Research Article

ISSN: 2574 -1241 DOI: 10.26717/BJSTR.2023.53.008392

On Stability of One Mathematical Model of the Epidemic

Spread Under Stochastic Perturbations
Leonid Shaikhet*
Department of Mathematics, Ariel University, Ariel 40700, Israel
*Corresponding author: Leonid Shaikhet, Department of Mathematics, Ariel University, Ariel 40700, Israel

ARTICLE INFO ABSTRACT

Received: September 09, 2023 This paper continues the study of the asymptotic properties of the known SAIRP epidemic model
under stochastic perturbations. The SAIRP epidemic model is described by a system of five nonlinear
Published: October 18, 2023 differential equations. It is assumed that the system is influenced by stochastic perturbations that are
of the type of white noise and are proportional to the deviation of the current system state from one of
Citation: Leonid Shaikhet. On Stability the system equilibriums. It is shown that sufficient conditions of stability in probability for two different
equilibria of the considered system are formulated via a simple linear matrix inequality (LMI), that can
of One Mathematical Model of the Epi- be easily studied via MATLAB. Two demonstrative examples illustrate the obtained results via numerical
demic Spread Under Stochastic Pertur- simulation of solutions of the considered system of five nonlinear Ito’s stochastic differential equations.
The research method used here can be applied to a lot of other more complicated models in different
bations. Biomed J Sci & Tech Res 53(3)-
applications.
2023. BJSTR. MS.ID.008392.
Keywords: Equilibria; Stability in Probability; Asymptotic Mean Square Stability; Lyapunov Function;
Linear Matrix Inequality (LMI); Numerical Simulation

Introduction ment rate Λ and the natural death rate µ are assumed to be constant.
The susceptible individuals S(t) become infected by contact with ac-
During the last years investigations of epidemic models are very
tive infected I(t) and asymptomatic infected individuals A(t), at a rate
popular in research (see, for instance, [1-3] and the references there- θ A(t ) + I (t )
of infection β N (t ) , where θ represents a modification parameter
in). The so-called SAIRP epidemic model is defined by the following
for the infectiousness of the asymptomatic infected individuals A. It
system of ordinary differential equations [1]:
is supposed also that all parameters of the system (1.1) are positive
.  θ A(t ) + I (t )  and, besides, p < 1, u < 1.
S (t ) = Λ −  β (1 − p (1 − u )) + ψ p (1 − u ) + µ  S (t ) + ω P (t ),
 N (t )  In [1] some properties of stability of the system (1.1) equilibria
. θ A(t ) + I (t ) are studied. Below, following the method from [4], stability in prob-
Α(t ) = β (1 − p (1 − u )) S (t ) − (v + µ) A(t ),
N (t )
ability of two equilibria of the system (1.1) is investigated by the as-
.
) vA(t ) − (δ + µ ) I (t ),
I (t= sumption that the system (1.1) is exposed to stochastic perturbations
. that are of the type of the white noise and are directly proportional to
R(t ) δ I (t ) − µ R(t ),
=
.
the deviation of a system state from an appropriate equilibrium.
P(t )= ψ p (1 − u ) S (t ) − (ω + µ ) P(t ). (1.1)
Equilibria
Here it is supposed that the population is subdivided into five
distinct classes: susceptible individuals (S(t)); asymptomatic infected Assuming that all variables in the system (1.1) are constants, we
individuals (A(t)); active infected individuals (I(t)); removed (R(t)); obtain that the equilibria of the system (1.1) are defined by the sys-
protected individuals (P (t)). The total population, N (t) = S(t) + A(t) tem of algebraic equations
+ I(t) + R(t) + P (t), with t ≥0 has a variable size, where the recruit-

Copyright@ : Leonid Shaikhet | Biomed J Sci & Tech Res | BJSTR.MS.ID.008392. 44630
 Volume 53- Issue 3 DOI: 10.26717/BJSTR.2023.53.008392

 θ A+ I  the system of Ito’s stochastic differential equations [5]

