Qualitative Behaviour of A Model of An SIRS Epidemic: Stability and Permanence

Download as pdf or txt
Download as pdf or txt
You are on page 1of 19

Applied Mathematics & Information Sciences 5(2) (2011), 220-238

An International Journal
c 2011 NSP
Qualitative Behaviour of a Model of an SIRS Epidemic:
Stability and Permanence
Hamid El Maroufy
1
, Adil Lahrouz
2
and PGL Leach
3
1
Department of Applied Mathematics, Beni-Mellal Faculty of Sciences and Technology,
B.P : 523, Beni-Mellal, Morocco
Email Address : [email protected]
2
Department of Mathematics, Sidi Mohammed Ben Abdellah University,
B.P : 1796, Atlas, Fez, Morocco.
3
School of Mathematical Sciences, University of KwaZulu-Natal,
Private Bag X54001 Durban 4001, Republic of South Africa.
We consider a classical model of a SIRS epidemic in an open population. The positivity
and permanence are studied and explicit formul are obtained by which the eventual
lower bound of the density of infectives can be computed. The stability of the model
is studied. We mainly use the Lyapunov functional to established the global stability
of disease-free and endemic equilibrium points for both the deterministic and stochas-
tic models. In addition we illustrate the dynamic behaviour of the deterministic and
stochastic models via a numerical example.
Keywords: model of SIRS epidemic, Local and global stability, Lyapunov functional.
1 Introduction
Epidemiology is the study of the spread of infectious diseases with the objective to
trace factors that contribute to their dynamics and stabilities. Formerly and recently mod-
els, which have the population subclasses (i) the susceptibles (S), (ii) the infectives (I)
and (iii) the removals (R), have been studied by a number of authors (see for example
Bailey [2], Tornatore [12], Beretta and Takeuchi [3], Zhang and Teng [13]). The basic
and important research subjects for recent studies are the existence of the threshold values
which distinguish whether the disease dies out, the stability of the disease-free and the en-
demic equilibria, permanence and extinction. Most of these works deal with local stability
PGLL thanks the University of KwaZulu-Natal and the National research Foundation of South Africa for their
continued support. Any opinions expressed in this paper should not be construed as being of either institution.
Qualitative Behaviour of a Model of an SIRS Epidemic 221
of equilibria of the deterministic model in a closed population. There are very few works
which study both deterministic and stochastic stabilities of the model.
In this paper we consider a model of an SIRS epidemic which is the extension of the clas-
sical SIR model (see [5] and [13]) for which it is assumed that all newborn are susceptibles
and the population grows at a rate b > 0. The susceptibles, infectives and removals die at
different rates, d
1
, d
2
and d
3
. It is biologically natural to assume that d
1
< min(d
2
, d
3
).
In addition we suppose that an individual in the class (R) can be cured and becomes a new
susceptible or infected another time, respectively, with rates
2
and
3
. Mathematically the
model that we consider is dened as follows. At time t the variables, S(t), I(t) and R(t),
represent, respectively, the density of susceptibles, infected and removed individuals. The
epidemial process is thus completely determined by {(S(t), I(t), R(t)); t 0} while its
dynamics is governed by the system of ordinary differential equations
_

_
S

= b SI d
1
S +
3
R,
I

= SI (d
2
+
1
) I +
2
R,
R

=
1
I (d
3
+
2
+
3
) R,
(1)
with the initial conditions S (0) = S
0
, I (0) = I
0
and R(0) = R
0
, where the parameter
is the average number of contacts per infective per unit of time. As is known, systems like
(1) are very important mathematical models which describe epidemiological dynamics. As
mentioned in the rst paragraph, the most basic and important questions to ask for these
systems in the theory of mathematical epidemiology concern permanence and stability.
Recently Ma et al [8] and Zhand and Teng [13] studied an SIRS system with time delay.
Under certain conditions they proved the permanence of the disease and stability. Motivated
by the above works the present analysis aims to establish some conditions on the positivity,
boundness of solution and permanence of the epidemic. By using methods in [8] we obtain
explicit formul of the eventual lower bound of infectious individuals and the total size of
the epidemic. Applying the Lyapunov functional, we give some sufcient conditions for
local and global stabilities of the deterministic model governed by (1) and of its stochastic
version, which is obtained by random perturbation of the deterministic model.
The organization of this paper is as follows. In the next Section we give the equilibrium
points of the model. In Section 3 we establish the positivity and we give the sufcient
conditions of the ultimate boundedness of system (1). In Section 4 we give the sufcient
and necessary conditions for the local and global stability of the disease-free and endemic
equilibria. In Section 5 the stochastic stability of the disease-free and endemic equilibria
is proved. In Section 6 numerical simulations are performed to complement the analytical
results. Finally in Section 7 a brief discussion of the main results and some ideas for future
research are given.
222 El Maroufy, Lahrouz, Leach
2 The Equilibrium Points
The equilibria of (1) are the solutions of the system
_

_
b SI d
1
S +
3
R = 0
SI (d
2
+
1
) I +
2
R = 0

1
I (d
3
+
2
+
3
) R = 0.
(2)
It is evident that (2) has two solutions, E
0
= (b/d
1
, 0, 0) which called the disease-free
equilibrium point and another solution, E

