > 1.
64 W. Wang / Mathematical Biosciences 201 (2006) 58–71

700

600

500

400

I
300

200

100

0
0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4

R0

Fig. 3. The figure of I*, I1 and I2 versus R0 that shows a backward bifurcation with endemic equilibria when R0 < 1,
where Corollary 2.3 holds.

By analyzing the eigenvalues of the Jacobian matrices of (1.3) at the equilibria, which is similar
to the discussions in [19, Theorem 2.1] or [4, p. 62], we have the following results for the stability
of the endemic equilibria.
Theorem 2.5. E0 is asymptotically stable if R0 < 1 and is unstable if R0 > 1. E* is asymptotically
stable if I* < I0. E1 is a saddle whenever it exists. For E2, we have
(i) E2 is stable if either
kA  3d 2  dðc þ Þ  2d 3 =ðc þ Þ 6 kk ð2:17Þ
or
8
>
> kk < kA  3d 2  dðc þ Þ  2d 3 =ðc þ Þ;
< " sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi!#
1 4kA ð2:18Þ
>
: kk < 2 2kA þ ð2d þ c þ Þðd þ c þ Þ 1  1 þ ðc þ  þ dÞ2 ;
>

(ii) E2 is unstable if
8
>
> kk < kA  3d 2  dðc þ Þ  2d 3 =ðc þ Þ;
< " sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi!#
1 4kA ð2:19Þ
>
: kk < 2 2kA þ ð2d þ c þ Þðd þ c þ Þ 1  1 þ ðc þ  þ dÞ2 .
>
 W. Wang / Mathematical Biosciences 201 (2006) 58–71 65

3. Global analysis and simulations

We begin from the global stability of the disease free equilibrium E0.

Theorem 3.1. The disease free equilibrium E0 is globally stable, i.e., the disease dies out, if one of the
following conditions is satisfied:
(i) R0 < 1 and p0 > 1.
(ii) R0 < 1, p0 < 1 and p1 P 1.
Proof. R0 < 1 implies that E* does not exist. Suppose p0 P 1. It follows from the discussions
for Theorem 2.2 that E1 or E2 exists only if R0 > p0 , which is impossible since we have R0 < 1.
Let us now suppose p0 < 1 and p1 P 1. If dr > k(d + c + )I0, since p1 < p2, it follows from the
discussions for (i), (ii) of Theorem 2.2 that E1 or E2 exists only if R0 > p1 , which is impossi-
ble since we have R0 < 1. If dr 6 k(d + c + )I0, since 1 < p2, it follows from (iii) of Theorem
2.2 that E1 and E2 do not exist. In summary, endemic equilibria do not exist under the
assumptions.
It is easy to verify that positive solutions of (1.3) are ultimately bounded. Note that the
nonnegative S-axis is positively invariant and that the nonnegative I-axis repels positive solutions
of (1.3). Since E0 is asymptotically stable, it follows from the Bendixson Theorem that every
positive solution of (1.3) approaches E0 as t approaches infinity. h
The limit cycles of (1.3) play crucial roles on the structure of dynamical behaviors of the model.
For example, if there is no limit cycle and its endemic equilibrium is unique, the unique endemic
equilibrium is globally stable. For this reason, we adopt Dulac functions to obtain conditions for
the nonexistence of a limit cycle in (1.3). Generally speaking, the approach of Dulac functions
applies to smooth vector fields. However, it is applicable to (1.3) where the vector field defined
by (1.3) is not smooth at the line I = I0. Indeed, denote the right-hand sides of (1.3) by f1 and
f2. Then we have

Lemma 3.2. If there is a continuous function D in int R2þ such that D is continuously differentiate
when I 5 I0, and

oðDf 1 Þ oðDf 2 Þ
þ < 0; when I 6¼ I 0 ;
oS oI

then (1.3) does not have a limit cycle.

Proof. Without loss of generality, we suppose that Q is a limit cycle of (1.3) across the line I = I0
with a period T. We denote its part below I = I0 by Q1 and its part above I = I0 by Q2 (see Fig. 4).
Then we consider auxiliary lines I = I0  g, I = I0 + g, denoted by L1 and L2, respectively, where
g > 0 is sufficiently small. Let G1 denote the region enclosed by Q1 and L1, and let G2 denote the
region enclosed by Q2 and L2. If C1 and C2 be the boundaries of G1 and G2, respectively with the
directions indicated in Fig. 4, then Green’s Theorem implies that
66 W. Wang / Mathematical Biosciences 201 (2006) 58–71

7
Q2
I
6
G2

5
L2

I0 4
L1

G1
2
Q1

0
0 1 2 3 4 5 6 7 8 9 10
S

Fig. 4. The diagram of Q1 and Q2 split from the limit cycle and the diagram of G1 and G2.