Λ −  β (1 − p (1 − u )) +ψ p (1 − u ) + µ  S + ω P =0,
 N (t ) 
  θ A(t ) + I (t )  
θ A+ I dS (t ) =  Λ −  β (1 − p(1 − u )) + ψ p(1 − u ) + µ  S (t ) + ω P(t )  d (t ) + σ 1 ( S (t ) − S ∗ )dw1 ( s ),
β (1 − p(1 − u )) S − (v + µ ) A = 0,   N (t )  
N (t )  θ A(t ) + I (t ) 
dA(t )=  β (1 − p (1 − u )) S (t ) − (v + µ ) A(t )  d (t ) + σ 2 ( A(t ) − A∗ )dw2 ( s ), (3.1)
vA − (δ + µ ) I = 0,  N (t ) 
t ) [ vA(t ) − (δ + µ ) I (t )] d (t ) + σ 3 ( I (t ) − I ∗ )dw3 ( s ),
dI (=
δ I − µR = 0,
dR(t ) =[δ I (t ) − µ R(t ) ] d (t ) + σ 4 ( R(t ) − R∗ )dw4 ( s ),
ψ p(1 − u ) S − (ω + µ ) P = 0, (2.1) ) [ψ p (1 − u ) S (t ) − (ω + µ ) P(t ) ] d (t ) + σ 5 ( P(t ) − P∗ )dw5 ( s ),
dP(t=

with the solutions: disease-free equilibrium where σ1, ..., σ5 are constants and w1(t), ..., w5(t) are the mutually
independent 𝔉t -adapted standard Wiener processes.
E0∗ = ( S0∗ , A0∗ , I 0∗ , R0,∗ P0∗ ), Note that the equilibrium (S*, A*, I*, R*, P*) of the deterministic
A= I= R= 0,
∗
0
∗
0
∗
0
system (1.1) is also the solution of the system of Ito’s stochastic dif-
(ω + µ )Λ ferential equations (3.1). Stochastic perturbations of this type were
S0∗ = , first proposed in [6] for SIR epidemic model and later also for a lot of
µ[ω + µ +ψ p(1 − u )]
other different applied models (see [7] and the references therein).
ψ p(1 − u )Λ
P0∗ = , (2.2)
µ[ω + µ +ψ p(1 − u )] Centralization and Linearization
and endemic equilibrium Consider the nonlinear differential equation
.
E+∗ = ( S +∗ , A+∗ , I +∗ , R+∗, P+∗ ), x(t ) = F ( x(t )), (4.1)
(ω + µ )Λ
S +∗ = R −1 ,
µ[ω + µ +ψ p (1 − u )] 0 where x(t) ∈ Rn and the equation F(x) = 0 has a solution x∗ that is
Λ an equilibrium of the differential equation (4.1). Using the new vari-
=A+∗ (1 − R0−1 ),
v+µ able y(t) = x(t) − x∗, represent the equation (4.1) in the form
vΛ
=I +∗ (1 − R0−1 ), (2.3) .
(v + µ )(δ + µ )
y (t ) F ( x∗ + y (t )).
= (4.2)
vδΛ
R+∗ (1 − R0−1 ),
µ (v + µ )(δ + µ ) It is clear that stability of the zero solution of the equation (4.2)
ϕ p (1 − u )Λ is equivalent to stability of the equilibrium x* of the equation (4.1).
P+∗ = R −1 ,
µ[ω + µ +ψ p (1 − u )] 0
∂Fi
=JF = , i, j 1,....., n,
where the basic reproduction number R0 > 1, Let be the Jacobian matrix of the func-
∂xi
o( y )
= tion F {= F1 ,.... Fn } and lim y →0 0, where y is the Eu-
y
( β (1 − p (1 − u )))(θ (δ + µ ) + v)(ω + µ ) clidean norm in R . Using Taylor’s expansion in the form
n
R0 = . (2.4)
(v + µ )(δ + µ )(ω + µ +ψ p (1 − u )) ) F ( x* ) + J F ( x∗ ) y + o( y ) and the equality F(x*) = 0, we
F ( x∗ + y=
Note also that, summing all equations of the system (2.1), we ob- obtain the linear approximation
tain N ∗ = Λ for both equilibria (2.2) and (2.3).
µ .
Stochastic Perturbations z (t ) = J F ( x∗ ) z (t ) (4.3)
Let {Ω, 𝔉, P} be a complete probability space, {𝔉t, t ≥ 0} be a nonde-
of the nonlinear differential equation (4.2). So, a condition for the
creasing family of sub-σ-algebras of 𝔉, i.e., 𝔉t1 ⊂ 𝔉t2 ⊂ 𝔉 for t1 < t2, E be
asymptotic stability of the zero solution of the linear equation (4.3)
the mathematical expectation with respect to the measure P.
is also a condition for the local stability of the equilibrium x* of the
Let us suppose that the system (1.1) is exposed to stochastic per- initial nonlinear equation (4.1).
turbations that are of the type of the white noise and are directly pro-
To construct the linear approximation of the system (3.1) let us
portional to the deviation of the system state (S(t), A(t), I(t), R(t), P
put (′ is the sign of transpose)
(t)) from one of the equilibria (S*, A*, I*, R*, P*). As a result, we obtain