= (S

, I

, R

), where
_

_
S

=
1

.
_
d
2
+
1

2
d
3
+
2
+
3
_
,
I

=
_
S

3
d
3
+
2
+
3
_
1
(b d
1
S

) ,
R

=

1
d
3
+
2
+
3
I

is positive if and only if


b
d
1
> d
2
+
1

2
d
3
+
2
+
3
. Under the last condition E

is
called the endemic-disease point.
3 Positivity, Boundedness and Permanence
3.1 Positivity and boundedness of the solution
The application of the classical theory of Ordinary Differential Equations implies that
for every set of initial data, (S
0
, I
0
, R
0
), there exists a unique solution, (S (t) , I (t) , R(t)),
dened in the maximal open interval (T, T) with T > 0.
Proposition 3.1. Let (S, I, R) be the solution of (1).
(i) If S
0
> 0, I
0
> 0 and R
0
> 0, then S (t) > 0, I (t) > 0 and R(t) > 0 for every t
[0, T).
(ii) The solution (S, I, R) is dened in [0, ) and limsup
t
N(t)
b
d
1
, where N(t) =
S(t) +I(t) +R(t).
Proof. We rstly prove (i). To do this we suppose that there exists t
0
(0, T) such that
S (t
0
) = 0, S

(t
0
) 0 and S (t) > 0 for t [0, t
0
) . Then I (t) > 0 for t [0, t
0
]. If
this be not the case, there exists t
1
[0, t
0
] such that I (t
1
) = 0, I

(t
1
) 0 and I (t) >
0 fort [0, t
1
) . Integration of the third equation of (1) leads to
R(t) = R
0
exp [(d
3
+
2
+
3
) t]
+
1

t
0
exp [(d
3
+
2
+
3
) (t )] I () d > 0 for t [0, t
1
].
Qualitative Behaviour of a Model of an SIRS Epidemic 223
Then I

(t
1
) =
2
R(t
1
) > 0. This is a contradiction. Hence R(t) > 0 for every t
[0, t
0
] . Therefore S

(t
0
) = b +
3
R(t
0
) > 0, but this leads to a contradiction to the
supposition that S

(t
0
) 0, which completes the proof of (i).
For (ii) we note that N

= b d
1
S d
2
I d
3
R b d
1
N. By integrating the last
inequality we obtain
N (t)
b
d
1
_
1 e
d
1
t
_
for every t [0, T)

2b
d
1
.
The solution (S, I, R) is bounded in the interval [0, T). Therefore N (t)
b
d
1
_
1 e
d
1
t
_
for every t [0, ). Finally limsup
t
N (t)
b
d
1
.
Remark Following the same method used to demonstrate Proposition 3.1, we see that
system(1) with the initial conditions S
0
0, I
0
0 and R
0
0 has a nonnegative solution
dened in all Rand the set =
_
(S, I, R) /S > 0, I > 0, R > 0 and S +I +R
b
d
1
_
is invariant by (1).
3.2 Permanence of the Epidemic
Lemma 3.1. Let (S, I, R) be the solution of system (1). If there exists a sequence (t
n
) such
that t
n
, S (t
n
) l, I (t
n
) 0 and R(t
n
) 0, then l =
b
d
1
.
Proof. We have 0 l
b
d
1
. Suppose that 0 l <
b
d
1
. Since S (t
n
) l, I (t
n
) 0 and
R(t
n
) 0, it follows that (l, 0, 0) W (S
0
, I
0
, R
0
) which is the set of W limit.
Consider (S, I, R), the solution of (1) with the initial condition (l, 0, 0). Therefore
(S(t), I(t), R(t)) W (S
0
, I
0
, R
0
) for every t because the set W (S
0
, I
0
, R
0
) is in-
variant by (1). It is easy to verify that S(t) =
b
d
1
+
_
l
b
d
1
_
e
d
1
t
and I (t) = R(t) =
0 for every t R. Since 0 l <
b
d
1
, contrary to the positivity of S in all R (Remark 2),
we have S(t) < 0 for t .
Proposition 3.2. Let (S, I, R) be the solution of (1) such that
b
d
1
> d
2
+
1

2
d
3
+
2
+
3
.
Then I

> 0. If I is bounded below by the real positive number m, then R



1
m
d
3
+
2
+
3
.
Proof. Suppose that I

= 0. We have in this case two possibilities: lim


t
I (t) = 0
or 0 = liminf
t
I (t) < limsup
t
I (t) . When lim
t
I (t) = 0, on account of the proof of
Proposition 3.1 we have R(t)
t
0 and S (t)
t
b
d
1
. Then for sufciently small and
224 El Maroufy, Lahrouz, Leach
t sufciently large S (t) >
b
d
1
(1 ) and
I

= SI (d
2
+
1
) I +
2
R
>
_

b
d
1
(1 ) (d
2
+
1
)
_
I +
2
R
=
_

b
d
1
(1 ) (d
2
+
1
)
_
I +

1

2
d
3
+
2
+
3
I

2
d
3
+
2
+
3
R

.
Therefore for t sufciently large
_
I +

2
d
3
+
2
+
3
R
_

>
_

b
d
1
(1 ) (d
2
+
1
) +

1

2
d
3
+
2
+
3
_
I. (4)
Since
b
d
1
> d
2
+
1

2
d
3
+
2
+
3
, we can choose sufciently small such that

b
d
1
(1 ) (d
2
+
1
) +

1

2
d
3
+
2
+
3
> 0.
Then for t sufciently large
_
I +

2
d
3
+
2
+
3
R
_

> 0. However, I +

2
d
3
+
2
+
3
R > 0 and
_
I +

2
d
3
+
2
+
3
R
_

t
0. This is a contradiction in this case. If 0 = liminf
t
I (t) <
limsup
t
I (t) , then there exists a sequence (t
n
)
n
such that I (t
n
)
n
0 and I

(t
n
) = 0.
From the second equation of (1)
2
R(t
n
) = (d
2
+
1
) I (t
n
) S (t
n
) I (t
n
) . Since
(S (t
n
))
n
is bounded, it follows that R(t
n
)
n
0 and liminf
t
R(t) = 0 (R(t) > 0).
Therefore liminf
t
_
I (t) +