Z Z   I
oðDf 1 Þ oðDf 2 Þ
þ dS dI ¼ Dðf1 dI  f2 dSÞ;
oS oI C1
G1
Z Z   I ð3:1Þ
oðDf 1 Þ oðDf 2 Þ
þ dS dI ¼ Dðf1 dI  f2 dSÞ.
oS oI C2
G2

If G0  G1, We have
Z Z  
oðDf 1 Þ oðDf 2 Þ
n :¼ þ dS dI < 0.
oS oI
G0

It follows from (3.1) that

I I
n> Dðf1 dI  f2 dSÞ þ Dðf1 dI  f2 dSÞ. ð3:2Þ
C1 C2

Taking g ! 0 in (3.2), we obtain

I Z T
0>n> Dðf1 dI  f2 dSÞ ¼ Dðf1 f2  f2 f1 Þdt ¼ 0.
Q 0

This precludes the existence of the limit cycle Q. h

Now, we apply Lemma 3.2 to obtain sufficient conditions for the nonexistence of a limit cycle in
(1.3).
 W. Wang / Mathematical Biosciences 201 (2006) 58–71 67

Theorem 3.3. (1.3) does not have a limit cycle if r < d.

Proof. By the first equation of (1.3), it is easy to see that positive solutions of (1.3) eventually
enter and remain in the region
 
A
P :¼ ðS; IÞ : S 6 .
d
Thus, a limit cycle, if it exists, must lie in the region P. Take a Dulac function D = 1/(IS). Then we
have
oðDf 1 Þ oðDf 2 Þ A
þ ¼  2 < 0;
oS oI S I
if 0 < I < I0. If I > I0, it is easy to see that

oðDf 1 Þ oðDf 2 Þ A k 1 I 0A
þ ¼  2 þ 2 ¼ 2 2 ðAI þ kSÞ 6 2 2 ðr  dÞ < 0.
oS oI S I SI S I dS I
Hence, (1.3) does not have a limit cycle. h
Theorem 3.3 implies that there is no limit cycle in (1.3) if the treatment rate is less than the
death rate. We now present a different condition for the nonexistence of limit cycles of (1.3) which
is independent of the treatment rate.
Theorem 3.4. (1.3) does not have a limit cycle if
kA < ð2d þ c þ Þd. ð3:3Þ

Proof. We consider a Dulac function D defined by

(
1
; if I 6 I 0 ;
D ¼ SI1
SI 0
; if I > I 0 .

Then we have
oðDf 1 Þ oðDf 2 Þ A
þ ¼ 2 <0
oS oI S I

if 0 < I < I0. Note that (3.3) implies that



 Að2d 2  kA þ dc þ dÞ
A þ kS  ðd þ c þ ÞS S¼A=d ¼ 
2
< 0.
d2
It follows that for I > I0,

oðDf 1 Þ oðDf 2 Þ 1

þ ¼ 2 A þ kS 2  ðd þ c þ ÞS < 0;
oS oI S I0

where the domain P is considered. Therefore, (1.3) does not have a limit cycle. h
68 W. Wang / Mathematical Biosciences 201 (2006) 58–71

70

60

50

40

I
30

20

10

0
100 120 140 160 180 200 220 240 260 280 300
S

Fig. 5. E* is globally stable when A = 220, I0 = 40, k = 0.01,  = 0.01, c = 0.01, d = 0.01, d = 0.8, r = 1.5.

200

180

160

140

120

I 100

80

60

40

20

0
50 100 150 200 250 300 350 400 450 500
S

Fig. 6. One region of disease persistence and one region of disease extinction where A = 500, I0 = 20, k = 0.01, c = 0.1,
 = 0.1, d = 1, r = 6, where Corollary 2.3 holds.
 W. Wang / Mathematical Biosciences 201 (2006) 58–71 69

140

120

100

80

I
60

40

20

0
100 150 200 250 300 350 400
S

Fig. 7. A bistable case where endemic equilibria E* and E2 are stable when A = 260, I0 = 40, k = 0.01, c = 0.01,
 = 0.01, d = 0.8, r = 1.5, where (i) of Theorem 2.2 holds.

If the conditions of Theorems 2.1, 3.3 or 3.4 are satisfied, then E* is one unique endemic equi-
librium and is globally stable. Numerical calculations show that the unique endemic equilibrium
E* is also globally stable if Theorems 3.3 and 3.4 are not satisfied (see Fig. 5). If R0 > p2 . Theorem
2.2 implies that E2 is a unique endemic equilibrium of (1.3). Numerical calculations show that this
unique endemic equilibrium is also globally stable. However, by direct calculations, we see that
I2 > I* if R0 > p2 This means that we have more infected members in the limited resources for
treatment than those with the unlimited resources for treatment.
Let the conditions of Corollary 2.3 hold. If R0 < 1, the stable manifolds of the saddle E1 split R2þ
into two regions. The disease is persistent in the upper region and dies out in the lower region (see
Fig. 6).
For the case where there are three endemic equilibria, typical dynamical behaviors are the fol-
lowing: the stable manifolds of the saddle E1 split the feasible region into two parts; positive orbits
in the lower part approach to the endemic equilibrium E* and positive orbits in the upper part
approach to the endemic equilibrium E* (see Fig. 7). Thus, we have a bistable case.

