Mathematical Epidemiology of Typhoid Fever

A project report submitted to the course instructor in partial fulfilment of the

requirements for the course on

Mathematical Modelling
(MA539)

Submitted by

MR. SONU
Research Scholar
Roll No.- 2021MA23

DEPARTMENT OF MATHEMATICS
INDIAN INSTITUTE OF TECHNOLOGY PATNA
BIHTA- 801106, PATNA, BIHAR, INDIA

November 2020
 Abstract
The main goal of this mini-project is to study the mathematical modelling and stability analysis of
typhoid disease. An epidemic compartmental model is used to formulate a system of ordinary differ-
ential equations. The stability theory of differential equations is used to investigate the qualitative
behaviour of the system. The basic reproduction number that acts as a threshold value is obtained
by using the next generation matrix. The local and global stability for the disease free and endemic
equilibrium points is then discussed with the help of basic reproduction number. The numerical
simulation is also done using MATLAB, which suggests that both protection and treatment are
effective to control the transmission of typhoid disease.

Keywords: Basic reproduction number, next generation matrix, stability theory, typhoid disease.

Contents
Abstract 1

1 Introduction 2

2 Description and Formulation of Model 2

2.1 Model Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2 Model Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.3 Model Flow Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.4 Model Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

3 Analysis of the Model 4

3.1 Invariant Region . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.2 Existence of the Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.3 Stability Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.3.1 Disease Free Equilibrium (DFE) . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.3.2 Endemic Equilibrium (EE) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.3.3 Basic Reproduction Number (R0 ) . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.3.4 Stability Analysis of DFE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.3.5 Stability Analysis of EE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.3.6 The Case When R0 = 1 (Exchange of Stability) . . . . . . . . . . . . . . . . . 10
3.4 Numerical Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

4 Conclusion 12

Bibliography 12

Acknowledgement
I would like to extend my sincere thanks to the course instructor Dr. P. K. Srivastava,
Associate Professor, Department of Mathematics, IIT Patna for teaching the course on
Mathematical Modelling (MA539). His wonderful teaching & research skills made him a visionary
and inspiring professor and an ideal supervisor for me.

I also thank my senior colleague Ms. Tanuja Das for helping me in understanding some typical
points from the numerical simulation part.
1 Introduction
Typhoid fever is a serious disease, found only in human and spreads through contaminated food and
water causing high fever, abdominal pain and diarrhoea. It occurs due to systemic infection mainly
by the Salmonella typhi bacteria. According to the most recent (2019) WHO estimates, between
11 and 21 million cases and 128,000 to 161,000 typhoid-related deaths occur annually worldwide.
The disease is a major problem in most of the developing countries where there is lack of access to
clean water, proper sanitation systems, and proper health-care facilities. In figure 1, we can see that
typhoid is strongly endemic in most of the African & Asian countries.

Figure 1: Incidence of typhoid fever in different countries

Various models have been developed to investigate the dynamics of typhoid fever along with the
effects of some intervention plans against the spread of typhoid infections. For instance, the com-
partmental model proposed by Nthiiri et al. [1] which was used to investigate the dynamics of the
disease incorporating protection against infection.

Here we have taken the same epidemic compartmental model with some important modifications
while calculating and investigating the stability of its equilibria both at disease-free equilibrium
(DFE) and at endemic equilibrium (EE).

2 Description and Formulation of Model

The total population size at any time t is subdivided into four compartments; P (t) is the protected
class, S(t) is the susceptible class, I(t) is the infected class, and T (t) is the treated class. Therefore,
the total population size at any time t is given by N (t) = P (t) + S(t) + I(t) + T (t).

2.1 Model Parameters

The description of parameters used in this model is listed below;
1. αΛ is the per capita rate of recruitment for the protected individuals.
2. (1 − α)Λ is the per capita rate of recruitment for the susceptible individuals.

3. α is the proportion of progress to the per capita rate of recruitment.

2
 4. γ is the rate at which protected individuals are becoming susceptible.
5. λ is the rate at which susceptible individuals are being infected (effective force of infection1 ).
6. β is the rate of treatment.

