Applied Mathematics and Mechanics (English Edition), 2006, 27(5):667672

c
Editorial
Committee of Appl. Math. Mech., ISSN 0253-4827

QUALITATIVE ANALYSIS OF AN SEIS EPIDEMIC MODEL

WITH NONLINEAR INCIDENCE RATE
WANG La-di (

)1,2 , LI Jian-quan (

)3

(1. Department of Applied Mathematics, Shanxi University of Finance and Economics,

Taiyuan 030006, P. R. China;
2. Department of Mathematics, Shanghai University, Shanghai 200444, P. R. China;
3. Department of Applied Mathematics and Physics, Air Force Engineering University,
Xian 710051, P. R. China)
(Communicated by LIU Zeng-rong)

Abstract: By means of limit theory and Fondas theorem, an SEIS epidemic model with
constant recruitment and the disease-related rate is considered. The incidence term is of
the nonlinear form, and the basic reproduction number is found. If the basic reproduction
number is less than one, there exists only the disease-free equilibrium, which is globally
asymptotically stable, and the disease dies out eventually. If the basic reproduction
number is greater than one, besides the unstable disease-free equilibrium, there exists
also a unique endemic equilibrium, which is locally asymptotically stable, and the disease
is uniformly persistent.
Key words: epidemic model; equilibrium; stability; persistence
Chinese Library Classification: O175.12
2000 Mathematics Subject Classification: 34D23; 92D30
Digital Object Identifier(DOI): 10.1007/s 10483-006-0513-z

Introduction
For classical epidemic models, one attends mainly to the existence, local and global stability
of equilibrium. Recently, since some more general incidence rates are introduced such that the
corresponding epidemic models are of complex dynamic behavior, it is sometimes dicult to
do this. However, in the epidemiological sense, studying the persistence of the disease has the
same signicance. The persistence of the disease has been considered by some researchers[13] .
Bilinear and standard incidence rates have been frequently used in many classical epidemic
models[4, 5] . Simple dynamics of these models seem related to such functions. Several dierent
incidence rates have been proposed by researchers. Let S(t) be the number of susceptible
individuals, E(t) be the number of exposed (latent) individuals, and I(t) be the number of
infective individuals at time t. After a study of the cholera epidemic spread in Bari 1973,
Capasso and Serio[6] introduced a saturated incidence rate g(I)S into epidemic models. This is
important because the number of eective contacts between infective individuals and susceptible
individuals may saturate at high infective levels due to crowding of infective individuals or due
to the protection measures by the susceptible individuals. Nonlinear incidence rates of the
form I p S q were investigated in Refs.[7,8]. The incidence rate of the form I p S/(1 + aI q )
was discussed in Refs.[8,9]. A very general form of nonlinear incidence rate was considered in
Ref.[10].
Received Jul.31, 2004; Revised Feb.10, 2006
Project supported by the National Key Technologies R & D Program of China (No.2004BA719A01)
Corresponding author LI Jian-quan, Professor, Doctor, E-mail: jianq [email protected]

668

WANG La-di and LI Jian-quan

In this paper, we consider the following SEIS model:

dS
SI

= A S
+ I,

dt
(I)

dE
SI
=
( + )E,

dt
(I)

dI = E ( + + )I,
dt

(1)

where A is the recruitment rate of the population, is the natural death rate,  is the transfer
rate from the exposed (latent) group to the infection group, is the transfer rate from the
infectious group to the susceptible group, is the disease-related death rate. The incidence
rate in this paper is SI/(I), which generalizes the incidence rate I p S/(1 + aI q ) and is a
bilinear incidence rate as (I) 1. We assume that function (I) satises (0) = 1 and
 (I) 0, which implies (I) 1 for I 0. 1/(I) may measure the the inhibition eect from
the behavior change of the susceptible individuals when the number of the infectious individuals
increases, which implies that the contact number between susceptible individuals and infectious
individuals decreases as the number of the infectious individuals increases. In this sense this
incidence rate seems more reasonable.
Our paper is organized as follows. In Section 1, the existence and stability of equilibria are
investigated. In Section 2, we show that the system (1) is persistent as the endemic equilibrium
exists. A brief summary and conclusion is given in Section 3.