Copyright@ : Leonid Shaikhet | Biomed J Sci & Tech Res | BJSTR.MS.ID.008392. 44631
 Volume 53- Issue 3 DOI: 10.26717/BJSTR.2023.53.008392

and
x(t ) = ( S (t ), A(t ), I (t ), R(t ), P(t ))' ,
S∗
x∗ = ( S ∗ , A∗ , I ∗ , R∗ , P∗ )' a22= θβ (1 − p (1 − u )) − (v + µ ),
N∗
(t ) x(t ) − x∗ ,
y= S∗
a23 = β (1 − p (1 − u )) ∗ ,
N ∗ = S ∗ + A∗ + I ∗ + R∗ + P∗ . (4.4) N
a=21 a=24 a=
25 0. (4.10)
Representing the system (1.1) in the form (4.1) and calculating
the Jacobian matrix, we get the linear part of the system (3.1) in the
Remark 1 Let V = V (z) be a twice differentiable function
form.
5
5
of z ∈ R . The generator of the equation (4.5) has the form [5]
dz (t ) Az (t )dt + ∑ Bi z (t )dwi (t ),
= (4.5)
1 5 '
i =1 (∇V )' Az +
LV = ∑ z Bi∇2VBi z.
2 i =1
(4.11)

where Bi is the matrix with all zero elements besides of bii = σi, i
=1,…,5: Stability
 a11 a12 a13 a14 a15  Definition 1
 
 a a22 a23 a24 a25 
 21
 Put y(t) = (S(t), A(t), I(t), R(t), P(t))-(S*, A*, I*, R*, P*). The solution
 
=A  0 v − (δ + µ ) 0 0  (4.6) (S*, A*, I*, R*, P*) of the system (3.1) is called stable in probability if for


0 0 δ −µ 0 

any ε1 > 0 and ε 2 > 0 there exists δ > 0 such that y(t) satisfies the
 ψ p (1 − u ) 0 0 0 − (ω + µ )  condition P{sup y (t ) > ε1} < ε 2 for any y(0), such that P{ y (0) < δ } = 1
 
  t ≥0

with
 θ A∗ + I ∗ S∗  Definition 2
a11 =−  β (1 − p (1 − u )) ∗
(1 − ∗ ) +ψ p (1 − u ) + µ  ,
 N N 
The zero solution of the equation (4.5) is called:
S∗  θ A∗ + I ∗ 
− β (1 − p (1 − u ))
a12 = θ − ,
N∗  N∗  -mean square stable if for each ε > 0 there exists a δ > 0 such
2
S ∗  θ A∗ + I ∗ 
− β (1 − p (1 − u )) ∗ 1 −
that E z (t ) < ε , t ≥ 0, , provided that E z (0) 2 < δ ;
a13 = , (4.7)
N  N∗ 
S ∗ (θ A∗ + I ∗ )
-asymptotically mean square stable if it is mean square stable
a14 = β (1 − p (1 − u )) , 2
and for each initial value z(0), such that E z (0) < ∞, ,the solution z(t)
( N ∗ )2
a=
15 a14 + ω , of the equation (4.5) satisfies the condition lim E z (t ) 2 = 0.
t →∞

and Remark 2

θ A∗ + I ∗ 
S∗  It is known [7] that sufficient conditions for asymptotic mean
a21 = β (1 − p (1 − u )) 1 − ∗  ,
N∗
 N  square stability of the zero solution of the linear part of a stochastic
S∗  θ A∗ + I ∗  nonlinear system with the order of nonlinearity higher than one at
a22 = β (1 − p (1 − u )) ∗  θ −  − (v + µ ), (4.8)
N  N∗  the same time are sufficient conditions for stability in probability of
S ∗  θ A∗ + I ∗  the initial nonlinear system solution. So, for investigation of stability
a23 = β (1 − p (1 − u )) 1 − ,
N∗  N∗  in probability of the equilibrium (S*, A*, I*, R*, P*) of the system (3.1)
S ∗ (θ A∗ + I ∗ ) it is enough to get conditions for asymptotic mean square stability of
a24 = − β (1 − p (1 − u ))
a25 = ,
( N ∗ )2 the zero solution of the linear equation (4.5).

