University of Malawi

School of Natural and Applied Sciences

Department of Mathematical Sciences

Master of Science in Mathematical Sciences

MAT629: Mathematical Biology − Reference Notes

Prepared by:
Milliward Maliyoni, PhD (Appl. Math − Math Bio)

October 2024
2 Mathematical Modelling: Introduction and Applications

1 Review of Ordinary Differential Equations (ODEs)

1.1 Motivation
Differential equations (DEs) arise naturally when we attempt to describe real-life problems mathemati-
cally. This makes DEs very applicable directly to our daily lives. DEs provide the thread and needle with
which one works out problems from such diverse fields as physics, chemistry, electrical engineering, biol-
ogy, economics, finance just to mention a few. In this course, we shall use differential (ordinary) equations
to discuss infectious diseases models that arise in biology; in particular those dealing with transmission
and control of infections (communicable) diseases. But before we get to that, we review some of the
methods that are used to solve first order ordinary differential equations.

1.2 Definition and Terminology

In calculus, given a function y = f (x), the derivative

dy
= f 0 (x)
dx
is a function of x itself and is found by some appropriate rule. In differential equations, given an equa-
dy
tion = f (x) we want to find a function y = f (x) that satisfies the given equation.
dx

1.2.1 Differential Equation (DE)

This is an equation containing the derivatives or differentials of one or more dependent variables with
respect to one or more independent variables. Examples of DEs are:

dy ∂ 2y ∂ 2u
= −2xy and = − .
dx ∂x2 ∂t2

1.2.2 Classification of DEs

(a) Classification by Type

i. If an equation contains only ordinary derivatives of one or more dependent variables with re-
spect to a single independent variable, then it is said to be an ordinary differential equation
(ODE). For example,

dy dy
+ 10y = e2x and − xy = 0 are ODEs.
dx dx
ii. An equation involving the partial derivatives of one or more dependent variables with respect
two or more independent variables, is called a partial differential equation (PDE). For example,

∂t ∂u ∂u ∂u
=− and x +y =u are PDEs.
∂y ∂x ∂x ∂y
(b) Classification by Order
The order of a DE is the order of the highest derivative in a DE. For example,

M. Maliyoni, PhD (Appl. Math − Math Bio)

3 Mathematical Modelling: Introduction and Applications

d2 y dy
i. + − y = 0 is a second-order DE.
dx dx
dy
ii. (y − x) + 4x = 0 is a first-order DE.
dx
A general nth − order ODE is often represented by

dn y
 
dy
F x, y, , · · · , n = 0 (1)
dx dx

or equivalently F x, y, y 0 , · · · , y (n) = 0.

The following property is a special case of (1).

(c) Classification by Linearity (as linear or nonlinear)

A DE is said to be linear if it can be written in the form

dn y dn−1 y dy
an (x) n
+ a n−1 (x) n−1
+ · · · + a1 (x) + a0 (x)y = g(x). (2)
dx dx dx
A similar definition applies to PDEs.

Note: A linear DE (2) is characterized by two properties and these are;

i. The dependent variable y and all its derivatives are of degree 1 i.e. the power of each term
containing y is 1.
ii. Each coefficient depends only on the independent variable x. In other words, all coefficients are
functions of x.

For example

i. (y + x)dx − 6xdy = 0 is a linear first-order ODE.

d2 y
ii. + y 2 = 0 is a nonlinear second-order ODE.
dx2

1.3 Solution Methods for First Order ODEs

A solution of a DE given by (1), is a function y = f (x) that possesses at least n derivatives and satisfies
equation (1) for every x in I.