2
d
3
+
2
+
3
R(t)
_
= 0. Hence there exists a sequence, denoted
also (t
n
)
n
, such that
I (t
n
) +

2
d
3
+
2
+
3
R(t
n
)
n
0 and I

(t
n
) +

2
d
3
+
2
+
3
R

(t
n
) = 0.
We have I (t
n
)
n
0 and R(t
n
)
n
0 since (S (t
n
))
n
is bounded. There exists, then,
a subsequence, denoted also (t
n
)
n
, such that (S (t
n
))
n
is convergent and by using Lemma
1 we deduce that S (t
n
)
n
b
d
1
. Applying the inequality (4) we get, for n sufciently
large,
0 = I

(t
n
) +

2
d
3
+
2
+
3
R

(t
n
)
>
_

b
d
1
(1 ) (d
2
+
1
) +

1

2
d
3
+
2
+
3
_
I (t
n
) > 0.
This is also a contradiction.
We now turn to prove the second result. If I is bounded below by m > 0, then R

1
I (d
3
+
2
+
3
) R
1
m (d
3
+
2
+
3
) R. Integrating the above inequality we
get
R(t)

1
m
d
3
+
2
+
3
+
_
R
0


1
m
d
3
+
2
+
3
_
exp [(d
3
+
2
+
3
) t] .
The result follows when t tends towards .
Qualitative Behaviour of a Model of an SIRS Epidemic 225
Theorem 3.1. Let (S, I, R) be the solution of (1) such that
b
d
1
> d
2
+
1


1

2
d
3
+
2
+
3
.
Then liminf
t
I (t) > e
(d
1
+)
, where and satisfy
b
d
1
+
_
1 e
(d
1
+)
_
> d
2
+
1

2
2
d
3
+
2
+
3
. (5)
Proof. Since
b
d
1
> d
2
+
1

2
d
3
+
2
+
3
, there exist sufciently small and sufciently
large such that (5) is satised. Firstly we claim that there exists t
0
> 0 such that I (t
0
) > .
If this be not the case, I (t) < for every t > 0. Then
S

= b SI d
1
S +
3
R > b (d
1
+) S.
Integrating the above inequality we obtain for every t > 0 that
S (t)
b
d
1
+
+
_
S
0

b
d
1
+
_
e
(d
1
+)t
.
Therefore for every t > S (t)
b
d
1
+
_
1 e
(d
1
+)
_
S

. Combining the second


and third equations in (1) we see that for every t >
_
I +

2
d
3
+
2
+
3
R
_

= SI (d
2
+
1
) I +

1

2
d
3
+
2
+
3
I
>
_
S

(d
2
+
1
) +

1

2
d
3
+
2
+
3
_
I

,
where I

= inf
t0
I (t) . By Proposition 3.1 we have I

> 0. Since
S

(d
2
+
1
) +

1

2
d
3
+
2
+
3
> 0
, we deduce that I (t) +

2
d
3
+
2
+
3
R (t)
t
, which contradicts the fact that
_
I +

2
d
3
+
2
+
3
R
_
is bounded (Proposition 3.1). Hence there exists t
0
> 0 such that
I (t
0
) > and we cannot have I (t) < for large t. Therefore we have two possi-
bilities: I (t) > for large t or I oscillates about . We claim in the second case that
I (t) > e
(d
1
+)
. If I oscillates about such that I (t
1
) = I (t
2
) = and I (t) <
for t (t
1
, t
2
) , we have I

= SI (d
2
+
1
) I +
2
R > (d
2
+
1
) I. It follows
by the integration of the previous inequality on [t
1
, t
2
] that I (t) I (t
1
) e
(d
2
+
1
)t

Re
(d
2
+
1
)(t
2
t
1
)
. If t
2
t
1
< , then I (t) > e
(d
1
+)
for every t [t
1
, t
2
] . If
this be not the case, we have I (t) > e
(d
1
+)
in [t
1
, t
1
+h] . We claim that the above
inequality remains true in the interval [t
1
+h, t
2
]. If this be not the case, there exist T
1
and
T
2
such that [T
1
, T
2
] [t
1
+h, t
2
] and
_

_
I (T
1
) = I (T
2
) = e
(d
1
+)
I (t) < e
(d
1
+)
for t (T
1
, T
2
)
I

(T
1
) < 0 < I

(T
2
) .
226 El Maroufy, Lahrouz, Leach
Using Proposition 2 we have
I

(T
1
) = S (T
1
) I (T
1
) (d
2
+
1
) I (T
1
) +
2
R(T
1
)
>
_
S

(d
2
+
1
)

e
(d
1
+)
+
2
R(T
1
)
>
_
S

(d
2
+
1
)

e
(d
1
+)
+

1

2
I

d
2
+
2
+
3
.
Since 0 < I

= inf
t0
I (t) I (t) < e
(d
1
+)
for t [T
1
, T
2
] , it follows that
I

(T
1
)
_
S

(d
2
+
1
) +

1

2
d
3
+
2
+
3
_
I

> 0.
This contradicts the fact that I

(T
1
) < 0. Therefore I (t) > e
(d
1
+)
for t sufciently
large. This completes the proof of the Theorem.
4 Deterministic Stability
4.1 The local stability of the disease-free point
Theorem 4.1. The disease-free point,
_
b
d
1
, 0, 0
_
, is locally asymptotically stable for (1) if
and only if
b
d
1
< d
2
+
1

2
d
3
+
2
+
3
.
Proof. Let u = (u
1
, u
2
, u
3
) = (S
b
d
1
, I, R). By (1) the t-derivative of u is
_