E0∗ Theorem 1
In particular, for the equilibrium the elements (4.7) and (4.8)
of the matrix (4.6) are Let for the matrices A and Bi, i = 1,…,5, of the equation (4.5) there
exists a positive definite matrix P such that the following linear matrix
a11 =−(ψ p (1 − u ) + µ ),
inequality
S∗
−θβ (1 − p (1 − u ))
a12 = , (4.9)
N∗ 5
S ∗
− β (1 − p (1 − u )) ∗ ,
a13 = PA + A' P + ∑ Bi' PBi < 0 (5.1)
N i =1
=a14 0,= a15 ω
holds. Then the equilibrium (S*, A*, I*, R*, P*) of the system (3.1)
is stable in probability.

Copyright@ : Leonid Shaikhet | Biomed J Sci & Tech Res | BJSTR.MS.ID.008392. 44632
 Volume 53- Issue 3 DOI: 10.26717/BJSTR.2023.53.008392

Proof ( S0∗ , A0∗ , I 0∗ , R0∗ , P0∗ ) = (12.5445, 0, 0, 0, 2.4555). (5.4)

Let L be the generator [5] of the equation (4.5). Using the Lyapun- Via MATLAB the following maximal values of the white noise lev-
ov function V ( z ) = z ' Pz via (4.11) we have els were obtained, for which the LMI (5.1) holds and, therefore, the
5
LV( z ) 2 z ' PAz + ∑ z ' Bi' PBi z
= equilibrium (5.4) is stable in probability: σ1 = 1.4, σ2 = 0.93, σ3 = 1.2,
i =1
σ4 = 1.4, σ5 = 1.4.
 5

= z '  PA + A' P + ∑ Bi' PBi  z. (5.2)
 
i =1
In Figure 1 100 trajectories of the solution of the system (3.1) are
So, if the linear matrix inequality (5.1) holds then via (5.2) shown with the initial values S(0) = 22, A(0) = 11, I(0) = 4, R(0) = 8,
2
LV ( z ) ≤ −c z for some c > 0 and, therefore, the zero solution of the P(0) = 17. All trajectories (S(t)-brown, A(t)-violet, I(t)-blue, R(t)-red,
linear stochastic differential equation (4.5) is asymptotically mean P(t)-green) converge to the stable in probability equilibrium (5.4).
square stable [7]. Via Remark 2 it means that the appropriate equi-
librium (S*, A*, I*, R*, P*) of the nonlinear system (3.1) is stable in Example 2 Putting
probability. The proof is completed.
Λ 15, =
= µ 1,=θ 1, = ψ 0.08,
Example 1 Putting = v 0.18,
= δ 0.033,
= ω 0.0013,
= p 0.4,
= u 0.3,
= β 2, (5.5)
=Λ 15, = µ 1, =θ 1,=ψ 0.4,
=v 0.15,
= δ 0.033,
= ω 0.0013, (5.3)
from (2.4) and (2.3) we obtain R0 = 1.3541 and N0=15.
=p 0.7,
= u 0.3,
= β 1.5,
( S , A+∗ , I +∗ , R+∗ , P+∗ ) = (10.8264,3.3317, 0.4509, 0.1488, 0.2422).
∗
+ (5.6)
from (2.4) and (2.2) we obtain R0 = 0.6371 and N0=15.

Figure 1: 100 trajectories of the system (3.1) solution are shown with

Λ 15, =
= µ 1, = θ 1,=
ψ 0.4,= v 0.15,=δ 0.033, = ω 0.0013,
=p 0.7,
= u 0.3,
= β 1.5,
= σ 1 1.4,
= σ 2 0.93,
= σ 3 1.2,
= σ 4 1.4,
= σ 5 1.4,=
S (0) 22,=
A(0) 11,
I (0) 4,=
= R(0) 8,=
P(0) 17, ( S0∗ , A0∗ , I 0∗ , R=
0 , P0 )
∗ ∗
(12.5445, 0, 0, 0, 2.4555)

Copyright@ : Leonid Shaikhet | Biomed J Sci & Tech Res | BJSTR.MS.ID.008392. 44633
Volume 53- Issue 3 DOI: 10.26717/BJSTR.2023.53.008392