The situation is simpler for first-order ODEs since we only have the first derivative of y.
Recall that a first-order ODE takes the form
dy
a1 (x) + a0 (x)y = g(x). (3)
dx
where a1 (x) 6= 0. Dividing (3) through by a1 (x) leads to

M. Maliyoni, PhD (Appl. Math − Math Bio)

4 Mathematical Modelling: Introduction and Applications

dy
+ p(x)y = q(x) (4)
dx
a0 (x) g(x)
where p(x) = and q(x) = . We wish to discuss methods for solving equation (4).
a1 (x) a1 (x)

1.3.1 Direct Integration

dy
If p(x) = 0 in (4), then the resulting differential equation = q(x) may be solved by integrating both
dx
sides directly with respect to (w.r.t) x. Thus we obtain
Z Z Z
dy
dx = q(x) dx =⇒ y = q(x) dx + C.
dx
dy
Example 1 Solve the equation = (x + 1)2 .
dx
Solution
We expand the right hand side of the equation and then integrate both sides to obtain
Z
y = (x2 + 2x + 1) dx
x3
= + x2 + x + C.
3
Example 2 Solve the equation y 0 = cos 2x.

Solution
We integrate both sides of the equation and apply substitution technique. Hence,
Z
y = cos 2x dx (5)

Now, let u = 2x =⇒ du = 2 dx. Thus, equation (5) becomes

Z
y = cos 2x dx
Z Z
du 1
= cos u = cos u du
2 2
1
= sin u + C
2
1
= sin 2x + C.
2

Exercise
dy 2
Solve the equation = xe−x .
dx
1 2
Answer: y = − e−x + C.
2

M. Maliyoni, PhD (Appl. Math − Math Bio)

5 Mathematical Modelling: Introduction and Applications

1.3.2 Separation of Variables

A differential equation of the form

dy g(x)
= (6)
dx h(y)
is said to be separable or to have separable variables.

We can re-write (6) as

dy
h(y) = g(x). (7)
dx
dy
When h(y) = 1, (7) reduces to = g(x).
dx

Now, if y = f (x) denotes a solution of (7), h(f (x))f 0 (x) = g(x) and therefore
Z Z
0
h(f (x))f (x) dx = g(x) dx + C. (8)

But dy = f 0 (x) dx, so (8) becomes

Z Z
h(y) dy = g(x) dx + C. (9)

Equation (9) is used to solve separable differential equations. This is achieved by integrating both sides of
the equation.

Note: Only one constant of integration should be used in the integration of separable equations.

dy
Example 3 Solve the equation = y.
dx
Solution
Re-writing the equations by separating the variables and integrating both sides yield

Z Z
dy
= dx
y
Z Z
1
=⇒ dy = dx
y
ln |y| = x + C
=⇒ eln |y| = ex+C
Thus, y = ex .eC
∴ y = Cex .

Exercise
Solve (1 + x) dy − y dx = 0.
Answer: y = C(1 + x).

M. Maliyoni, PhD (Appl. Math − Math Bio)

6 Mathematical Modelling: Introduction and Applications

1.3.3 Exact Differential Equations

An equation
M (x, y) dx + N (x, y) dy = 0

is said to be an exact differential if M (x, y) dx + N (x, y) dy is an exact differential.

∂f ∂f
Note: If z = f (x, y), then dz = dx + dy.
∂x ∂y

Thus,
∂f ∂f
M (x, y) dx + N (x, y) dy = dx + dy.
∂x ∂y
where
∂f
• M (x, y) dx = dx.
∂x
∂f
• N (x, y) dy = dy.
∂y

Example 4 The equation x3 y 4 dx + x4 y 3 dy = 0 is an exact equation since

 
1 4 4
d xy = x3 y 4 dx + x4 y 3 dy
4
M (x, y) dx = x3 y 4 .
N (x, y) dy = x4 y 3 .

Reading Exercise
Read about total differential, test for an exact differential and method of solution for solving exact differ-
ential equation.

Example 5 Solve (5x + 4y) dx + (4x − 8y 3 ) dy = 0.

Solution
∂M
M (x, y) = 5x + 4y =⇒ = 4.
∂y
∂N
N (x, y) = 4x − 8y 3 =⇒ = 4.
∂x
∂M ∂N
=⇒ = = 4.
∂y ∂x

∴ The equation is exact and has to be solved for f (x, y).

Now
∂f ∂f
(5x + 4y) dx + (4x − 8y 3 ) dy = dx + dy
∂x ∂y

M. Maliyoni, PhD (Appl. Math − Math Bio)

7 Mathematical Modelling: Introduction and Applications

for some function, f .