_
u

1
=
_
u
1
+
b
d
1
_
u
2
d
1
u
1
+
3
u
3
u

2
=
_
u
1
+
b
d
1
_
u
2
(d
2
+
1
) u
2
+
2
u
3
u

3
=
1
u
2
(d
3
+
2
+
3
) u
3
.
(6)
Linearising the system (6) at the point (0, 0, 0) we obtain u

= Mu, where
M =
_
_
_
d
1

b
d
1

3
0
b
d
1
(d
2
+
1
)
2
0
1
(d
3
+
2
+
3
)
_
_
_.
The matrix M has three eigenvalues,
_

1
= d
1

2
=
1
2
_
b
d
1
(d
2
+
1
) (d
3
+
2
+
3
) +

3
=
1
2
_
b
d
1
(d
2
+
1
) (d
3
+
2
+
3
)

_
with
=
_
b
d
1
(d
2
+
1
) (d
3
+
2
+
3
)
_
2
+ 4
1

2
.
The disease-free point
_
b
d
1
, 0, 0
_
is locally asymptotically stable if and only if the real parts
of the eigenvalues are negative. This is equivalent to
b
d
1
< d
2
+
1

2
d
3
+
2
+
3
.
Qualitative Behaviour of a Model of an SIRS Epidemic 227
4.2 The global stability of the disease-free point
Theorem 4.2. If
b
d
1
< d
2
+
1


1

2
d
3
+
2
+
3
, then the disease-free point, (
b
d
1
, 0, 0), is
globally asymptotically stable.
In order to prove the above theorem we need the following results.
Lemma 4.1. ( [11]) Let D be a bounded interval in R and g : (t
0
, ) R R be
a bounded and uniformly continuous function. Furthermore let x : (t
0
, ) R be a
solution of x

= g (t, x), which is dened on the whole interval (t


0
, ) . Then
(i) liminf
t
g (t, x

) 0 limsup
t
g (t, x

) and
(ii) liminf
t
g (t, x

) 0 limsup
t
g (t, x

) ,
where x

= limsup
t
x(t) and x

= liminf
t
x(t).
Proof. From the second equation of (1) we have I

(t) = g (t, I (t)), where g (t, I) =


S (t) I (t) (d
2
+
1
) I (t) +
2
R(t) . Using (ii) of Lemma 2 we deduce that
0 limsup
t
g (t, I

)
or
0 limsup
t
[S (t) I

(d
2
+
1
) I

+
2
R(t)]
and hence
0 S

(d
2
+
1
) I

+
2
R

.
Applying Lemma 2 to (1) we get
R



1
d
3
+
2
+
3
I

. (7)
Therefore 0 S

(d
2
+
1
) I

+

1

2
d
3
+
2
+
3
I

. Use of the previous inequality


and the fact that S


b
d
1
lead to
0
_
b
d
1
(d
2
+
1
) +

1

2
d
3
+
2
+
3
_
I

.
However,
b
d
1
(d
2
+
1
) +

1

2
d
3
+
2
+
3
< 0. Hence I

= 0 = lim
t
I (t) and by (7) we
obtain R

= 0 = limR(t)
t
. It remains to show that lim
t
S (t) =
b
d
1
. To do this it is
enough to see that S


b
d
1
. Applying (i) of Lemma 2 to the equation of (1) we obtain
(b S

I d
1
S

+
3
R)

0. Then b S

d
1
S

+
3
R

0. Since
I

= R

= 0, it follows that b d
1
S

0.
228 El Maroufy, Lahrouz, Leach
4.3 The stability of the endemic point
Theorem 4.3. If
b
d
1
> d
2
+
1

2
d
3
+
2
+
3
, then the endemic point is locally asymptotically
stable. Moreover there exists an explicit attractive region A for the solution of (1), that is,
for any initial condition (S
0
, I
0
, R
0
) such that (S
0
S

, I
0
I

, R
0
R

) A we have
lim
t
(S(t) S

) = lim
t
(I(t) I

) = lim
t
(R(t) R

) = 0.
Proof. Let v
1
= S S

, v
2
= I I

, v
3
= R R

and v = ( v
1
, v
2
, v
3
). By (1) the
t-derivatives of v
1
, v
2
and v
3
are
_

_
v

1
= (d
1
+I

) v
1
v
1
v
2
S

v
2
+
3
v
3
v

2
= (d
2
+
1
S

) v
2
+v
1
v
2
+I

v
1
+
2
v
3
v
3
=
1
v
2
(d
3
+
2
+
3
) v
3
.
(8)
Consider the functional
V
1
(v) =
1
2
_
w
1
v
2
1
+w
2
v
2
2
+w
3
v
2
3
+w
4
(v
1
+v
2
+v
3
)
2
_
.
The rst derivative of V
1
along the trajectory of a solution of (8) is

V
1
= [ w
1
v
2
+ (d
1
+I

) w
1
+d
1
w
4
] v
2
1
[d
2
w
4
+ (d
2
+
1
S

) w
2
w
2
v
1
] v
2
2
[(d
3
+
2
+
3
) w
3
+d
3
w
4
] v
2
3
[S

w
1
+ (d
1
+d
2
) w
4
I

w
2
] v
1
v
2
[(d
1
+d
3
) w
4

3
w
1
] v
1
v
3
[(d
2
+d
3
) w
4

2
w
2

1
w
3
] v
2
v
3
. (4.9)
Choose w
1
, w
2
and w
4
such that S

w
1
+(d
1
+d
2
)w
4
I

w
2
= 0 and (d
1
+d
3
)w
4

3
w
1
= 0. Then w
2
= kw
4
, where k =
1
I

_
d
1
+d
3

3
S

+d
1
+d
2
_
. The relation in (9)
can be expressed in terms of the previous variables as