7. δ is disease induced mortality rate.

8. µ is natural mortality rate.

2.2 Model Assumptions

1. All the parameters are positive.
2. No treatment failure. There is no re-infection once an individual is treated.
3. As µ is the natural death rate of individuals so it is considered in each of the compartments.

4. Since we are dealing with a population, we expect that the size of all sub-population classes
is non-negative at any time t; that is, all variables P (t), S(t), I(t), and T (t) are non-negative
for any time t (though it can be proved but for making our life simpler we assumed it).
5. The total population at any time t is given by N = P + S + I + T with non-negative initial
conditions P (0) = P0 , S(0) = S0 , I(0) = I0 , and T (0) = T0 .

2.3 Model Flow Diagram

The flow diagram of the compartmental model is depicted in the figure 2 below;

Figure 2: Flow diagram of the compartmental model

1 The πθ(1−υ)I
effective force of infection λ is given by, λ = N
,
where π is the probability rate of acquiring typhoid, θ is contact rate of the infection, and υ is the effective rate of
protection against infection.

3
2.4 Model Equations
Considering all the variables, parameters, and assumptions the general form of this model is given
by the following system of ordinary differential equations;
dP
= αΛ − (γ + µ)P
dt
dS
= (1 − α)Λ + γP − (λ + µ)S
dt (1)
dI
= λS − (δ + β + µ)I
dt
dT
= βI − µT
dt
These equations can easily be comprehended by putting the flow diagram (figure 2) alongside. It
is clear that the inward and outward arrows for a compartment corresponds to the plus and minus
signs respectively in the equation of that compartment. It can easily be noted that our model is an
open, deterministic, and dynamic model.

3 Analysis of the Model

The implementation of any mathematical model largely depends on whether the obtained system of
equations are well posed or not. So before jumping into the stability analysis, let us first check the
epidemiological and mathematical well-posedness of our model.

3.1 Invariant Region

We have to obtain the feasible region Γ such that P (t), S(t), I(t), and T (t) ∈ Γ ⊂ R4+ , this feasible
region is known as the invariant region.

Theorem 3.1.1 The variables P (t), S(t), I(t), and T (t) are bounded for all t or equivalently, the
feasible solution set of the model enters and remains in the region Γ = {(P, S, I, T ) ∈ R4+ : N ≤ Λ
µ }.
Proof: As we have N (t) = P (t) + S(t) + I(t) + T (t), differentiating N both sides with respect to t
we have,
dN dP dS dI dT
= + + + (2)
dt dt dt dt dt
By putting corresponding values from (1) into (2), we get

dN
= αΛ − (γ + µ)P + (1 − α)Λ + γP − (λ + µ)S + λS − (δ + β + µ)I + βI − µT
dt
dN
=⇒ = Λ − µ(P + S + I + T ) − δI
dt
dN
=⇒ = Λ − µN − δI (3)
dt
In absence of disease induced mortality rate i.e., δ = 0, equation (3) becomes,

dN
≤ Λ − µN
Z dt Z
dN
=⇒ ≤ dt
Λ − µN
1
=⇒ − ln(Λ − µN ) ≤ t + c1
µ

4
Taking anti-logarithm and arranging the constant of integration, we have;

Λ − µN ≥ e−µ(t+c1 )

or,
Λ − µN ≥ Ce−µt , where C = e−µc1 (4)
Applying the initial condition, N (0) = N0 in equation (4) we get C = Λ − µN0 , which leads to,

Λ − µN ≥ (Λ − µN0 )e−µt
=⇒ µN ≤ Λ − Λe−µt + µN0 e−µt
Λ Λ
=⇒ N ≤ − e−µt + N0 e−µt (5)
µ µ
As t → ∞, the equation (5) gives us,
Λ Λ
N≤ or 0≤N ≤
µ µ

Thus we have obtained the invariant region Γ = {(P, S, I, T ) ∈ R4+ : N ≤ Λ

µ } and therefore, our
model is well posed epidemiologically and mathematically. Hence, it is sufficient to analyse the
dynamics of the model in Γ region. 

