Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Chaos, Solitons and Fractals 30 (2006) 10121019 www.elsevier.

com/locate/chaos

Global stability of an SEIR epidemic model with constant immigration q


Guihua Li
a

a,b,c,*

, Wendi Wang b, Zhen Jin

Key Laboratory of Eco-environments in Three Gorges Reservoir Region (Ministry of Education), Faculty of Life Science, Southwest China Normal University, Chongqing 400715, PR China b Department of Mathematics, Southwest China Normal University, Chongqing 400715, PR China c Department of Mathematics, North University of China, Taiyuan Shanxi 030051, PR China Accepted 1 September 2005

Abstract An SEIR epidemic model with the infectious force in the latent (exposed), infected and recovered period is studied. It is assumed that susceptible and exposed individuals have constant immigration rates. The model exhibits a unique endemic state if the fraction p of infectious immigrants is positive. If the basic reproduction number R0 is greater than 1, sucient conditions for the global stability of the endemic equilibrium are obtained by the compound matrix theory. 2005 Elsevier Ltd. All rights reserved.

1. Introduction Many infectious diseases in nature incubate inside the hosts for a period of time before the hosts become infectious. We assume the population can be partitioned into four compartments: susceptible, latent or exposed, infectious and recovered, with sizes denoted by S, E, I and R, respectively. The total population N = S + E + I + R. Using a compartmental approach, one may assume that a susceptible individual rst goes through a latent period (in class E ) after infection before becoming infectious. The resulting models are of SEI, SEIR or SEIRS type, respectively. Liu [1] discussed dynamical behavior of epidemiological models with non-linear incidence rates. Greenhalgh [2] considered SEIR models that incorporate density dependence in the death rate. Cooke and van den Driessche [3] introduced and studied SEIRS models with delays. Greenhalgh [4] studied Hopf bifurcations in models of the SEIRS type with density dependent contact rate and death rate. Li and Muldowney [5] and Li et al. [6] studied the global dynamics of the SEIR models with a non-linear incidence rate and with a standard incidence, respectively. Li et al. [7] analyzed the global dynamics of a SEIR model with vertical transmission and a bilinear incidence. Zhang et al. [8] considered SEIR with saturating contact rate. In [9], Korobeinikov considers the global properties for SEIR and SEIS by means of Lyapunov functions. Li and Jin [10,11] consider the global stability of the SEI and SEIR model with

This work is supported by the National Science Foundation of China (10571143, 10471040) and Youth Science Foundation of Shanxi Province (20041004). * Corresponding author. Address: Department of Mathematics, Southwest China Normal University, Chongqing 400715, PR China. E-mail address: liguihua@nuc.edu.cn (G. Li). 0960-0779/$ - see front matter 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.chaos.2005.09.024

G. Li et al. / Chaos, Solitons and Fractals 30 (2006) 10121019

1013

infectious force in latent and infected period. Research on epidemic models of SEIR with infectious force in both latent, infected and recovered period are scarce in the literature. In the present paper, we consider an SEIR model with infectious force in latent, infected and recovered period and having constant immigration. To establish the global stability of the unique endemic equilibrium, we introduce a change of variables by which our four-dimensional SEIR model can be reduced to a three-dimensional asymptotical autonomous dierential system. This paper is organized as follows. In Section 2, we formulate the model with infectious force in latent, infected and recovered periods and having constant immigration and discuss the existence of equilibrium. In Section 3, we consider the global stability of the endemic equilibrium when p P 0.

2. Model formulation The total host population is partitioned into susceptibles, exposed (in the latent period), infectious, and recovered, with the densities, respectively, denoted by S(t), E(t), I(t) and R(t). The total population size at time t is denoted by N(t), with N = S + E + I + R. The transfer diagram is depicted in the following gure:

The model is described by the following system of dierential equations: 8 0 S 1 pA b10 SE b20 SI b30 SR lS ; > > > < E0 pA b SE b SI b SR c E lE; 10 20 30 0 0 > I c E k I a I l I ; 0 0 > 0 > : 0 R k 0 I lR;

2:1

where the positive parameters l is the rate of natural death. a0 is non-negative constant and denote the rate of diseasecaused death. The parameter c0 denote the transfer rates between the exposed and the infectious. The constant k0 is the rate at which the infectious individuals recover. b10, b20, b30 are the rate of the ecient contact in the latent, infected and recovered period. (1 p)A, pA are constant recruitment of susceptibles, exposed, respectively. Therefore c10 is the mean latent period and k10 is the mean infectious period. Let s lt; b1 b10 ; l b2 b20 ; l b3 b30 ; l c c0 ; l k k0 ; l c A ; l a a0 . l