∂f
Thus, = M (x, y) = 5x + 4y.
∂x
Z
=⇒ f (x, y) = (5x + 4y) dx + g(y) (10)
5x2
= + 4xy + g(y). (11)
2
∂f
But = 4x + g 0 (y) = 4x − 8y 3 = N (x, y)
∂y

=⇒ g 0 (y) = 4x − 8y 3 − 4x = −8y 3 . (12)

8y 4
Z
∴ g(y) = − 8y 3 dy = − = −2y 4 . (13)
4

Substituting (13) into (11) gives

5x2
f (x, y) = + 4xy − 2y 4 = C.
2
The solution is therefore
5x2
+ 4xy − 2y 4 = C.
2
Exercise
Solve the initial value problem (IVP)

(ex + y) dx + (2 + x + yey ) dy = 0; y(0) = 1.

Answer: ex + xy + 2y + ey (y − 1) = 3.

1.3.4 Integrating Factor

The standard form of a first-order linear equation is

dy
+ P (x)y = f (x). (14)
dx
We want to find the solution of equation (14)R
on an interval I for which P (x) and f (x) are continuous.
P (x) dx
We will use the integrating factor, I.F. = e to solve equation (14).

Exercise
Read about the derivation of the integrating factor and the method of solution of (14).

dy
Example 6 Solve the equation x + 2y = 3.
dx

M. Maliyoni, PhD (Appl. Math − Math Bio)

8 Mathematical Modelling: Introduction and Applications

Solution
We re-write the equation as
dy 2 3
+ y= . (15)
dx x x
2
Thus, P (x) = . Now the integrating factor is given by
x
1 2
R
I.F. = e x
dx
= e2 ln |x| = eln x = x2 . (16)

Multiplying (15) by (16) gives

dy 2 2 3 2
x2 .+ y.x = .x
dx x x
dy
=⇒ x2 . + 2xy = 3x
dx
Thus,
d 2
[x y] = 3x (17)
dx
1
R
dx
where I.F. = e x = x2 .

Integrating both sides of (17) gives

3x2
Z
2
xy= 3x dx = + C.
2
3
=⇒ y = + Cx−2 .
2
Exercise
dy
Solve = 1 − 2y; y(0) = 0.
dx

Answer: y = 1
2
− 21 e−2x = 12 (1 − e−2x ).

1.3.5 Homogeneous Differential Equations

A differential equation M (x, y) dx + N (x, y) dy = 0 is said to be a homogeneous DE if M (x, y) and

N (x, y) are homogeneous functions of the same degree, that is, if

M (tx, ty) = tn M (x, y) and N (tx, ty) = tn N (x, y).

Reading Exercise
Read about homogeneous functions and method of solution of homogeneous DEs.

Example 7 Solve the differential equation (x + y) dx + x dy = 0.

Solution
M (x, y) = x + y =⇒ M (tx, ty) = tx + ty = t(x + y) = tM (x, y).

M. Maliyoni, PhD (Appl. Math − Math Bio)

9 Mathematical Modelling: Introduction and Applications

N (x, y) = x =⇒ N (tx, ty) = tx = tN (x, y).

∴ Both functions are homogeneous of the same degree.

Now, we make the substitution y = ux =⇒ dy = u dx + x du. Thus, equation (x + y) dx + x dy = 0

becomes

(x + ux) dx + x(u dx + x du) = 0

x dx + ux dx + ux dx + x2 du = 0
x dx + 2ux dx + x2 du = 0
x(1 + 2u) dx + x2 du = 0
(1 + 2u) dx + x du = 0

which is separable. That is

dx du
+ = 0. (18)
x 1 + 2u
Integrating both sides of equation (18) leads to

1
ln |x| + ln |1 + 2u| = k, where k is a constant. (19)
2
But y = ux ⇒ u = xy . Hence, equation (19) becomes

1 2y
ln |x| + ln 1 + = k.
2 x

Exercise
Solve xy 2 dy + (x3 − y 3 ) dx = 0.

y3
Answer: + ln |x| = k.
3x3

Note: Epidemic models are nonlinear systems in nature as such they are analyzed as any other nonlin-
ear system. Thus, we will briefly look at nonlinear systems as indicated in Section 2.