Existence and Stability of Equilibria

To be concise in notations, let n = + , m = + + , then the system (1) becomes

dS
SI

= A S
+ I,

dt
(I)

dE
SI
(2)
=
nE,

dt
(I)

dI = E mI.
dt

Summing up the three equations in the system (2) gives

d
(S + E + I) = A (S + E + I) I.
dt
Therefore, from biological considerations, we study the system (2) in the closed set =
3
3
{(S, E, I) R+
: 0 < S + E + I A/}, where R+
denotes the non-negative cone of R3
including its lower dimensional faces. It is easy to see that is positively invariant with respect
to the system (2).
Let R0 = A/(mn). R0 will be called the basic reproduction number, which is the number
of secondary infectious cases produced by an infectious individual during his or her eective
infectious period when introduced in a population of susceptible.
It is obvious that the system (2) always has the disease-free equilibrium (trivial equilibrium)
P0 (A/, 0, 0). About P0 we have
Theorem 1.1 The disease-free equilibrium P0 of the system (2) is globally asymptotically
stable in if R0 < 1 and it is unstable if R0 > 1.
In order to prove the global stability of the disease-free equilibrium P0 , we introduce the
following lemma.

Qualitative Analysis of SEIS Epidemic Model

Lemma 1.2[3]

669

For a bounded real-valued function f on [0, ), dene

f = lim inf f (t),
t

f = lim sup f (t).

t

Let f : [0, ) R be bounded and twice dierentiable with bounded second derivative. Let
tk and f (tk ) converge to f or f for k . Then limk f  (tk ) = 0.
Proof of Theorem 1.1 It is easy to see that = is an eigenvalue of J(P0 ), which is
the Jacobian matrix of the system (2) at disease-free equilibrium P0 . The other eigenvalues of
J(P0 ) are determined by equation ( + n)( + m) A/ = 0. It is easy to know that all of its
roots are with negative real parts if and only if mn A/ > 0, i.e., R0 < 1. So P0 is locally
asymptotically stable if R0 < 1 and unstable if R0 > 1.
By Lemma 1.2 we may choose a sequence tk (k ) such that E(tk ) E ,
dE(tk )/dtk 0, and another sequence k (k ) such that I(k ) I , dI(k )/dtk 0.
From the middle equation of the system (2) we have
E =

A
S(tk )I(tk )

lim
I ,
n k [I(tk )]
n

(3)

where (I) 1 and S A/ are used.

From the last equation of the system (2) we have
I =



lim E(k )
E.
k
m
m

(4)

From the inequalities (3) and (4) we have

E

A
E = R0 E ,
mn

A
I = R0 I .
mn

These imply that E 0 and I 0 if R0 < 1. However, E 0 and I 0, then we

have that E = E = 0, and I = I = 0, that is, limt E(t) = 0 and limt I(t) = 0.
Further, limt S(t) = A/.
According to the local stability of P0 , we can know that the disease-free equilibrium P0 is
globally asymptotically stable if R0 < 1.
Theorem 1.1 is proved completely.
Theorem 1.3 If R0 > 1, the system (2) has a unique endemic equilibrium(positive
equilibrium) P (S , E , I ) in the interior of , which is locally asymptotically stable, where
S = mn(I )/(),

E = mI /,

and I is the unique root of the equation (mn/ ) I + mn(I)/() = A in the interval
(0, A/).
Proof The coordinates of the endemic equilibrium P (S , E , I ) of the system (2) are
the positive solution of the equations

SI

A S
+ I = 0,

(I)

SI
(5)
nE = 0,

(I)

E mI = 0
in the interior of .