We obtain the following system analogous to (2.1): 8 dS > > 1 pc b1 SE b2 SI b3 SR S ; > > > ds > > > > > dE > > pc b1 SE b2 SI b3 SR cE E; < ds > dI > > cE kI aI I ; > > > ds > > > > > dR > : kI R. ds

2:2

According to the fact, we assume b2 > b3. The total population size N(t) can be determined by N(t) = S(t) + E(t) + I(t) + R(t) or from the dierential equation dN c N aI . ds

1014

G. Li et al. / Chaos, Solitons and Fractals 30 (2006) 10121019

It is convenient to use E, I, R and N as variables and replace S by N E I R. This gives the model 8 dE > pc b1 E b2 I b3 RN E I R dE; > > > ds > > > > dI > > < c E xI ; ds > dR > > kI R; > > > ds > > > > : dN c N aI ; ds

2:3

where d = 1 + c, x = 1 + k + a. The system (2.3) is equivalent to the system (2.2). This allows us to attack (2.2) by studying the system (2.3). From biological considerations, we study the system (2.3) in the closed set & ' A . : 0 6 E I R 6 N 6 c T E; I ; R; N 2 R4 l It can be veried that T is positively invariant with respect to the system (2.3). If b1, b2, b3 = 0, so that the exposed are those who have entered the population from outside, this reduces to the linear non-homogeneous system 8 dE > > pc dE; > > dt > > > > dI > > < cE xI ; dt 2:3 > > dR kI R; > > > dt > > > > d > : N c N aI . dt Every solution approaches the equilibrium E E0 pc ; d I I0 cpc ; dx R R0 ckpc ; dx N N0 cdx acp . dx

We would expect an equilibrium with b1, b2, b3 > 0 to satisfy E P E0, I P I0, R P R0, N 6 N0. So let the right side of each of the four dierential equations equal to zero in system (2.3), obtaining the equation 8 pc b1 E b2 I b3 RN E I R dE 0; > > > < cE xI 0; 2:4 > kI R 0; > > : c N aI 0. In order to nd equilibria for b1, b2, b3 > 0, we substitute N = c aI into pA b1 E b2 I b3 RN E I R dE 0 to obtain the quadratic equation GI b1 x b2 c b3 k cdxI 2 cdx cb1 x b2 c b3 k cI c2 pc 0. If p = 0, one root is I = 0, and there is a second root I ccb1 x b2 c b3 k c dx ; dxb1 x b2 c b3 k c

which is positive if and only if r cb1 x b2 c b3 k c dx > 0. If p > 0, the quadratic equation G(I) = 0 has one positive and one negative root [12]. The disease-free equilibrium, I = 0, that occurs when p = 0 now become negative (not biologically feasible). The positive root is p cr c r2 4pcdxb1 x b2 c b3 k c I ; 2dxb1 x b2 c b3 k c

G. Li et al. / Chaos, Solitons and Fractals 30 (2006) 10121019

1015

with 8 < 0; r < 0; cr jrj lim I ccb1 x b2 c b3 k c dx p!0 2dxb1 x b2 c b3 k c : dxb1 x b2 c b3 k c

r > 0.

Thus r = 0 is a threshold, there is a basic reproduction number R0 cb1 x b2 c b3 k c cb1 x cb2 c cb3 k c . dx dx dx dx

3. Global stability analysis Theorem 3.1 [11]. If R0 6 1, the disease-free equilibrium P0 when p = 0 is globally asymptotically stable in T. The system (2.3) only has the endemic equilibrium P* = (E *, I *, R*, N*). In the following we will prove the global stability of the equilibrium. Firstly, we prove the local stability of the equilibrium. Theorem 3.2. For the system (2.3) if R0 > 1, the endemic equilibrium P* when p P 0 is local asymptotically stable. Proof. The Jacobian matrix of the system (2.3) at a point P* = (E *, I *, R*, N *) is 0 b1 n d m pc b2 n m pc b3 n m pc n n n B c x 0 B J P B @ 0 k 1 0 a 0 n 1 0 C C C; 0 A 1

dx , n = b1E * + b2I * + b3R*. Its characteristic equation is det(kI J(P*)) = 0, where I is the unit mawhere m xb1 cb 2 k cb3 trix and

N E I R

dE pc ; b1 E b2 I b3 R

x I ; c

R kI .