2 Introduction to Nonlinear Systems

2.1 Analysis of Nonlinear Systems
An nth-order system of continuous differential equations has the following form:

M. Maliyoni, PhD (Appl. Math − Math Bio)

10 Mathematical Modelling: Introduction and Applications

ẋ1 (t) = f1 (x1 (t), x2 (t), · · · , xn (t)),

ẋ2 (t) = f2 (x1 (t), x2 (t), · · · , xn (t)),
.. (20)
. = ··· ··· ··· ···
ẋn (t) = fn (x1 (t), x2 (t), · · · , xn (t))
which can be expressed in matrix form as

Ẋ(t) = f (X(t), t),

   
x1 f1
 x2   f2 
   
where X =   .. 
 and f =  . .
.
. .
xn fn
In general, one does not seek detailed solutions of system (20) either in numerical or analytical form,
but rather to characterize some aspects of system behavior. For example, we may ask whether there are
equilibrium points and whether or not they are stable. In nonlinear systems, one might look, in addition to
equilibrium points, for threshold effects. The approach therefore, includes characterizing in broad terms
the critical aspects of the system behavior.

2.2 Main Tools for Analysis of Nonlinear Systems

2.2.1 Equilibrium Points (or Steady States)

Definition: A vector X̄ is an equilibrium point for a dynamical system if once the state vector is equal to
X̄ it remains equal to X̄ for all future time.

Thus, if
Ẋ(t) = f (X(t), t)

then an equilibrium point is a state X̄ satisfying

f (X̄, t) = 0

for all time t.

An analysis of a nonlinear dynamical system may devote considerable attention to the characterization of
the equilibrium points. In linear systems, equilibrium points are basically solutions to linear equations.
The nonlinear case is different in two essential aspects:

i) Since the equilibrium points are solutions, in this case, to nonlinear equations, finding such solutions
is somewhat more of an accomplishment than in the linear case.

ii) The equilibrium point distribution is potentially more complex in the nonlinear case than in the linear
case.

M. Maliyoni, PhD (Appl. Math − Math Bio)

11 Mathematical Modelling: Introduction and Applications

Note: A system may have none, one or any number of equilibrium points in virtually a spatial pattern in
state space.

Thus, characterization of equilibrium points is not only technically more difficult, it is a much broader
question. Ultimately, interest centers not just on the existence of the equilibrium points but also on their
stability properties.

2.2.2 Stability

Stability properties characterize how a system behaves if its state is initiated close to, but not precisely at
a given equilibrium point.

(a) If a system is initiated with the state exactly equal to an equilibrium point, then by definition it will
never move.

(b) When initiated close by, however, the state may remain close by, or it may move away.

Roughly speaking, an equilibrium point is stable whenever the system state is initiated near that point, the
state remains near it, perhaps even tending towards the equilibrium point as time increases.

Suppose X̄ is an equilibrium point of

Ẋ(t) = f (X(t)).

For a precise definition of stability, it is convenient to introduce the notation S(X̄, R) to denote a spheri-
cal region in the state space with center at X̄ and radius R.

Definition: An equilibrium point X̄ is stable if there exists R0 > 0 for which the following is true. For
every R < R0 there exists r, 0 < r < R, such that if X(0) is inside S(X̄, R), then X(t) is inside
S(X̄, R) for all t > 0.

2.2.3 Linearization and Stability

According to the basic definitions, stability properties depend only on the nature of the system near the
equilibrium point. Therefore, to conduct stability analysis, it is often theoretically legitimate and mathe-
matically convenient to replace the full nonlinear description by a simpler description that approximates
the true system near the equilibrium point. Often a linear approximation is sufficient to reveal the stability
properties.
The linearization of a nonlinear system is based on linearization of the nonlinear function f in its descrip-
tion. For the first-order system, defined by a single function f (x) of a single variable, the procedure is to
approximate f near x̄ by

df (x̄)
f (x̄ + y) = f (x̄) + y. (21)
dx

M. Maliyoni, PhD (Appl. Math − Math Bio)

12 Mathematical Modelling: Introduction and Applications

An nth-order system is defined by n functions, each of which depends on the n variables. In this case
each function is approximated by the relation

