On the global stability of the endemic state in an epidemic model with vaccination

Abstract

This paper investigates an SIS epidemic model with variable population size including a vaccination program. Dynamics of the endemic equilibrium of the model are obtained, and it will be shown that this equilibrium exists and is locally asymptotically stable when R0>1 1$$\end{document}]]> . In this case, the disease uniformly persists, and moreover, using a geometric approach we conclude that the model is globally asymptotically stable under some conditions. Also, a numerical discussion is given to verify the theoretical results.

Introduction

The susceptible–infected–susceptible (SIS) model is one of the most well-known type of epidemic models. These models are appropriate for some infections, for instance, common cold and influenza, or bacterial diseases such as meningitis and cholera, or sexually transmitted diseases, that do not cause permanent immunity after recovery. To immunize individuals from infection and control the infectious diseases, vaccination is usually preferred because of its efficiency compared with other drug and non-drug interventions. There are many epidemic models [ 1 , 3 , 8 , 13 , 14 , 16 , 19 ] and also some SIS epidemic models in which temporary or permanent vaccination has been included [ 6 , 9 , 10 – 11 , 17 , 22 ]. These models may be deterministic [ 10 , 11 ] or stochastic [ 4 , 6 , 22 ], with constant [ 10 , 13 ] or variable [ 9 ] population size, and with general incidence [ 7 , 21 ] or a particular incidence such as standard, bilinear, and saturated [ 2 , 5 , 13 ]. In this paper, we introduce an SIS epidemic model with vaccination and standard incidence. In the next section, we first describe the model and then some basic properties of it will be given. Sections “ Local asymptotical stability ” and “ Global asymptotical stability ” are devoted to investigating the local and global stability of endemic equilibrium of the model, respectively. Finally, we summarize the conclusions in section “ Conclusions ”, after a numerical consideration in section “ Numerical examples ”.

Model description

Consider the following system:

dSdt=(1-ϑ)Λ-βSIN-(δ+σ)S+ηI+ξV,dIdt=βSIN-(δ+η+α)I,dVdt=ϑΛ+σS-(δ+ξ)V.

Here, the population has been divided into three subpopulations as susceptible, infected, and vaccinated individuals. The size of population in each class at time t is denoted by S , I , and V , respectively, whereas the number of all individuals in this time is given by N=S+I+V . The recruiting is done by entering a number of Λ individuals into population per unit time that may be either immigrants or newborns. The vaccination program is applied on the new members by a proportion ϑ . A proportion σ of susceptible individuals are also vaccinated, and the rate of losing immunity of vaccination is ξ . The vaccine is supposed to be completely effective, and no vaccinated individual becomes infected. Susceptible individuals become infected at standard incidence rate βSIN, where β is transmission coefficient, whereas the rate of recovery is η . The rate of natural death in population is δ , and the mortality due to disease is also included in the model with a rate α . All values in the model are nonnegative except Λ and δ that are assumed positive.

From system ( 1 ), it can be seen that the total number of members N is not fixed in general and it is expressed by the following equation:

dNdt=Λ-δN-αI.

We have dNdt≤Λ-δN , and thus, lim supN(t)t→∞≤Λδ . This also states that I , S , and V are bounded. We can see that the feasible region

Γ=(I,S,V)∈R+3:I+S+V≤Λδ
is a positively invariant set for system ( 1 ), and the system is well posed for mathematical and epidemiological considerations in Γ .

The equilibria of the model ( 1 ) are obtained by solving the following equations:

(1-ϑ)Λ-βS¯I¯N¯-(δ+σ)S¯+ηI¯+ξV¯=0,βS¯N¯-(δ+η+α)I¯=0,ϑΛ+σS¯-(δ+ξ)V¯=0.
The system has two solutions: When I¯=0 , the disease-free equilibrium (DFE) E0
E0=(I0,S0,V0)=0,Λ(δ(1-ϑ)+ξ)δ(δ+ξ+σ),Λ(δϑ+σ)δ(δ+ξ+σ),
and when I¯>00$$\end{document}]]> , the endemic equilibrium E∗=(I∗,S∗,V∗) , where
I∗=Λ(R0-1)δR~0+α(R~0-1),S∗=Λ(δ+η+α)βδ1-α(R0-1)δR~0+α(R~0-1),V∗=ϑΛδ+ξ1+σ(δ+αϑ+ξ)/ϑ(δ+σ+ξ)δR~0+α(R~0-1).