3.2 Existence of the Solution

For doing stability analysis of any system of differential equations, it is highly recommanded and
important to first check the existence of the solution. Therefore, here we shall look for the existence
of the solution of our model.

Theorem 3.2.1 Solution of the system (1) with defined initial conditions exist in Γ region i.e., the
solution of our model exists for all t and will remain in the feasible region.
Proof: Writing the right hand sides of all the four equations of our system (1) as,

f1 (P, S, I, T ) = αΛ − (γ + µ)P
f2 (P, S, I, T ) = (1 − α)Λ + γP − (λ + µ)S
f3 (P, S, I, T ) = λS − (δ + β + µ)I
f4 (P, S, I, T ) = βI − µT

For the feasible region Γ, Derrick and Grossman’s theorem2 states that the system (1) have unique
∂fi
solution if ∂x i
; i, j = 1, 2, 3, 4 are continuous and bounded in Γ. Here x1 = P , x2 = S, x3 = I, and
x4 = T . Now, let us verify the continuity and boundedness;
For f1 :
       
 ∂f1   ∂f1   ∂f1   ∂f1 
 ∂P  = | − (γ + µ)| < ∞,  ∂S  = 0 < ∞,  ∂I  = 0 < ∞,  ∂T  = 0 < ∞.
       

For f2 :
         
 ∂f2   ∂f2   ∂f2   πθ(1 − υ)   ∂f2 
 ∂P  = |γ| < ∞,  ∂S  = | − (λ + µ)| < ∞,
=− S  < ∞,  ∂T  = 0 < ∞
      
 ∂I   N
2 Derrick and Grossman’s theorem, 1976
dx(t)
For any system dt = f(t, x) defined within some bounded region R, suppose f(t, x) satisfies the Lipschitz condition
||f(t, x1 ) − f(t, x2 )|| ≤ k||x1 − x2 || in R, for some positive constant k, ∃ δ ≥ 0 such that the system has a unique
vector solution x(t) in the interval t − t0 ≤ δ.
∂fi
Note: This condition can be replaced by the sufficient requirement that ∂x ; i, j = 1, 2, 3, 4, ... are continuous and
i
bounded in R and this is what we are using here.

5
 For f3 :
       
 ∂f3   ∂f3   ∂f3   ∂f3 
 ∂P  = 0 < ∞,  ∂S  = |λ| < ∞,  ∂I  = | − (δ + β + µ)| < ∞,  ∂T  = 0 < ∞
       

For f4 :
       
 ∂f4   ∂f4   ∂f4   ∂f4 
 ∂P  = 0 < ∞,  ∂S  = 0 < ∞,  ∂I  = |β| < ∞,  ∂T  = | − µ| < ∞
       

∂fi
Thus all the partial derivatives ∂x i
; i, j = 1, 2, 3, 4 are continuous and bounded in Γ. Hence by
using Derrick and Grossman’s theorem, we can say that solution for this model exists and unique.
Therefore, now it is legitimate to discuss the stability analysis of the model. 

3.3 Stability Analysis

In this section, we shall analyse the model quantitatively to discuss the stability of its equilibria
both at disease-free equilibrium (DFE) and at endemic equilibrium (EE), for that we need to find
both the equilibrium (critical) points of system (1).

3.3.1 Disease Free Equilibrium (DFE)

The disease free equilibrium points (DFE) are the solutions of steady state of the model in the
absence of the disease. Therefore, for obtaining disease free equilibrium, consider the steady state
of the system (1) i.e.,

αΛ − (γ + µ)P = 0
(1 − α)Λ + γP − (λ + µ)S = 0
λS − (δ + β + µ)I = 0
βI − µT = 0

Since there is no disease i.e., there is no infection and treatment and therefore, we set I = 0, T = 0
in the above system of equations. From the first equation we have,
αΛ
P = (6)
γ+µ
Now solving the second equation for P, we have;

(λ + µ)S − (1 − α)Λ
P = (7)
γ

Comparing equation (6) & (7) and solving it for S by setting λ = 0 (for no infection), we get;

(µ + γ − αµ)Λ
S= (8)
µ(µ + γ)
 
αΛ (µ+γ−αµ)Λ
Therefore, E 0 = γ+µ , µ(µ+γ) , 0, 0 is the disease free equilibrium points (DFE).