∂fi (x̄1 , x̄2 , · · · , x̄n ) ∂fi (x̄1 , x̄2 , · · · , x̄n )

f (x̄1 + y1 , x̄2 + y2 , · · · , x̄n + yn ) ≈ fi (x̄1 , x̄2 , · · · , x̄n ) + y1 + y2
∂x1 ∂x2
∂fi (x̄1 , x̄2 , · · · , x̄n ) ∂fi (x̄1 , x̄2 , · · · , x̄n )
+ y3 + · · · · · · · · · · · · · · · + yn
∂x3 ∂xn
(22)
where i = 1, 2, 3, · · · , n. In matrix form, (22) can be written as

f (x̄ + y) ≈ f (x̄) + F y, (23)

where  
∂f1 ∂f1 ∂f1
 ∂x1 ···
∂x2 ∂xn 
 ∂f2 ∂f2 ∂f2 
 
 ··· 
 ∂x1 ∂x2 ∂xn 
··· ··· ··· ··· .
 
F =
  (24)
··· ··· ··· ··· 
 
··· ··· ··· ··· 
 
 ∂fn ∂fn ∂fn 
···
∂x1 ∂x2 ∂xn

The matrix F is called the Jacobian matrix of f .

Now consider the matrix equation

ẋ = f (x(t)).

Setting x(t) = x̄ + y(t) we obtain

ẏ = f (x̄ + y(t)) = f (x̄) + F y(t). (25)

Since x̄ is an equilibrium point of f , f (x̄) = 0. Therefore

ẏ(t) = F y(t). (26)

Thus, the stability properties of the original system, (22), can be inferred from the linearized system, (26),
using the following results.

(a) If all eigenvalues of F have negative real part, then x̄ is asymptotically stable for the nonlinear system.

(b) If at least one eigenvalue of F has a positive real part, then x̄ is unstable for the nonlinear system.

M. Maliyoni, PhD (Appl. Math − Math Bio)

13 Mathematical Modelling: Introduction and Applications

3 Epidemic Models
3.1 A Basic SIR Epidemic Model
Epidemic models are constructed to enable us understand the spread of an infectious disease in a host
population. The major factors are:

1. Variables
These are subgroups of the host population. We have the following subgroups for a basic epidemic
model:

(a) S: The number of susceptibles. These are individuals who can catch the disease if they come in
contact with it).
(b) I: The number of Infectives. Individuals who are infected with with the disease. Can be infec-
tious or not.
(c) R: The number of Removals. Individuals who have recovered from the infection or have been
quarantined or have succumbed to the disease.

2. Parameters
These are divided into two groups.

(a) Epidemiological Parameters

Includes but not limited to the following:
i. Infection rate.
ii. Progression rate.
iii. Recovery rate.
iv. Vaccination rate.
v. Efficacy rate (the strength of a particular drug - how effective the drug is) etc.
(b) Demographical Parameters
i. Related to the Natural dynamics of the community/population.
ii. Natural birth rate.
iii. Natural death rate.

3. Assumptions

• Should concern the disease under study, variables and parameters.

• Must be realistic in the practical sense. For example, the assumption that Ebola infection does
not kill is unrealistic.
• All assumptions should be reflected in the model.

M. Maliyoni, PhD (Appl. Math − Math Bio)

14 Mathematical Modelling: Introduction and Applications

3.2 Definition of Some Terms

(a) Force of Infection, λ
βI
This is the rate at which susceptibles catch the disease, i.e. βI or , where N is the total population.
N
Thus, λ = βIS is the mass-action term.

(b) Herd Immunity

A population is said to have herd immunity if a large enough fraction has been immunized to ensure
that the disease cannot be endemic. The goal is to reduce R0 to below 1 i.e. R0 < 1. This is achieved
by immunization.
For example, if a fraction ρ of say, µN of newborns (µ is the birthrate) is successfully immunized, N
is replaced by N (1 − ρ) and therefore reduce R0 by R0 (1 − ρ).
Thus, R0 (1 − ρ) < 1 leads to ρ > 1 − R10 . Hence, ρ should always be greater than 1 − R10 if we are to
have the disease eliminated.