So the characteristic equation become to (k + 1)(k3 + a1k2 + a2k + a3) = 0, where a1 1 x n d pcb1 mb1 n pcb1 dcb2 k cb3 1xn > 0; xb1 cb2 k cb3 n     pcb1 pcb2 a2 x n d 1 x c n mb1 1 x cmb2 n n x nd x a3 n 1dx k c pcb1 pcb2 dcb2 k 1 xb3 1 x c > 0; xb1 cb2 k cb3 n n

pcb3 pcb1 pcb2 x c mxb1 cb2 k cb3 n n n pcb3 pcb1 pcb2 x c > 0. ndx k c n n n

We calculate easily a1a2 a3 > 0 for b2 > b3. According to Hurwitz criterion, the epidemic equilibrium P* is local asymptotical stability. h To prove the global stability behavior of this equilibrium, we let a = 0 in the system (2.3) and dene the new variables X l S; A Y l E; A Z l I; A R1 l R. A

1016

G. Li et al. / Chaos, Solitons and Fractals 30 (2006) 10121019

Using these change of variables, the system (2.3) becomes 8 dX > > 1 p cb1 XY cb2 XZ cb3 XR1 X ; > > ds > > > > > dY > p cb1 XY cb2 XZ cb3 XR1 cY Y ; < ds > dZ > > cY Z kZ ; > > ds > > > > > : dR1 kZ R1 ; ds with N1(t) = X(t) + Y(t) + Z(t) + R1(t). The equation for the total population N1 is dN 1 1 N 1. ds Since N ! 1 as s ! +1, substituting X = 1 Y Z R1 into the system (3.1), we can obtain limit system 8 dY > > p cb1 Y b2 Z b3 R1 1 Y Z R1 cY Y ; > > > ds > < dZ cY Z kZ ; > ds > > > dR > > : 1 kZ R1 . ds

3:1

3:2

We make the change of variable x = 1 Y Z R1, y = Y, z = Z, then the following system (3.3) is equivalent to (3.2): 8 dx > > 1 p x cb1 y b2 z b3 1 x y zx; > > > ds > < dy 3:3 p cb1 y b2 z b3 1 x y zx cy y ; > ds > > > > > : dz cy z kz; ds with N2(t) = x(t) + y(t) + z(t). The equation for the total population N2 is dN 2 1 N 2 kz. ds It is easy to see that the feasible region is C x; y ; z 2 R3 : x; y ; z P 0; x y z N 2 6 1 . Let E * = (x*, y*, z*) is the unique positive equilibrium of the system (3.3). The Jacobian matrix J of the system (3.3) is 2 3 cb1 b3 x cb2 b3 x a11 6 7 J 4 a21 cb1 b3 x c 1 cb2 b3 x 5; 3:4 0 where a11 1 cb3 x cb1 y b2 z b3 1 x y z; a21 cb3 x cb1 y b2 z b3 1 x y z. In the following, using the geometrical approach of Li and Muldowney in [13], we obtain simple sucient conditions that the disease steady state E* is globally asymptotically stable. At rst, we give a brief outline of this geometrical approach. Let x 7! f x 2 Rn be a C1 function for x in an open set D & Rn . Consider the dierential equation x0 f x. Denote by x(t, x0) the solution to (3.5) such that x(0, x0) = x0. We make the following two assumptions: (H1) There exists a compact absorbing set K & D. (H2) Eq. (3.5) has a unique equilibrium  x in D. 3:5 c 1 k

G. Li et al. / Chaos, Solitons and Fractals 30 (2006) 10121019

1017

The equilibrium  x is said to be globally stable in D if it is locally stable and all trajectories in D converge to  x. For n P 2, by a Bendixson criterion we mean a condition satised by f which precludes the existence of non-constant periodic solutions of (3.5). The classical Bendixsons condition divf(x) < 0 for n = 2 is robust under C1 local perturbations of f. For higher dimensional systems, the C1 robust properties are discussed in [1315]. A point x0 2 D is wandering for (3.5) if there exists a neighborhood U of x0 and T > 0 such that U \ x(t, U ) is empty for all t > T. Thus, for example, all equilibria and limit points are non-wandering. The following global-stability principle is established in Li and Muldowney [13] for autonomous systems in any nite dimension. Theorem 3.3. Suppose that assumptions (H1) and (H2) hold. Assume that (3.5) satises a Bendixson criterion that is robust under C1 local perturbations of f at all non-equilibrium non-wandering points for (3.5). Then  x is globally stable in D provided it is stable. The following criterion is given in [13] and shown to have the robustness required by Theorem 3.3. Let  Bendixson    n n x # P(x) be an matrix-valued function that is C1 for x 2 D. Assume that P1(x) exists and is continuous 2 2 for x 2 K, the compact absorbing set. A quantity  q2 is dened as Z t 1  lBxs; x0 ds; 3:6 q2 lim sup sup t!1 x0 2 K t 0 where B P f P 1 P of 2 1 P ox 3:7