V
1
= [ w
1
v
2
+ (d
1
+I

) w
1
+d
1
w
4
] v
2
1
[d
2
w
4
w
2
v
1
] v
2
2
+P (v
2
, v
3
) ,
where
P (v
2
, v
3
) = (d
2
+
1
S

) w
2
v
2
2
[(d
2
+d
3
) w
4

2
w
2

1
w
3
] v
2
v
3
[(d
3
+
2
+
3
) w
3
+d
3
w
4
] v
2
3
.
P(v
2
, v
3
) is negative if the discriminant
= [(d
2
+d
3
) w
4

2
w
2

1
w
3
]
2
4 (d
2
+
1
S

) w
2
[(d
3
+
2
+
3
) w
3
+d
3
w
4
]
Qualitative Behaviour of a Model of an SIRS Epidemic 229
is negative. Since d
2
+
1
S

=

1

2
d
3
+
2
+
3
, we obviously have
< 0 if and only if [(d
2
+d
3
) w
4

2
w
2

1
w
3
]
2
< 4 (d
2
+
1
S

)
w
2
[(d
3
+
2
+
3
) w
3
+d
3
w
4
]
if and only if
_

1
w
3
w
4
(d
2
+d
3

2
k)
_
2
< 4k

2
w
3
w
4
+d
3
(d
2
+
1
S

)
_
if and only if Q
_
w
3
w
4
_
< 0
with
Q(
w
3
w
4
) =
2
1
_
w
3
w
4
_
2
2
1
(d
2
+d
3
+
2
k)
w
3
w
4
+ (d
2
+d
3

2
k)
2
4kd
3
(d
2
+
1
S

) .
The discriminant of Q
_
w
3
w
4
_
is

=
2
1
(d
2
+d
3
+
2
k)
2

2
1
(d
2
+d
3

2
k)
2
+ 4
2
1
kd
3
(d
2
+
1
S

)
= 4
2
1
k [
2
(d
2
+d
3
) +d
3
(d
2
+
1
S

)] .
Let w and w

be the roots of Q
_
w
3
w
4
_
such that w

< w and w > 0. We choose w


3
and w
4
such that max (0, w

) <
w
3
w
4
< w. In this case, < 0, we have also P (v
2
, v
3
) < 0. Set

1
=
d
2
w
4
w
2
=
d
2
k
,
2
=
(d
1
+I

)w
1
+d
1
w
4
w
1
= d
1
+I

+
d
1

3
(d
1
+d
3
)
and = min (
1
,
2
) .
If
v
1
< and v
2
> , then

V
1
< 0. (10)
Let |v| = max (|v
1
| , |v
2
| , |v
3
|) , = min
|v|=
V
1
(v) and
A =
_
v R
3
/ |v| < , V
1
(v) <
_
. We claim that, if v (0) A, then | v (t)| <
for every t. If this be not the case, there exists > 0 such that |v ()| = and
|v (t)| < for every t [0, ) . (11)
Then the denition of implies that V
1
(v ()) since | v ()| = . Now com-
bining (10) with (11) we obtain
d
dt
V
1
(v(t)) < 0. Then V
1
(v (t)) is decreasing and
V
1
(v (t)) < V
1
(v (0)) for all t [0, ) , but V
1
(v (t)) is continuous and therefore
V
1
(v ()) V
1
(v (0)) . Hence V
1
(v (0)) since v (0) A. This is a contradic-
tion. Finally, if v (0) A, we have

V
1
< 0. Therefore V
1
is a Lyapunov functional and this
completes the proof of the Theorem.
5 Stochastic Stability
In this Section we discuss the stochastic stability of the following model
dX = f (X) dt +g (X) dB, (12)
230 El Maroufy, Lahrouz, Leach
where g = (g
1
, g
2
, g
3
) , g
i
, i = 1, 2, 3, are locally Lipchitz functions, B is three-
dimensional brownian motion (see [4] or the references given therein) and
f (X) =
_
_
_
b SI d
1
S +
3
R
SI (d
2
+
1
) I +
2
R

1
I (d
3
+
2
+
3
)
_
_
_.
We denote by L the differential operator associated with (12) , dened for a nonnegative
function, V (t, x) C
1,2
(R R
n
), by
LV =
V
t
+f
T
.
V
x
+
1
2
Tr
_
g
T
.

2
V
x
2
.g
_
,
where
V
x
=
_
V
x
1
,
V
x
2
,
V
x
3
_
T
and

2
V
x
2
=
_

2
V
x
i
x
j
_
ij
, i, j = 1, 2, 3, T and Tr
mean, respectively, transposition and trace.
With the reference to the book by Afnasev et al [1] the following auxiliary results hold.
Theorem 5.1. Suppose that there exist a nonnegative function V (t, x) C
1,2
(R, R
n
) and
two real positive continuous functions, a and b, and constant K > 0 such that, for |x| < K,
a (|x|) V (t, x) b (|x|).
(i) If LV 0, |x| ]0, K[, then the trivial solution of (12) is stable in probability.
(ii) If there exists a continuous function : R
0
+
R
0
+
, positive on R
+
, such that
LV (|x|) ,
then the trivial solution of (12) is asymptotically stable.
The best general reference to stability and related results can be found in [6] and [7].
Proposition 5.1. Let g be a locally Lipchitz function such that supp g

. Then the set


is stable by (12) .
Proof. Let u(0) . Suppose by contradiction that there exists t
0
such that u(t
0
) / ,
then there exists
0
such that u(t) /

for every t [
0
, t
0
] . Since suppg

, it follows
that du = f (u) dt for every t [
0
, t
0
] . This is a contradiction to the invariance of by
(1) .
5.1 Stability of the disease-free point
Theorem 5.2. Let
b
d
1
< d
2
+
1