670

WANG La-di and LI Jian-quan

From the last equation of Eqs.(5) we have E = mI/. Substituting E = mI/ into the
middle equation of Eqs.(5) can give S = mn(I)/(), since the root I = 0 is not considered
here. Further, from the rst equation of Eq.(5) we have


mn
 mn
I +
(I) A = 0.
(6)
F (I) =


Notice that mn/ = ( + + )/ + ( + ) > 0, then the function F (I) is increasing
since  (I) 0. It is obvious that F (A/) > 0, again F (0) = mn/() A (here (0) = 1 is
used). Therefore, Eq.(6) has a unique root, I , in the interval (0, A/) if and only if F (0) < 0,
that is, R0 > 1.
Substituting I = I into E = mI/ and S = mn(I )/() gives E and S .
From F (A/( + )) > 0 we know I < A/( + ), then 0 < S +E = [A ( + )I ]/ <
A/. So point P (S , E , I ) is in the interior of .
The existence of the endemic equilibrium P is proved.
Using equalities I S /(I ) = nE , S /(I ) = mn/ and E = mI / gives the Jacobian matrix of the system (2) at the endemic equilibrium P ,



mn
I  (I )
nE
0
1

S

(I )


.

(I
)
nE
mn
I
J(P ) =

n
1

)
S

(I

0
Its characteristic equation is

3 + a1 2 + a2 + a3 = 0,

where
nE
> 0,
S


nE
I  (I )
+
(m
+
n)

+
a2 = mn
> 0,
(I )
S
nE
I  (I )
.
a3 = (mn ) + mn
S
(I )

a1 = + m + n +

Further,



nE
I  (I )
a1 a2 a3 = m + n +
+
(m
+
n)
+ (m + n)2
mn
S
(I )



nE
nE
+ m + n + (m + n) mn +  .
+
S
S
Notice that m = + + and n = + , then a3 > 0 and a1 a2 a3 > 0. Therefore, the
Routh-Hurwitz stability conditions are satised. It follows that the endemic equilibrium P is
locally asymptotically stable.
This completes the proof of Theorem 1.3.

Persistence

First, let us recall some denitions in order to introduce the abstract theorem. Assume X is
a locally compact metric space with metric d and let F be a closed subset of X with boundary
F and interior intF . Let be a semi-dynamical system dened on F .

Qualitative Analysis of SEIS Epidemic Model

671

We say that is persistent if for all u intF , limt+ infd((u, t), F ) > 0, and that is
uniformly persistent if there is > 0 such that for all u intF , limt+ infd((u, t), F ) > .
In Ref.[11], Fonda gives a result about persistence in terms of repellers. A subset of
F is said to be a uniform repeller if there is a > 0 such that, for each u F \ ,
limt+ infd((u, t), ) > . A semiow on a closed subset F of a locally compact metric
space is uniformly persistent if the boundary of F is repelling. The result of Fonda is the
following.
Lemma 2.1[11] Let be a compact subset of X such that X \ is positively invariant. A necessary and sucient condition for to be a uniform repeller is that there exists a
neighborhood U of and a continuous function P : X R+ satisfying
(i) P (u) = 0 if and only if u .
(ii) For all u U \ there is a Tu > 0 such that P ((u, Tu )) > P (u).
For any u0 = (S0 , E0 , I0 ) there is a unique solution (u0 , t) = (S, E, I)(t; u0 ) of the
system (2), which is dened in R+ and satises (u0 , 0) = (S0 , E0 , I0 ). Since is a positively
invariant set of the system (2), then (u0 , t) for t R+ and is a semi-dynamical system in
.
We are now in position to prove that = {(S, E, I) : I = 0} is a uniform repeller,
which implies that the semi-dynamic system is uniformly persistent.
Theorem 2.2
persistent in .