the matrix Pf is obtained by replacing each entry p of P by its derivative   in the direction of f, pijf , and l(B) is the n N , dened by [18, p. 41] Lozinski measure of B with respect to a vector norm j j in R , N 2 jI hBj 1 . lB lim h h!0 2 < 0 rules out the presence of any orbit that gives It is shown in [13] that, if D is simply connected, the condition q rise to a simple closed rectiable curve that is invariant for (3.5), such as periodic orbits, homoclinic orbits, and heteroclinic cycles. Moreover, it is robust under C1 local perturbations of f near any non-equilibrium point that is non-wandering. In particular, the following global-stability result is proved in Li and Muldowney [13]. Theorem 3.4. Assume that D is simply connected and that the assumptions (H1) and (H2) hold. Then the unique equilibrium  x of (3.5) is globally stable in D if  q2 < 0. Now, we study the global stability of the disease steady state E*, and obtain Theorem 3.5. If R0 > 1, b1 > b3 and c + cb1(1 p) < 1, cb3(1 p) < 1 + p, then the disease steady state E* of the system (3.3) is globally asymptotically stable. Proof. The Jacobian matrix J associated with a general solution to (3.3) is (3.4), its second additive compound matrix J[2] is 2 3 b11 cb2 b3 x cb2 b3 x 6 7 J 2 4 c b22 cb1 b3 x 5; 3:8 0 b32 b33 where b11 2 c cb1 x cb1 y b2 z cb3 1 x y z; b22 2 k cb3 x cb1 y b3 z cb3 1 x y z; b32 cb3 x cb1 y b3 z cb3 1 x y z; b33 2 c k cb1 cb3 x. A comprehensive survey on compound matrices and their relations to dierential equations is given in [16]. Set the function

1018

G. Li et al. / Chaos, Solitons and Fractals 30 (2006) 10121019

z 6 P x; y ; z 4 0 0 Then

3 0 0 7 y 0 5. y y

& 0 0 0' z y y ; ; P f P 1 diag z y y


2

f and the matrix B P f P 1 P o P 1 in (3.7) can be written in block form ox

B where

B11 B21

! B12 ; B22

z0 2 c cb1 x cb1 y b2 z cb3 1 x y z; z 2 cy 3 ! cb2 b3 xz 6 7 ; B21 4 cz B12 0 y 5; y z 2 0 y 2 k cb1 x cb1 y b2 z b3 1 x y z 6y B22 6 4 c B11
n

cb1 b3 x y 2kc y
0

3 7 7. 5

Let (u, v, w) denotes the vectors in R3 R2 , we select a norm in R3 as j(u, v, w)j = max{juj,jvj + jwj} and let l denotes the Lozinsk measure with respect to this norm. Following the method in [17], we have the estimate l(B) 6 sup{g1, g2}, where g1 l1 B11 jB12 j; g2 jB21 j l1 B22 .

jB12j, jB21j are matrix norms with respect to the l1 vector norm, and l1 denotes the Lozinsk measure with respect to the 0 l1 norm, see [18, p. 41]. More specically, l1 B11 zz 2 c cb1 x cb1 y b2 z cb3 1 x y z, b3 xz jB12 j cb2 , jB21 j czy . To calculate l1(B22), add the absolute value of the off-diagonal elements to the diagonal y one in each column of B22, and then take the maximum of two sums, see [18, p. 41]. Since b1 > b3, we have & 0 ' y y0 l1 B22 max 2 k c cb1 x cb1 y b2 z b3 1 x y z; 2 k c cb1 b3 x y y 0 y 6 2 k c cb1 x. y From the system (3.3), we have x < 1 p; y0 p cb2 b3 xz cb3 1 xx 1 c; cb1 x cb3 x y y y y z0 c y 1 k. z z So we have y 0 z0 p cb 1 xx 1 cb3 x cb1 y b2 z cb3 1 x y z 3 y z y y y 0 z0 < 1 p cb3 1 p; y z y 0 z0 g2 < 1 c cb1 x y z y 0 z0 < 1 c cb1 1 p. y z g1