2
d
3
+
2
+
3
. For any locally Lipchitz function g such that
supp g

and
g
2
1
(S, I, R)
1
_
S
b
d
1
_
2
, where
1
2

1
< d
1
, (13)
the disease-free point
_
b
d
1
, 0, 0
_
is globally asymptotically stable.
Qualitative Behaviour of a Model of an SIRS Epidemic 231
Proof. Let u
1
= S
b
d
1
, u
2
= I and u
3
= R. Consider the functional,
V
2
(u) =
1
2
m
1
u
2
1
+m
2
u
2
+m
3
u
3
, u = (u
1
, u
2
, u
3
) R R

+
R

+
,
where the constants m
1
, m
2
and m
3
are to be chosen in the course of the proof.
f
T
.
V
2
u
= m
1
u
1
_

_
u
1
+
b
d
1
_
u
2
d
1
u
1
+
3
u
3
_
+m
2
_

_
u
1
+
b
d
1
_
u
2
(d
2
+
1
) u
2
+
2
u
3
_
+m
3
[
1
u
2
(d
3
+
2
+
3
) u
3
]
= d
1
m
1
u
2
1
[(d
2
+
1
) m
2

1
m
3
] u
2
[(d
3
+
2
+
3
) m
3

2
m
2
] u
3
m
1
_
u
1
+
b
d
1
_
u
1
u
2
+m
2
_
u
1
+
b
d
1
_
u
2
+
3
m
1
u
1
u
3
= d
1
m
1
u
2
1
m
1
u
2
1
u
2

_
b
d
1
m
1
m
2
_
u
1
u
2
+
3
m
1
u
1
u
3

__
d
2
+
1

b
d
1
_
m
2

1
m
3
_
u
2
[(d
3
+
2
+
3
) m
3

2
m
2
] u
3
.
We choose m
1
such that
b
d
1
m
1
m
2
= 0. We obtain
f
T
.
V
2
u
= d
1
m
1
u
2
1
m
1
u
2
1
u
2

__
d
2
+
1

b
d
1
_
m
2

1
w
3
_
u
2
[(d
3
+
2
+
3
) m
3

2
m
2
] u
3
+

3
d
1
b
m
2
u
1
u
3
.
Since S +I +R
b
d
1
in , we have u
1
0, 0 < u
2

b
d
1
and 0 < u
3

b
d
1
. Hence
f
T
.
V
2
u
d
1
m
1
u
2
1

__
d
2
+
1

b
d
1
_
m
2

1
m
3
_
u
2
[(d
3
+
2
+
3
) m
3

2
m
2
] u
3
.
We choose m
2
and m
3
such that
_
d
2
+
1

b
d
1
_
m
2

1
m
3
> 0 and (d
3
+
2
+
3
) m
3

2
m
2
> 0
which are equivalent to

1

2
(d
3
+
2
+
3
)
m
2
<
1
m
3
<
_
d
2
+
1

b
d
1
_
m
2
. Hence the choice
232 El Maroufy, Lahrouz, Leach
of w
2
and m
3
is possible since

1

2
d
3
+
2
+
3
< d
2
+
1

b
d
1
. Using (13) we have
LV
2
=
V
2
t
+f
T
.
V
2
u
+
1
2
Tr
_
g
T
.

2
V
2
u
2
.g
_
= f
T
.
V
2
u
+
1
2
m
1
g
2
1
d
1
m
1
u
2
1

__
d
2
+
1

b
d
1
_
m
2

1
m
3
_
u
2
[(d
3
+
2
+
3
) m
3

2
m
2
] u
3
+
1
2

1
m
1
u
2
1

_
d
1

1
2

1
_
m
1
u
2
1

b
d
1
__
d
2
+
1

b
d
1
_
m
2

1
m
3
_
u
2
2

b
d
1
[(d
3
+
2
+
3
) m
3

2
m
2
] u
2
3
.
By Theorem 5 the proof is complete.
5.2 Stability of the endemic point
In this Section we use the notation of Theorems 4 and 6 and their proofs.
Lemma 5.1. Let a R

, b, c R and set = b
2
4ac. For all x R
ax
2
+bx +c <

4a
.
Theorem 5.3. If
b
d
1
> d
2
+
1


1

2
d
3
+
2
+
3
for any locally Lipchitz function g such
that g
2
1
(S, I, R)
1
(S S

)
2
, g
2
2
(S, I, R)
2
(I I

)
2
and g
2
3
(S, I, R)

3
(R R

)
2
, where
1
2

1
< d
1
,
1
2

2
<
d
2
k+1
and
1
2

3
<
w
4
Q
(
w
3
w
4
)
4k(d
2
+
1
S

)(w
3
+w
4
)
for
any
w
3
w
4
such that max (0, w

) <
w
3
w
4
< w, then the endemic point is asymptotically stable.
Proof. We have
LV
1
=
V
1
t
+
1
2
(w
1
+w
4
) g
2
1

+
1
2
(w
2
+w
4
) g
2
2
+
1
2
(w
3
+w
4
) g
2
3
= [ w
1
v
2
+ (d
1
+I

) w
1
+d
1
w
4
] v
2
1
[d
2
w
4
w
2
v
1
] v
2
2
+P (v
2
, v
3
) +
1
2
(w
1
+w
4
) g
2
1
+
1
2
(w
2
+w
4
) g
2
2

+
1
2
(w
3
+w
4
) g
2
3
.
Qualitative Behaviour of a Model of an SIRS Epidemic 233
Using the estimations satised by the functions g
1
, g
2
and g
3
we obtain
LV
1
[ w
1
v
2
+ (d
1
+I