3.3.2 Endemic Equilibrium (EE)

The endemic equilibrium E ∗ = (P ∗ , S ∗ , I ∗ , T ∗ ) occurs when the disease found and difficult to get
rid of i.e., the disease persists in the community. For obtaining E ∗ , we need to solve the steady
state of the model without setting any variable equal to zero.

6
Solving the first equation of the steady state, we get;
αΛ
P∗ = (9)
γ+µ
From the third equation, putting value of λ we get,
 
πθ(1 − υ)I
S − (δ + β + µ)I = 0
N

N (δ + β + µ)
=⇒ S ∗ = (10)
πθ(1 − υ)

Solving the fourth equation we get,

βI ∗
T∗ = (11)
µ
Now, again taking the third equation,
 
πθ(1 − υ)I
S − (δ + β + µ)I = 0
N
N (δ + β + µ)
=⇒ S =
πθ(1 − υ)
N (δ + β + µ)
=⇒ N − P − I − T =
πθ(1 − υ)

Using equation (9) and (11), we get,

αΛ βI ∗ N (δ + β + µ)
N− − I∗ − =
γ+µ µ πθ(1 − υ)
 
µ αΛ N (δ + β + µ)
=⇒ I ∗ = N− − (12)
µ+β γ+µ πθ(1 − υ)

Ans therefore, using equation (12) in equation (11) we have,

 
β αΛ N (δ + β + µ)
T∗ = N− − (13)
µ+β γ+µ πθ(1 − υ)

Hence, the endemic equilibrium is,

   !
αΛ N (δ + β + µ) µ αΛ N (δ + β + µ) β αΛ N (δ + β + µ)
E∗ = , , N− − , N− −
γ+µ πθ(1 − υ) µ+β γ+µ πθ(1 − υ) µ+β γ+µ πθ(1 − υ)

3.3.3 Basic Reproduction Number (R0 )

As we have obtained the equilibria, so it’s high time to jump into the stability analysis. But before
doing that, let us first discuss an important term called as the basic reproduction number, denoted
by R0 . It is the average number of secondary cases a typical infectious individual will cause in a
completely susceptible population. Refer the diagram (figure 3) for understanding it easily.

Here we shall obtain the basic reproduction number R0 , which is the threshold parameter that gov-
erns the spread of the disease. To obtain R0 , we shall use the usual method i.e., the next generation
matrix method and R0 is the spectral radius of the next generation matrix.

7
 Figure 3: Basic Reproduction Number

Consider the primary (new) infected group λS and secondary infected group (δ + β + µ)I in the
third equation of the model (1) i.e., dI
dt = λS − (δ + β + µ)I. Writing the primary infected group as,
" #
πθ(1 − υ)I
f = [λS] = S
N
" #
∂f πθ(1 − υ)
=⇒ F = = S (14)
∂I N

Since at the beginning (before spread of disease), there is disease free environment and therefore
(from equation 8) we have S = (µ+γ−αµ)Λ Λ
µ(µ+γ) . Further P + S = N gives us N = µ , putting these values
of S and N in equation (14) we get,
" #
πθ(1 − υ)(µ + γ − αµ)
F = (15)
(µ + γ)

Now for obtaining the matrix V , we consider the secondary infected group g = (δ + β + µ)I;
∂g
V = = [δ + β + µ]
∂I
 
1
=⇒ V −1 = (16)
δ+β+µ

The basic reproduction number is spectral radius of the next generation matrix i.e., R0 = ρ(F V −1 ),
using equation (15) and (16) we get,
 !

−1
 πθ(1 − υ)(µ + γ − αµ)
R0 = ρ [F V ] = ρ
(µ + γ)(δ + β + µ)
πθ(1 − υ)(µ + γ − αµ)
=⇒ R0 = (17)
(µ + γ)(δ + β + µ)
It the required value of R0 , which acts as a threshold value for governing the spread of the disease.