4. Concluding remarks

In this paper, we have proposed an epidemic model to simulate the limited resources for
the treatment of patients, which can occur because patients have to be hospitalized but there
70 W. Wang / Mathematical Biosciences 201 (2006) 58–71

are limited beds in hospitals, or there is not enough medicine for treatments. We have shown in
Corollary 2.3 that backward bifurcations occur because of the insufficient capacity for treatment.
We have also shown that (1.3) has bistable endemic equilibria because of the limited resources.
This means that driving the basic reproduction number below 1 is not enough to eradicate the
disease. The level of initial infectious invasion must be lowered to a threshold so that the disease
dies out or approaches a lower endemic steady state for a range of parameters.

Acknowledgments

I am very grateful to two anonymous referees for their careful reading and valuable comments
which led to the improvement of this paper.

References

[1] J. Arino, C.C. McCluskey, P. van den Driessche, Global results for an epidemic model with vaccination that
exhibits backward bifurcation, SIAM J. Appl. Math. 64 (2003) 260.
[2] R.M. Anderson, R.M. May, Infectious Diseases of Humans, Dynamics and Control, Oxford University, Oxford,
1991.
[3] F. Brauer, C. Castillo-Chavez, Mathematical models in population biology and epidemiology, Texts in Applied
Mathematics, 40, Springer, New York, 2001.
[4] N.F. Britton, Essential Mathematical Biology, Springer, London, 2003.
[5] O. Diekmann, J.A.P. Heesterbeek, Mathematical epidemiology of infectious diseases. Model building, analysis
and interpretation, Wiley Series in Mathematical and Computational Biology, John Wiley and Sons, Chichester,
2000.
[6] J. Dushoff, W. Huang, C. Castillo-Chavez, Backwards bifurcations and catastrophe in simple models of fatal
diseases, J. Math. Biol. 36 (1998) 227.
[7] Z. Feng, H.R. Thieme, Recurrent outbreaks of childhood diseases revisited: the impact of isolation, Math. Biosci.
128 (1995) 93.
[8] K.P. Hadeler, P. van den Driessche, Backward bifurcation in epidemic control, Math. Biosci. 146 (1997) 15.
[9] H.W. Hethcote, The mathematics of infectious disease, SIAM Rev. 42 (2000) 599.
[10] J.M. Hyman, J. Li, Modeling the effectiveness of isolation strategies in preventing STD epidemics, SIAM J. Appl.
Math. 58 (1998) 912.
[11] H. Inaba, H. Sekine, A mathematical model for Chagas disease with infection-age-dependent infectivity, Math.
Biosci. 190 (2004) 39.
[12] W.M. Liu, S.A. Levin, Y. Iwasa, Influence of nonlinear incidence rates upon the behavior of SIRS epidemiological
models, J. Math. Biol. 23 (1986) 187.
[13] M. Lizana, J. Rivero, Multiparametric bifurcations for a model in epidemiology, J. Math. Biol. 35 (1996) 21.
[14] M. Martcheva, H.R. Thieme, Progression age enhanced backward bifurcation in an epidemic model with super-
infection, J. Math. Biol. 46 (2003) 385.
[15] S. Ruan, W. Wang, Dynamical behavior of an epidemic model with a nonlinear incidence rate, J. Differ. Equat. 188
(2003) 135.
[16] P. van den Driessche, J. Watmough, A simple SIS epidemic model with a backward bifurcation, J. Math. Biol. 40
(2000) 525.
[17] Y. Takeuchi, Wanbiao Ma, E. Beretta, Global asymptotic properties of a delay SIR epidemic model with finite
incubation times, Nonlinear Anal. 42 (2000) 931.
 W. Wang / Mathematical Biosciences 201 (2006) 58–71 71

[18] W. Wang, Z. Ma, Global dynamics of an epidemic model with time delay, Nonlinear Anal. Real World Appl. 3
(2002) 365.
[19] W. Wang, S. Ruan, Bifurcation in an epidemic model with constant removal rate of the infectives, J. Math. Anal.
Appl. 291 (2004) 775.
[20] L. Wu, Z. Feng, Homoclinic bifurcation in an SIQR model for childhood diseases, J. Differ. Equat. 168 (2000) 150.