Here,

R0=β(δ(1-ϑ)+ξ)(δ+η+α)(δ+ξ+σ),R~0=β(δ+ξ)(δ+η+α)(δ+ξ+σ),
and thus, we have
R~0-R0=βδϑ(δ+η+α)(δ+ξ+σ).

The quantity R0 is called the basic reproduction number of the model. Furthermore, the total population sizes at two equilibria E0 and E∗ are obtained, respectively, as follows:

N0=Λδ,
and
N∗=Λδ1-α(R0-1)δR~0+α(R~0-1).

In the following sections, we investigate dynamics of the model ( 1 ) at the endemic equilibrium E∗ and its asymptotic stability will be obtained. The next section is devoted to consider the local asymptotic stability.

Local asymptotical stability

Letting F=βSIN , the Jacobian matrix of the model ( 1 ) at E∗ has the following form:

J∗=a1-(δ+η+α)a2a3-a1+η-a2-(δ+σ)-a3+ξ0σ-(δ+ξ),
in which
a1=∂F∂I|E∗=βS∗N∗2(N∗-I∗),a2=∂F∂S|E∗=βI∗N∗2(N∗-I∗),a3=∂F∂V|E∗=-βS∗I∗N∗2.
The characteristic equation of matrix J∗ is
p(λ)=λ3+b1λ2+b2λ+b3,
where
b1=-tr(J∗)=-a1+a2+(δ+η+α)+(2δ+σ+ξ),b2=-12(tr(J∗2)-tr2(J∗))=-a1(2δ+σ+ξ)+a2(2δ+α+ξ)+a3σ+δ(δ+σ+ξ)+(δ+η+α)(2δ+σ+ξ),b3=-det(J∗)=-a1δ(δ+σ+ξ)+a2(δ+α)(δ+ξ)+a3σ(δ+α)+δ(δ+η+α)(δ+σ+ξ).

We have

a1=(δ+η+α)1-δ(R0-1)δR~0+α(R~0-R0),a2=δ(β-(δ+η+α))R0-1δR~0+α(R~0-R0),a3=-δ(δ+η+α)R0-1δR~0+α(R~0-R0).

Obviously a3<0 and a1=(δ+η+α)(δ+α)(R~0-R0)+δδR~0+α(R~0-R0)>00$$\end{document}]]> . Moreover, notice that the equilibrium E∗ exists if R0>1 1$$\end{document}]]> and this implies

β>βδ(1-ϑ)+ξδ+σ+ξ>(δ+η+α), \beta \left( \frac{\delta (1-\vartheta )+\xi }{\delta +\sigma +\xi }\right) >(\delta +\eta +\alpha ), \end{aligned}$$\end{document}]]>
and thus, a2>00$$\end{document}]]> .

Using ( 8 ), we obtain

In preceding relations, we got b1>00$$\end{document}]]> , b2>00$$\end{document}]]> , b3>00$$\end{document}]]> and b1b2-b3>00$$\end{document}]]> ; thus, by using Routh–Hurwitz criterion the real part of all eigenvalues of the Jacobian matrix J∗ must be negative. Therefore, the following theorem has been proven:

Theorem 1

For epidemic model ( 1 ), the endemic state E∗ exists and is stable if R0>11$$\end{document}]]> .

Corollary 1

The endemic state E∗ is stable if parameter values ϑ and σ lie under the following line in the ϑ - σ plane:

(βδ)ϑ+(δ+η+α)σ=(β-(δ+η+α))(δ+ξ).

Corollary  1 gives minimum amount of vaccination proportions that is needed to the vaccine be effective. Indeed, those values of ϑ and σ that lie above the mentioned line in ( 9 ) are sufficient to the disease eradicated from population. When the proportion of vaccination in new members and in susceptibles is equal, i.e., τ=ϑ=σ , this optimum amount of vaccination will be as

τ∗=(β-(δ+η+α))(δ+ξ)(δ+η+α)+βδ.