8
3.3.4 Stability Analysis of DFE
Local Stability
From the system of the equations (1), we can obtain the Jacobian matrix J for the correspond-
ing linearized system,
 
−(γ + µ)  0  0 0
− µ + πθ(1−υ)I − πθ(1−υ)S
 
 γ N N 0 
J =   (18)
πθ(1−υ)I πθ(1−υ)S

 0 N N − (δ + β + µ) 0 
0 0 β −µ
For investigating the stability of DFE, we shall now compute the Jacobian matrix J at DFE,
 
−(γ + µ) 0 0 0
 γ −µ − πθ(1−υ)(µ+γ−αµ)
µ+γ 0 
JE 0 =  (19)
 
πθ(1−υ)(µ+γ−αµ)
−

 0 0 µ+γ (δ + β + µ) 0 
0 0 β −µ

By solving |JE 0 − λI| = 0 (λ denotes eigenvalues & I is the identity matrix of order 4), we can
get all the four eigenvalues of JE 0 as follows (these can also be obtained by using MATLAB);

λ1 = −(γ + µ), λ2 = −µ, λ3 = −µ, and

πθ(1 − υ)(µ + γ − αµ)
λ4 = − (δ + β + µ)
µ+γ
or , λ4 = −(δ + β + µ)(1 − R0 )

Now as per the model assumptions made in section (2.2), it can be seen that the first three
eigenvalues λ1 , λ2 , λ3 are negative but negativity of the fourth eigenvalue λ4 depends on the
value of R0 and if R0 < 1 then only λ4 is negative.

Hence the disease free equilibrium E 0 of the model (1) is asymptotically stable whenever R0 < 1.
Global Stability
To establish the global asymptotic stability of DFE, here we have adopted the method of Lya-
punov functions.

1
We define L = δ+β+µ I such that,

• L(E 0 ) = 0 and L > 0 otherwise (∵ the parameters are positive).

• By simple calculations3 , we also have

dL 1 1 dL


dt = δ+β+µ I = δ+β+µ λS − (δ + β + µ)I = (R0 − 1)I < 0 i.e., dt < 0 if R0 < 1.

Thus the disease free equilibrium E 0 is globally asymptotically stable for R0 < 1.

3.3.5 Stability Analysis of EE

Let d1 = − πθ(1−υ)I
N and rewriting the Jacobian matrix from equation (18) we have,

−(γ + µ) 0 0 0
 
 γ d1 − µ − πθ(1−υ)S
N 0 
J = πθ(1−υ)S
 (20)
 0 −d1 N − (δ + β + µ) 0 
0 0 β −µ
3 Putting πθ(1−υ)I (µ+γ−αµ)Λ Λ
λ= N
, S= µ(µ+γ)
, N = µ
and solving.

9
Computing this Jacobian matrix (20) at the endemic equilibrium E ∗ we get,
 
−(γ + µ) 0 0 0
 γ d1 − µ −(δ + β + µ) 0 
JE ∗ =   (21)
 0 −d1 0 0 
0 0 β −µ

Using MATLAB, we have obtained the eigenvalues for JE ∗ ,

λ1 = −µ, λ2 = −(γ + µ),

√
d1 µ h
λ3 = − − , and
2 2 √2
d1 µ h
λ4 = − +
2 2 2
where, h = 2d1 µ + 4d1 β + 4d1 δ + µ2 + d21 and we have also obtained the determinant and trace
of JE ∗ , given by det(JE ∗ ) = −d1 µ(γ + µ)(δ + β + µ), trace(JE ∗ ) = d1 − 3µ − γ respectively. The
value of d1 in terms of R0 4 is given by,
µ
d1 = (1 − R0 )(δ + β + µ).
µ+β

It is very much clear that d1 < 0 whenever R0 > 1 (because all other parameters are positive).
It should also be noted that, for R0 > 1 we have det(JE ∗ ) = −d1 µ(γ + µ)(δ + β + µ) > 0 and
trace(JE ∗ ) = d1 − 3µ − γ < 0 (as d1 < 0).