If R0 > 1, then the set is a uniform repeller and hence is uniformly

Proof It is easy to show that I(t) > 0 for t > 0 if I(0) > 0, so \ is positively
invariant. Again the set is a compact subset of . Let P : R+ be dened by
P (S, E, I) = I and let U = {(S, E, I) : P (S, E, I) < }, where > 0 is small enough
so that R0 /[()( + 2)] > 1, since R0 > 1 and (0) = 1.
E,
I))
such that for each t > 0 we have P ((
Assume that there is u
U (
u = (S,
u, t)) <
P (
u) < , which implies that I(t; u) < for t > 0. From the rst equation of the system (2)
) A/( + ). So there is a suciently
we have dS/dt A S S, then limt infS(t; u
large T > 0 such that S(t; u
) > A/( + 2) for t T.
Dene the auxiliary function V (t) = I(t) + (1 )E(t)/n, where (0 < < 1) is a
suciently small constant so that R0 (1 )/[()( + 2)] > 1. Calculating directly gives
the derivative of V (t) along (
u, t) as follows:


(1 )S(t)
dV (t)
=
m I +  E,
dt
n(I(t))
then for t T we have



A(1 )
dV (t)

m I +  E
dt
n()( + 2)


R0 (1 )
=m
1 I +  E,
()( + 2)

where (I) () is used since I < and  (I) 0.

Let
 


R0 (1 )
n
= min m
> 0,
1 ,
()( + 2)
1
then

dV (t)
V (t).
dt
The last inequality implies that V (t) as t . However, V (t) is bounded on the set ,
so the assumption above is not true.

672

WANG La-di and LI Jian-quan

We have proved that for each u \ , with u belonging to a suitably small neighborhood
of , there is some Tu such that P ((u, Tu )) > P (u). Therefore, Lemma 2.1 allows us to reach
to the conclusion of Theorem 2.2.
The proof of Theorem 2.2 is complete.

Summary and Conclusion

This paper presents a mathematical study on the dynamical behavior of an SEIS epidemic
model that incorporates constant recruitment, exponential natural death as well as the diseaserelated death, so that the population size may vary in time. The incidence rate is nonlinear.
For the system (1) we nd the basic reproduction number, R0 = A/ /( + + )
/(d + ). It completely determines the dynamical behavior of the system (1) in the feasible
region . If R0 < 1, the disease-free equilibrium is globally asymptotically stable in and the
disease always dies out eventually. If R0 > 1, there is a unique endemic equilibrium which is
locally asymptotically stable and the disease persists if it is initially present.
Acknowledgements The authors would like to acknowledge the suppor provided by the
Fund of Scientic Research of Air Force Engineering University.

References
[1] Wang W, Ma Z. Global dynamics of an epidemic model with delay[J]. Nonlinear Analysis: Real
World Applications, 2002, 3(3):809834.
[2] Wang W. Global behavior of an SEIRS epidemic model time delays[J]. Applied Mathematics
Letters, 2002, 15(2):423428.
[3] Thieme R H. Persistence under relaxed point-dissipativity(with applications to an endemic
model)[J]. SIAM Journal of Mathematical Analysis, 1993, 24(2):407435.
[4] Hethcote H W. The mathematics of infectious diseases[J].SIAM Review, 2000, 42(3):599653.
[5] Li J, Ma Z. Qualitative analyses of SIS epidemic model with vaccination and varying total population size[J]. Mathematical and Computer Modelling, 2002, 20(5):12351243.
[6] Capasso V, Serrio G. A generalization of the Kermack-Mckendrick deterministic epidemic
model[J]. Mathematical Biosicences, 1978, 42(2):327346.
[7] Liu W M, Hethcote H W, Levin S A. Dynamical behavior of epidemiological model with nonlinear
incidence rates[J]. Journal of Mathematical Biology, 1987, 25(2):359380.
[8] Liu W M, Levin S A, Iwasa Y. Influence of nonlinear incidence rates upon the behavior of SIRS
epidemiological models[J]. Journal of Mathematical Biology, 1986, 23(1):187204.
[9] Ruan S, Wang W. Dynamical behavior of an epidemic model with a nonlinear incidence rate[J].
Journal of Dierential Equations, 2003, 188(1):135163.
[10] Derrick W R, van den Driessche P. A disease transmission model in a nonconstant population[J].
Journal of Mathematical Biology, 1993, 31(3):495512.
[11] Fonda A. Uniformly persistent semidynamical systems[J]. Proceedings of American Mathematical
Society, 1988, 104(1):111116.