) w
1
+d
1
w
4
] v
2
1
[d
2
w
4
w
2
v
1
] v
2
2
+P (v
2
, v
3
) +
1
2

1
(w
1
+w
4
) v
2
1
+
1
2

2
(w
2
+w
4
) v
2
2
+
1
2

3
(w
3
+w
4
) v
2
3
=
_
w
1
v
2
+
_
d
1

1
2

1
+I

_
w
1
+
_
d
1

1
2

1
_
w
4
_
v
2
1

_
d
2
w
4

1
2

2
(w
2
+w
4
) w
2
v
1
_
v
2
2
+P (v
2
, v
3
)
+
1
2

3
(w
3
+w
4
) v
2
3
.
Applying Lemma 3 to the polynomial P (v
2
, v
3
) like the function of v
2
we give, namely
P (v
2
, v
3
) <
w
4
Q
(
w
3
w
4
)
4k(d
2
+
1
S

)(w
3
+w
4
)
v
2
3
, it follows that
LV
1
<
_
w
1
v
2
+
_
d
1

1
2

1
+I

_
w
1
+
_
d
1

1
2

1
_
w
4
_
v
2
1

_
d
2
w
4

1
2

2
(w
2
+w
4
) w
2
v
1
_
v
2
2
+
_
_
1
2

3
(w
3
+w
4
) +
w
4
Q
_
w
3
w
4
_
4k (d
2
+
1
S

) (w
3
+w
4
)
_
_
v
2
3
.
The conditions satised by
1
and
2
lead to
k
1
=
_
d
1

1
2

1
+I

_
w
1
+
_
d
1

1
2

1
_
w
4
w
1
> 0
and
k
2
=
d
2
w
4

1
2

2
(w
2
+w
4
)
w
2
> 0.
Let k < min (k
1
,k
2
) . If |v| = max (|v
1
| , |v
2
| , |v
3
|) < k, we have
w
1
v
2
+
_
d
1

1
2

1
+I

_
w
1
+
_
d
1

1
2

1
_
w
4
> w
1
k +
_
d
1

1
2

1
+I

_
w
1
+
_
d
1

1
2

1
_
w
4
> 0
and d
2
w
4

1
2

2
(w
2
+w
4
) w
2
v
1
> d
2
w
4

1
2

2
(w
2
+w
4
) w
2
k > 0. It follows
234 El Maroufy, Lahrouz, Leach
that
LV
1
<
_
w
1
k +
_
d
1

1
2

1
+I

_
w
1
+
_
d
1

1
2

1
_
w
4
_
v
2
1

_
_
d
2
w
4

1
2

2
(w
2
+w
4
) w
2
k
_
_
v
2
2
+
_
_
1
2

3
(w
3
+w
4
) +
w
4
Q
_
w
3
w
4
_
4k (d
2
+
1
S

) (w
3
+w
4
)
_
_
v
2
3
.
The condition satised by
3
implies that
1
2

3
(w
3
+w
4
) +
w
4
Q
_
w
3
w
4
_
4k (d
2
+
1
S

) (w
3
+w
4
)
< 0.
Finally, when we apply Theorem 5, Theorem 7 follows.
6 Numerical Examples
0 2 4 6 8 10 12 14 16 18 20
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Times (years)
D
e
n
s
it
y
(
S
,I,R
)
(a)
0.5
0.6
0.7
0.8
0.9
1
0
0.02
0.04
0.06
0.08
0.1
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
D
ensity of susceptibles
(b)
D
ensity of infectives
D
e
n
s
i
t
y

o
f

r
e
m
o
v
a
l
s
Figure 6.1: The stochastic model and its deterministic model (black). (a) the density of the three
classes of individuals (S: blue, I: red,R: green) versus time, (b) the dynamic behaviour of
(S(t), I(t), R(t))(red ). Here b = 0.5, = 0.7, d
1
= 0.9, d
2
= 0.7, d
3
= 0.5, = 0.2, =
0.1, = 0.6 and we have R
0
= 0.5 and E = (0.55, 0, 0). (For interpretation of the references to
colour in the legend of this gure the reader is referred to the electronic version of this article.)
Qualitative Behaviour of a Model of an SIRS Epidemic 235
0 2 4 6 8 10 12 14 16 18 20
0
0.5
1
1.5
2
2.5
Time (years)
D
e
n
s
it
y
(
S
,I,R
)
(a)


0.8
1
1.2
1.4
1.6
1.8
2
2.2
0
0.2
0.4
0.6
0.8
1
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Density of susceptibles
(b)
Density of infectives
Figure 6.2: The stochastic model and its deterministic model (black). (a) the density of the
three classes of individuals (S: blue,I: red,R: green ) versus time, (b) the dynamic behavior of
(S(t), I(t), R(t)) (red). Here b = 0.5, = 0.7, d
1
= 0.1, d
2
= 0.4, d
3
= 0.2, = 0.6, =
0.5, = 0.5 and we have R
0
= 4.6 and E = (1.07, 0.78, 0.39). (For interpretation of the ref-
erences to colour in the legend of this gure the reader is referred to the electronic version of this
article.)
In this Section as an example of random perturbation we adopt the idea of Mukherje
in [9]. We allow the random perturbations of the variables (S, I, R) around the disease-
free point E
0
= (b/d
1
, 0, 0) if the constant R
0
=
b
d
1
d
2
+
1

2
d
3
+
2
+
3
< 1 and otherwise
around the endemic positive equilibrium point E

= (S

, I

, R

) in the case when it is


asymptotically stable. We assume that the random perturbations are a type of white noise
proportional to the distance of S, I and R from values of the equilibria. So the system (12)
becomes
_