Global asymptotical stability

The global asymptotical stability of the endemic state E∗ is discussed in this section. The analysis of the global stability of epidemic models is generally a difficult task, and the methods introduced for carrying this out are scant. To analyze the global asymptotic stability of the endemic state, we employ a geometric method developed by Li and Muldowney [ 12 ] and used by many authors [ 2 , 15 , 18 , 19 – 20 , 23 ]. The following result has been proved in [ 12 ].

Theorem 2

Assume that f is a C1 function on a simply connected open set D⊂Rn , the system x′=f(x) has a unique equilibrium x¯ in D , and there exists a compact absorbing set Θ⊂D . Then, x¯ is globally stable in D if

lim supt→∞supx0∈Θ1t∫0tΨ(Υ(x(r,x0)))dr<0,
in which Υ=ΣfΣ-1+ΣJ[2]Σ-1 . Here , Σ is an n2×n2 matrix-valued function, J[2] is the second additive compound matrix of the Jacobian matrix J , Σf is obtained by (σij)f=∂σij∂x⊤.f(x) , and Ψ(Υ) is the Lozinski ı˘ measure of Υ with respect to a vector norm |.| in Rm , with m=n2 , defined by Ψ(Υ)=limh→0+|I+hΥ|-1h .

The global stability of the endemic state is expressed in the next theorem.

Theorem 3

When R0>11$$\end{document}]]> and δ+σ>ξ\xi$$\end{document}]]> , the endemic equilibrium E∗ is globally asymptotic stable if δ+ξ>max{σ,α}+η \max \{\sigma , \alpha \}+ \eta$$\end{document}]]> .

Proof

Considering the feasible region Γ , we see that interior and boundary of Γ are Γ∘={(I,S,V)∈Γ:I>0}0 \}$$\end{document}]]> and ∂Γ=Γ\Γ∘={(I,S,V)∈Γ:I=0} , respectively. When R0>11$$\end{document}]]> , there exists an unique endemic state E∗ , and moreover, the disease-free equilibrium (DFE) E0∈∂Γ is unstable. Now, using Acyclicity Theorem similar to the proof of Theorem 3.2 in [ 18 ] we find that when R0>11$$\end{document}]]> , system ( 1 ) is uniformly persistent. On the other hand, we know that the solutions are bounded because the total population size N is bounded. Considering this with the uniform persistence of the system, it can be concluded that there exists a compact absorbing set Θ⊂Γ .

The Jacobian matrix of the system at E∗ was given by ( 6 ), and thus, its second additive compound matrix is

J∗[2]=a1-a2-n1-a3+ξ-a3σa1-n2a20-a1+η-a2-n3,
where
n1=2δ+η+α+σ,n2=2δ+η+α+ξ,n3=2δ+σ+ξ.

Matrix Σ in Theorem  2 acts as a Lyapunov function. So, for the method to be applicable and efficient, this matrix must be chosen suitable. Now, we choose matrix Σ as Σ=SII3 , where I3 is the 3×3 identity matrix. Then, Σ-1=ISI3 , Σf=dSdtI-SdIdtI2I3 and ΣfΣ-1=dSdtS-dIdtII3 .

Therefore,

Υ=ΣfΣ-1+ΣJ∗[2]Σ-1=dSdtS-dIdtI+a1-a2-n1-a3+ξ-a3σdSdtS-dIdtI+a1-n2a20-a1+ηdSdtS-dIdtI-a2-n3.

The matrix Υ can be written in the block form as Υ=Υ11Υ12Υ21Υ22, in which

Υ11=dSdtS-dIdtI+a1-a2-n1,Υ12=(-a3+ξ,-a3),Υ21=σ0,Υ22=dSdtS-dIdtI+a1-n2a2-a1+ηdSdtS-dIdtI-a2-n3.

Considering the vector norm |(μ1,μ2,μ2)|=max{|μ1|,|μ2|+|μ3|} for all (μ1,μ2,μ3)∈R3 as in [ 12 ], we have

Ψ(Υ)≤sup{Ω1,Ω2},
where
Ω1=Ψ(Υ11)+|Υ12|andΩ2=ψ(Υ22)+|Υ21|.