Via MATLAB the following maximal values of the white noise lev- In Figure 2 100 trajectories of the solution of the system (3.1) are
els were obtained, for which the LMI (5.1) holds and, therefore, the shown with the initial values S(0) = 7, A(0) = 4.5, I(0) = 9, R(0) = 5.5,
equilibrium (5.6) is stable in probability: σ1 = 1.5, σ2 = 0.79, σ3 = 1.2, P(0) = 2.7. All trajectories (S(t)-brown, A(t)-violet, I(t)-blue, R(t)-red,
σ4 = 1.3, σ5 = 1.3. P(t)-green) converge to the stable in probability equilibrium (5.6).

Figure 2: 100 trajectories of the system (3.1) solution are shown with

Λ 15, =
= µ 1, =θ 1,=
ψ 0.08,= v 0.18,=δ 0.033, =ω 0.0013,
=p 0.4,= β 2,=
u 0.3,= σ 1 1.5,=
σ 2 0.79,=σ 3 1.2,=
σ 4 1.3, =
σ 5 1.3, S=
(0) 7, A=
(0) 4.5,
I (0) 9,=
= R(0) 5.5,=
P(0) 2.7, ( S +∗ , A+∗ , I +∗ , R=
+ , P+ )
∗ ∗
(10.8264,3.3317, 0.4509, 0.1488, 0.2422)

Remark 3 be easily studied via MATLAB. Two demonstrative examples illustrate

Note that for the numerical simulation of the Wiener process tra-
jectories in Examples 1 and 2 a special algorithm has been used, de-
scribed in detail in [7].
Conclusion
Asymptotic properties of the SAIRP epidemic model, described
by a system of five nonlinear differential equations, are studied un-
der stochastic perturbations. It is shown that a sufficient condition
of stability in probability for two equilibria of the considered system
is formulated using a simple linear matrix inequality (LMI) that can
the obtained results via numerical simulation of solutions of the con-
sidered system of five nonlinear Ito’s stochastic differential equations.
These simulations can be continued for getting more detail analysis of
the considered epidemic model by real values of the system parame-
ters in some real situations. The research method used here can be
applied to a lot of other more complicated models in different appli-
cations.
References
1. G Cantin, CJ Silva, A Banos. (2022) Mathematical analysis of a hybrid mod-
el: impacts of individual behaviors on the spreading of an epidemic, Net-
works & Heterogeneous Media 17(3): 333-357.

Copyright@ : Leonid Shaikhet | Biomed J Sci & Tech Res | BJSTR.MS.ID.008392. 44634
Volume 53- Issue 3 DOI: 10.26717/BJSTR.2023.53.008392

2. CJ Silva, G Cantin, C Cruz, R Fonseca-Pinto, RP Fonseca, et al. (2021) Com- 5. II Gikhman, AV Skorokhod (1972) Stochastic differential equations.
plex network model for COVID-19: human behavior, pseudo-periodic solu- Springer, Berlin.
tions and multiple epidemic waves, Journal of Mathematical Analysis and
Applications 514(2): 125171. 6. E Beretta, V Kolmanovskii, L Shaikhet (1998) Stability of epidemic model
with time delays influenced by stochastic perturbations. Mathematics and
3. CJ Silva, C Cruz, DFM Torresetal (2021) Optimal control of the COVID-19 Computers in Simulation. 45(3-4): 269-277.
pandemic: Controlled sanitary deconnement inPortugal. Scientic Reports
11(1): 3451. 7. L Shaikhet (2013) Lyapunov functionals and stability of stochastic func-
tional differential equations. Springer Science & Business Media.
4. L Shaikhet (2022) Some generalization of the method of stability inves-
tigation for nonlinear stochastic delay differential equations. MDPI. Sym-
metry 14(8): 1734.

ISSN: 2574-1241 Assets of Publishing with us

DOI: 10.26717/BJSTR.2023.53.008392
• Global archiving of articles
Leonid Shaikhet. Biomed J Sci & Tech Res
• Immediate, unrestricted online access
This work is licensed under Creative • Rigorous Peer Review Process
Commons Attribution 4.0 License
• Authors Retain Copyrights
Submission Link: https://biomedres.us/submit-manuscript.php
• Unique DOI for all articles

https://biomedres.us/

Copyright@ : Leonid Shaikhet | Biomed J Sci & Tech Res | BJSTR.MS.ID.008392. 44635