_
dS = b SI d
1
S +
3
R +
1
(S E
1
)dW
1
t
dI = SI (d
2
+
1
) I +
2
R +
2
(I E
2
)dW
2
t
dR =
1
I (d
3
+
2
+
3
) R +
3
(R E
3
)dW
3
t
,
(12)
where (E
1
, E
2
, E
3
) is equal to E
0
or E

,
i
, i = 1, 2, 3, are real positives constants, W
1
t
,
W
2
t
and W
3
t
are standard Wiener processes independent from each other (Stroock and
Varadham [10]).
In the following we present some numerical simulations of two examples which validate the
theoretical results obtained in this paper. For simplicity we choose the initial conditions:
S
0
= 1, I
0
= 0.01 and
1
= 0.04,
2
= 0.01,
3
= 0.1 are supposed to satisfy the
conditions of Theorem 6 and Theorem 7. The values of the other parameters are explained
in each example.
It can be seen from Figure 6(a) and Figure 6(b) that, when R
0
< 1, it increases away from
the disease-free point. In this case the endemic equilibrium E

is asymptotically stable.
We can also see that the trajectory of the stochastic process remains close to the trajectory
of its deterministic analogue during a nite time interval. We should note that the path of
236 El Maroufy, Lahrouz, Leach
stochastic process eventually leaves the trajectory and is absorbed in the equilibrium point
(Figure 1(b) and Figure 2(b)).
7 Conclusion
The dynamic behaviour of deterministic as well as the stochastic model for the spread
of an SIRS epidemic are presented in this paper. We established the same properties of
stability. The numerical results also indicate that there exist positive-stable disease-free and
endemic equilibria. Moreover for future research it should be feasible to use the stochastic
differential equation, (12), with a general diffusion term g and to nd a suitable Lyapunov
functional for unconditional stability of the positive equilibrium of the model, (1), of the
SIRS epidemic. Another possible direction for future research is to consider how control
strategies may be devised. Finally, taking into account the available statistical data, we can
use the stochastic diffusion inference to estimate the parameters of the model.
References
[1] V. N. Afanasev, V. B. Kolmanowski and V. R. Nosov, Mathematical Theory of Global
Systems Design, Kluwer, Dordrecht, 1996.
[2] N. T. J. Bailey, The Mathematical Theory of Infectious Diseases, 2nd ed. Grifn,
London, 1975.
[3] E. Beretta and Y. Takeuchi, Global stability of a SIR epidemic model with time delay,
J. Math. Biol. 33 (1995), 250-260.
[4] F. L. Gregory, Introduction to Stochastic Processes, Chapman and Hall, 1995.
[5] Z. Jin, M. Zhien and H. Maoan, Global stability of an SIRS epidemic model with
delays, Acta Mat. Sci. 26 B (2006), 291306.
[6] V. Kolmanvskii and V. R. Nsov, Stability of Functional Differential Equations, Aca-
demic, New York, 1986.
[7] V. B. Kolmanvskii and A. D. Myshkis, Applied Theory of Functional Differential
Equations, Kluwer, Boston, 1992.
[8] W. Ma, Y. Takeuchi, T. Hara and E. Beretta, Permanence of an SIR epidemic model
with distributed time delays, Tohoku Math. J. 54 (2002), 581-591.
[9] D. Mukherjee, Stability analysis of a stochastic model for prey-predator system with
disease in the prey, Nonlin. Anal.: Mod. Control 8 (2003), 8392.
Qualitative Behaviour of a Model of an SIRS Epidemic 237
[10] D. W. Stroock and S. R. S. Varadhan, Multidimensional Diffusion Processes,
Springer-Verlag, Heidelberg, 2006.
[11] H. Thieme, Persistence under relaxed point-dissipativity, SIAM J. Appl. Math. 24
(1993), 407435.
[12] E. Tornatore, S. Maria Buccellato and P. Verto, Stability of a stochastic SIR system,
Physica A: Stat. Mech. 354 (2005), 111126.
[13] Tailei Zhang and Zhidong Teng, Permanence and extinction for a nonautonomous
SIRS epidemic model with time delay, Appl. Math. Mod. 33 (2009), 10581071.
Hamid El Maroufy, obtained his PhD (2001) degree in Probabil-
ity and Statistics at University Sidi Mohamed Ben Abdellah in Mo-
rocco. Since then, he has occupied academic position as lecturer and
researcher in Faculty of sciences and Technology of Beni-Mellal at
the University of Cady Ayyad Marrakech. Since 2007, he combines
a position as lecturer and coordinator of some research programs in
epidemic modelling and supervisor of MS and PhDs theses in this
eld. His PhD thesis was in the area of stochastic epidemic. Later on, he continued to
be interested in how probability analysis can be used in the areas of both in Kinetic and
epidemics models. He is, also, author/co-author of several papers and preprints in the eld
of stochastic epidemic models.
Adil Lahrouz, obtained his Master (2008) degree in mathemat-
ical sciences at Abdel Malek University in Morocco. Since 2009,
he combines the academic position in secondary education and stu-
dent researcher preparing his Ph. D degree in the deterministic and
stochastic stability of the epidemics models in Sidi Mohamed Ben
Abdellah in Morocco University.
Peter Leach is a Professor Emeritus of the University of
KwaZulu-Natal, South Africa, and is currently a Visiting Professor at
the University of Cyprus. His research interests are the development
and application of symmetry and singularity analyses to the resolu-
tion of differential equations, be they ordinary or partial, which arise
in divers applications including Finance, Medicine, Epidemiology,
Cosmology and General Relativity, Ecology, Classical and Quantum
238 El Maroufy, Lahrouz, Leach
Mechanics. He has also done work in Structural Chemistry in connection with the forma-
tion of organometallic complexes. He has served as a member of several editorial boards
including currently AMIS and has refereed for many journals.

You might also like