Thus, we obtain

Ψ(Υ11)=dSdtS-dIdtI+a1-a2-n1,|Υ12|=-a3+ξ,|Υ21|=σ
and
Ψ(Υ22)=maxdSdtS-dIdtI+η-n2,dSdtS-dIdtI-n3=maxdSdtS-dIdtI-(2δ+α+ξ),dSdtS-dIdtI-(2δ+σ+ξ)=dSdtS-dIdtI-(2δ+ξ)-min{α,σ}.

From the second equation in system ( 1 )

dIdtI=βSN-(δ+η+α).

Therefore, we have

Ω1=dSdtS-dIdtI+a1-a2-n1-a3+ξ=dSdtS-βIN1-SN-(δ+σ-ξ)≤dSdtS-(δ+σ-ξ),
and
Ω2=dSdtS-dIdtI-(2δ+ξ)-min{α,σ}+σ≤dSdtS-(min{α,σ}-(η+σ+α)+(δ+ξ)).

Hence, assuming

d=min{δ+σ-ξ,(δ+ξ)-(max{σ,α}+η)}>0,0, \end{aligned}$$\end{document}]]>
yields to
Ψ(Υ)≤dSdtS-d.

Thus, for each solution (I(0),S(0),V(0))∈Θ , we have

1t∫0tΨ(B)dr≤1t∫0tdSdtS-ddr=1tlnS(t)S(0)-d,
which implies
lim supt→∞sup1t∫0tΨ(B)dr<-d2<0.

Therefore, the conditions of Theorem  2 are hold and this concludes that E∗ is globally asymptotically stable. □

Numerical examples

Suppose the parameters in model ( 1 ) as Λ=30 , β=0.8 , δ=0.2 , α=0.1 , η=0.1 , and ξ=0.2 . For these values, the mentioned line in ( 9 ) is shown in Fig.  1 . The ordered pair (ϑ,σ)=(0.2,0.1) lies below the line and by corollary  1 the endemic state E∗=(I∗,S∗,V∗)=(34.74,66.32,31.58) is stable. The point (ϑ,σ)=(0.4,0.5) lies above the line. In this case, the disease-free equilibrium E0=(I0,S0,V0)=(0,53.33,96.67) is stable and the infection will be wiped out from population. Besides, for these values for (ϑ,σ) we have R0=1.44>11$$\end{document}]]> and R0=0.71<1 , respectively. Thus, these results can be confirmed also by Theorem 1 . Now, if it is supposed that the vaccination proportions in new members and in susceptibles are equal, then the optimal value of such quantity is given by relation ( 10 ) as τ∗=0.286 . For τ=0.05<τ∗ , the infection will remain in the population and the endemic state is stable, while for τ=0.4>τ∗\tau ^*$$\end{document}]]> the number of infected individuals vanishes and the disease-free state is stable (see Fig.  2 ). Figures  3 and 4 describe the behavior of the model for various initial values and show the solutions are stable. Figure  3 shows the number of infected individuals (sub-figure (a)) and the phase diagram of I ( t ) and S ( t ) (sub-figure (b)) when τ=0.4 , and it can be seen that E0 is stable. Figure  4 shows that I ( t ) does not vanish when τ=0.05 , and in this case, E∗ is stable .

Fig. 1

The line of optimal proportions of vaccine. For those values of ϑ and σ that lie below the line, E∗ is stable (red point), and for such values that lie above the line, E0 is stable (blue point)

Fig. 2

Solutions of model ( 1 ) for τ=0.05 (left) and τ=0.4 (right)

Fig. 3

Solutions of I ( t ) ( a ) and the phase diagram of I ( t ) and S ( t ) ( b ) for τ=0.4 with different initial values in each subpopulation. In this case, R0=0.80<1 and disease will be extinct with I0=0,S0=53.33,V0=96.67

Fig. 4

Diagrams of I ( t ) ( a ) and the phase diagram of I ( t ) and S ( t ) ( b ) for τ=0.05 with different initial values. This gives R0=1.73>11$$\end{document}]]> , and disease will persist with I∗=34.74,S∗=66.32,V∗=31.58

Conclusions

In this paper, a deterministic SIS epidemic model with temporary vaccination was studied. Vaccination includes both susceptible and new members, and disease transmission takes place at standard incidence rate. The number of individuals which are added to the population per unit time is constant and differs from the number of individuals that die and leave the population per unit time. Thus, the total population size is variable. Dynamics of the model at the endemic equilibrium were determined by the basic reproduction number R0 ; when R0>11$$\end{document}]]> , the unique endemic equilibrium E∗ is locally asymptotically stable and uniformly persists. In addition, it was shown that in this case E∗ is globally asymptotically stable under some conditions by employing a geometric approach and second additive compound matrix method. A numerical discussion was also performed to support the theoretical results.