Let us now analyse the eigenvalues with R0 > 1,

• Case 1: if h > 0, the first three eigenvalues λ1 , λ2 , and λ3 are negative. Since det(JE ∗ ) > 0
therefore, the fourth eigenvalue λ4 has to be negative. Thus all the four eigenvalues are negative
in this case.

• Case 2: if h < 0, in this case the real parts of all the four eigenvalues are negative.
Hence the endemic equilibrium E ∗ of the model (1) is asymptotically stable whenever R0 > 1.

3.3.6 The Case When R0 = 1 (Exchange of Stability)

It can be noted that R0 = 1 is the point where the disease free equilibrium and endemic equilibrium
exchange the stability behaviour. This phenomenon of change of stability is known as the forward
bifurcation.
4 Value of d in terms of R : For analysing the eigenvalues we need to simplify d by using the value of I ∗ from
1 0 1
the endemic equilibrium,
πθ(1 − υ)I ∗
d1 = −
N
 
πθ(1 − υ) µ αΛ N (δ + β + µ)
=⇒ d1 = − × N− −
N µ+β γ+µ πθ(1 − υ)

Λ
Putting N = µ
and simplifying it further we have,
µ
d1 = (1 − R0 )(δ + β + µ).
µ+β

10
3.4 Numerical Simulations
Numerical simulation are important to study the behaviour of system graphically. Here we have
made an attempt to understand the effect of protection and treatment on the transmission dynamics
of typhoid fever. The simulation is carried out using MATLAB software (ode45 solver) based on the
parameter values [2] listed in table (1).

Parameter Description Value

Λ Recruitment rate 2.2
α Proportion parameter 0.8
µ Natural death rate 0.022
λ Effective force of infection 0.0001
δ Disease induced death rate 0.05
γ Loss of protection rate 0.01
β Treatment rate 0.9
θ Contact rate of the infection 0.003
υ Effective rate of protection against infection 0.7
π Probability rate of transmission 0.015

Table 1: Parameters values of the model

Using these parameters, we first solved the system of differential equation of model (1) with the
help of ode45 solver, a simulation tool in MATLAB software and then plotted the solution curves.
Here we have considered the non-negative initial conditions P (0) = P0 = 100, S(0) = S0 = 50,
I(0) = I0 = 30, and T (0) = T0 = 0.

Figure 4: The graph of P , S, I and T against t

Figure (4) shows that with high success of protection and treatment the burden of the disease can
be reduced. An increment in β and decrement in γ will lead us to a sharp hike in the graph of
treated individuals and the infection in the population decreases sharply over time.

11
4 Conclusion
In this project, we have considered a deterministic compartmental model for the transmission dynam-
ics of typhoid fever disease with proper model assumptions. The epidemiological and mathematical
well-posedness of the model is established by proving the boundedness, existence and uniqueness
of the solution. The equilibria points of the model are obtained and their stability condition is
established with the help of the basic reproduction number R0 .

The disease free equilibrium of the model is locally and globally asymptotic stable whenever R0 < 1
while the endemic equilibrium is locally asymptotically stable if R0 > 1. The result that we obtained
from the numerical simulation revealed that both protection and treatment are effective to control
the transmission of typhoid disease. Hence an increase in protection and treatment leads to low
disease prevalence in a population, while with low protection there is high contact rate and that
lead us to a high disease prevalence in the population.

References
[1] Nthiiri et al., Mathematical Modelling of Typhoid Fever Disease Incorporating Protection against
Infection, BJMCS, 14(1), 2016, PP 2-9.
[2] Mamo S. Wameko et al., Mathematical Model for Transmission Dynamics of Typhoid Fever,
IOSRJM, 2019, PP 88-99.

[3] O. J. Peter et al., Series Solution of Typhoid Fever Model using Differential Transform Method,
Malaysian Journal of Computing, 3 (1), 2018, PP 67-80.
[4] Lectures delivered by Dr. Nitu Kumari and Dr. P. K. Srivastava during an online two-week
FDP on Biomathematics, recently organised by the University of Delhi.

?????

12