G. Li et al. / Chaos, Solitons and Fractals 30 (2006) 10121019

1019

We can choose t1 large enough such that y 0 z0 1 p cb3 1 p; y z 0 y z0 g 2 < 1 c c b 1 1 p y z g1 < for t P t1, where d can be chosen arbitrarily small. Therefore, lB 6 y 0 z0  b y z

b minf1 p cb3 1 p; 1 c cb1 1 pg 1 c cb1 1 p > 0 is a constant. Along each for t P t1, where  solution (x(t), y(t), z(t)) of (3.3) with (x(0), y(0), z(0)) 2 C, where C is the compact absorbing set, we have Z Z Z Z 1 t 1 t1 1 t 1 t1 1 y t 1 zt  log b; lB ds lB ds lB ds 6 lB ds log t 0 t 0 t t1 t 0 t y t 1 t zt 1 which implies that  q2 6  b=2 < 0 from (3.6). This complete the proof. h

4. Conclusion In the paper, we discuss the SEIR model with constant immigration and infectious force in the latent and recovered period, too. We derive a basic reproduction number R0 and that it determines the global dynamics of (2.1); if R0 > 1, a unique endemic equilibrium P* is globally asymptotically stable in the interior of the feasible region so that the disease persists at the endemic equilibrium level if it is initially present. In the limiting case when c0 ! 1, the latent period is negligible, and the model (2.1) reduces to an SIR model with bilinear incidence. When c0 = 0, there is no recovery from the disease, and (2.1) reduces to an SEI model. When p = 0, (2.1) becomes an SEIR with constant immigration only susceptibles.

References
[1] Liu W-M, Hethcote HW, Levin SA. Dynamical behavior of epidemiological models with nonlinear incidence rate. J Math Biol 1987;25:35980. [2] Greenhalgh D. Some results for a SEIR epidemic model with density dependence in the death rate. IMA J Math Appl Med Biol 1992;9:67. [3] Cooke K, van den Driessche P. Analysis of an SEIRS epidemic model with two delays. J Math Biol 1996;35:240. [4] Greenhalgh D. Hopf bifurcation in epidemic models with a latent period and nonpermanent immunity. Math Comput Model 1997;25:85. [5] Li MY, Muldowney JS. Global stability for the SEIR model in epidemiology. Math Biosci 1995;125:15564. [6] Li MY, Graef JR, Wang LC, Karsai J. Global dynamics of a SEIR model with a varying total population size. Math Biosci 1999;160:191213. [7] Li MY, Smith HL, Wang L. Global dynamics of an SEIR epidemic model with vertical transmission. SIAM J Appl Math 2001;62:58. [8] Zhang J et al. Global dynamic of SEIR epidemic model with saturating contact rate. Math Biosci 2003;185:1532. [9] Korobeinikov A. Lyapunov functions and global properties for SEIR and SEIS epidemic models. Math Med Biol 2004;21:7583. [10] Li G, Jin Z. Global stability of an SEI epidemic model with general contact rate. Chaos, Solitons & Fractals 2005;23:9971004. [11] Li G, Jin Z. Global stability of an SEIR epidemic model with infectious force in latent infected and immune period. Chaos, Solitons & Fractals 2005;25:117784. [12] Brauer F, Van den Driessche P. Models for transmission of disease with immigration of infectives. Math Biosci 2001;171:14354. [13] Li MY, Muldowney JS. A geometric approach to the global-stability problems. SIAM J Math Anal 1996;27:107083. [14] Li MY. Dulac criteria for autonomous systems having an invariant ane manifold. J Math Anal Appl 1996;199:37490. [15] Li MY, Muldowney JS. On RA. Smiths autonomous convergence theorem. Rocky Mount J Math 1995;25:36579. [16] Muldowney JS. Compound matrices and ordinary dierential equations. Rocky Mount J Math 1990;20:85772. [17] Martin Jr RH. Logarithmic norms and projections applied to linear dierential systems. J Math Anal Appl 1974;45:43254. [18] Coppel WA. Stability and asymptotic behavior of dierential equation. Boston: Heath; 1965.

You might also like