3.3 General Dynamics of an Epidemic

Individuals pass from one class to another with the passage of time. Movement is due to the status of the
individuals at that time. For instance, if a susceptible individual gets infected, he/she moves from class S
to class I.
Note: A mathematical model tries to capture this flow by using compartments.

3.3.1 The Compartmentalized Model

These are sets of ordinary or partial differential equations that describe the rate of flow between compart-
ments referring to numbers or proportions in each category. Refer to Section 3.3.2 for commonly used
compartmental model frameworks.

3.3.2 Common Epidemiological Model Frameworks/Structures

The following are some of the common model structures (applicable to different infections);

1. SI Framework
In this structure, individuals move from susceptible class to infectious class and remain there. It
could be that one cannot recover from the infection. See Figure 1.

Figure 1: An SI compartmental model/flow diagram

M. Maliyoni, PhD (Appl. Math − Math Bio)

15 Mathematical Modelling: Introduction and Applications

Transition: A change from being susceptible to infectious. This is also where the type of transmis-
I
sion is considered i.e. βS, βSI, βS .
N
2. SIS Framework
Individuals move from susceptible class to infectious then one can recover without immunity or with
temporary immunity and becomes susceptible again. See Figure 2.

Figure 2: An SIS compartmental model/flow diagram

3. SIR Framework
This is where recovery is with full or permanent immunity i.e. once you recover you do not become
susceptible again. See Figure 3.

Figure 3: An SIR compartmental model/flow diagram

4. SIRS Framework
In this framework, individuals move from susceptible to infectious class, then they recover with
temporary immunity and go back to susceptible class. See Figure 4.

Figure 4: An SIRS compartmental model/flow diagram

M. Maliyoni, PhD (Appl. Math − Math Bio)

16 Mathematical Modelling: Introduction and Applications

4 Modifications to the Models

- The infectives cannot remain infected forever.

- A more realistic model must take into account the fact that for most diseases infectives either recover
from the infection or they die.

4.1 The SIS Model

- It is simple but serves to illustrate the main features of all disease models.

- Mostly used in the study of sexually transmitted diseases.

- There is no recovered class due to vast number of strains.

- We modify the model by adding vital dynamics, that is, birth rate and death rate.

- Any flow in is represented by a +ve and flow out by a -ve.

Figure 5 shows the compartmental model of the diagram.

Figure 5: SIS modified model compartmental diagram

4.1.1 Assumptions of the Model

We make the following assumptions:

(a) The population is variable.

(b) There is homogeneous mixing in the population.

(c) There is no disease-related deaths.

(d) The recruitments do not have the disease at the time they are joining the population.

(e) The is no recovery with permanent immunity.

(f) No vertical transmission.

M. Maliyoni, PhD (Appl. Math − Math Bio)

17 Mathematical Modelling: Introduction and Applications

4.1.2 Equations of the Model

From Figure 5, we have the following equations of the model

I
S 0 (t) = π − cbS − µS + γI
N (27)
0 I
I (t) = cbS − (µ + γ)I
N
Model (27) must be analysed qualitatively i.e. finding the equilibrium points and their stability, computing
R0 etc.

4.2 SIR Model (Kermack-McKendrick Model)

Variables: S(t); I(t) and R(t).
Parameters: β - infection rate; γ - removal rate.

4.2.1 Assumptions

(a) Constant population (i.e. no births, deaths, recruitments) with initial I0 infectives and S0 susceptibles.

(b) Mass action incidence rate.

(c) People in the community mix freely.

(d) Recovery is with permanent immunity.

(e) The rate at which individuals contract the disease is proportional to the number of contacts between S
and I.

4.2.2 Model Equations

S 0 (t) = −βSI
I 0 (t) = βSI − γI (28)
R0 (t) = γI
with initial conditions S(0) = S0 > 0; I(0) = I0 > 0; R(0) = 0 and N = S + I + R.