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

  1. Allen (2007) Pearson/Prentice Hall
  2. Arino et al. (2003) Global results for an epidemic model with vaccination that exhibits backward bifurcation 64(1) (pp. 260-276) 10.1137/S0036139902413829
  3. Brauer and Castillo-Chavez (2001) Springer 10.1007/978-1-4757-3516-1
  4. Chalub and Souza (2014) Discrete and continuous SIS epidemic models: a unifying approach (pp. 83-95) 10.1016/j.ecocom.2014.01.006
  5. Farnoosh and Parsamanesh (2017) Disease extinction and persistence in a discrete-time SIS epidemic model with vaccination and varying population size 31(15) (pp. 4735-4747) 10.2298/FIL1715735F
  6. Farnoosh and Parsamanesh (2017) Stochastic differential equation systems for an SIS epidemic model with vaccination and immigration 46(17) (pp. 8723-8736) 10.1080/03610926.2016.1189571
  7. Hu et al. (2012) Stability analysis in a class of discrete SIRS epidemic models 13(5) (pp. 2017-2033) 10.1016/j.nonrwa.2011.12.024
  8. Jami et al. (2016) Numerical solution of stochastic SIR model via split-step forward Milstein method 2016(1) (pp. 38-45)
  9. Jianquan and Zhien (2004) Global analysis of SIS epidemic models with variable total population size 39(11) (pp. 1231-1242) 10.1016/j.mcm.2004.06.004
  10. Kribs-Zaleta and Velasco-Hernandez (2000) A simple vaccination model with multiple endemic states 164(2) (pp. 183-201) 10.1016/S0025-5564(00)00003-1
  11. Li and Ma (2002) Qualitative analyses of SIS epidemic model with vaccination and varying total population size 35(11) (pp. 1235-1243) 10.1016/S0895-7177(02)00082-1
  12. Li and Muldowney (1996) A geometric approach to global-stability problems 27(4) (pp. 1070-1083) 10.1137/S0036141094266449
  13. Liu et al. (2008) SVIR epidemic models with vaccination strategies 253(1) (pp. 1-11) 10.1016/j.jtbi.2007.10.014
  14. Parsamanesh (2017) Global stability analysis of a VEISV model for network worm attack 79(4) (pp. 179-188)
  15. Parsamanesh (2018) Global dynamics of an SIVS epidemic model with bilinear incidence rate (pp. 544-557)
  16. Rahmani et al. (2016) Numerical solution of stochastic SIR model by Bernstein polynomials 2016(1) (pp. 19-25)
  17. Safan and Rihan (2014) Mathematical analysis of an SIS model with imperfect vaccination and backward bifurcation (pp. 195-206) 10.1016/j.matcom.2011.07.007
  18. Sun and Hsieh (2010) Global analysis of an SEIR model with varying population size and vaccination 34(10) (pp. 2685-2697) 10.1016/j.apm.2009.12.005
  19. Yang et al. (2010) Global analysis for a general epidemiological model with vaccination and varying population 372(1) (pp. 208-223) 10.1016/j.jmaa.2010.07.017
  20. Yang (2015) Global stability of VEISV propagation modeling for network worm attack 39(2) (pp. 776-780) 10.1016/j.apm.2014.07.010
  21. Zhang and Liu (2009) Backward bifurcation and global dynamics of an SIS epidemic model with general incidence rate and treatment 10(2) (pp. 565-575) 10.1016/j.nonrwa.2007.10.011
  22. Zhao and Jiang (2014) The threshold of a stochastic SIS epidemic model with vaccination (pp. 718-727)
  23. Zhou and Cui (2011) Modeling and stability analysis for a cholera model with vaccination 34(14) (pp. 1711-1724)