4.2.3 Simple Analysis of the Model

dI
For an epidemic to initially start, equation 2 of model system (28) must be positive, that is, > 0. Thus,
dt
dI
= βSI − γI = (βS − γ)I
dt  
γ
= β S− I
 β 
γ
= β S0 − I0 putting initial conditions
β
γ
= β (S0 − ρ) I0 , where ρ = which is the relative removal rate
β

M. Maliyoni, PhD (Appl. Math − Math Bio)

18 Mathematical Modelling: Introduction and Applications

dI
Now, > 0 if S0 > ρ (the condition for the epidemic to start). Therefore, if the initial number of
dt
S0 < ρ, then there is no epidemic. ρ gives a threshold density of susceptibles.

Exercise
Verify that model system (28) is a constant population model.

4.3 SIR Modified Model

Variables: S(t); I(t); R(t).

Table 1 shows the parameters of the model.

Table 1: Parameter symbols and their definition of the SIR modified model
Parameter Symbol Parameter Definition
b Birth rate (aka flow rate)
δ Natural mortality rate
α Disease-induced death
p Proportion of offspring of infected parents who are also infectious
γ Removal rate
β Rate of infection

The flow diagram of the model is shown in Figure 6.

Figure 6: SIR modified model flow diagram

Note: bpI represents the proportion of those who are born infected; αR represents death due to disease of
those who are removed.

Exercise
State any three assumptions that are made in the model represented by Figure 6.

4.3.1 Equations of the Model

With reference to Figure 6, the equations of the model are:

M. Maliyoni, PhD (Appl. Math − Math Bio)

19 Mathematical Modelling: Introduction and Applications

S 0 (t) = bN − δS − βSI
= bN − (δ + βI)S
0
I (t) = βSI − αI − δI − γI + bpI
(29)
= βSI − (α + δ + γ)I + pbI
0
R (t) = γI − αR − δR
= γI − (α + δ)R
Note: Mean length of stay (MLS) in a class is the reciprocal of what is going out to another class. In
model (29),

1
M LS of I =
α+δ+γ
and
1
M LS of R = .
α+δ

4.4 SIRS Modified Model

This allows to for temporary immunity. The individuals will lose immunity as time goes on. This is shown
in Figure 7.

Figure 7: SIRS modified model flow diagram

The equations from the flow diagram, Figure 7 are

dS
= aN − (λ + µ)S + γR
dt
dI
= λS − (µ + α + ν)I (30)
dt
dR
= νI − (µ + γ)R
dt

M. Maliyoni, PhD (Appl. Math − Math Bio)

20 Mathematical Modelling: Introduction and Applications

I
where λ = βI or λ = β . Thus, λ depends on the type of incidence being used.
N

Exercise
State any three assumptions that are made in the model represented by Figure 7.

4.5 The SEIR Model

The SEIR model contains one more compartment, as it is apparent from its name. The new compartment is
called exposed compartment, E. These are the people who are infected but the symptoms of the disease are
not yet visible i.e. they are not infectious as such they cannot transmit the disease to others. These people
are said to be in latent period . For some diseases (e.g. tuberculosis, malaria etc), it takes certain time for
an infective agent to multiply inside the host up to the critical level so that the disease actually manifest
itself in the body of the host. This is called an incubation period. When the infective agents become active
in the host’s body, the individual progresses to the infectious class. In this model we assume homogeneous
mixing (mass action principle), horizontal transmission etc. Figure 8 shows a simple SEIR TB model flow
diagram with permanent immunity;

Figure 8: SEIR model flow diagram

In Figure 8, the definitions of the parameters are given below:

• π: Recruitment rate.

• λ: Infection rate.

• ρ: Progression rate.

• θ: Recovery rate.

• µ: Natural mortality rate.

• α: Disease-induced death.

From Figure 8, we get the following system of nonlinear ordinary differential equations;

S 0 (t) = π − (λ + µ)S
E 0 (t) = λS − (µ + ρ)E
(31)
I 0 (t) = ρE − (µ + α + θ)I
R0 (t) = θI − µR

M. Maliyoni, PhD (Appl. Math − Math Bio)

21 Mathematical Modelling: Introduction and Applications

I
where λ = βI or λ = β .
N

Exercise
Modify model system (31) by including loss of immunity and vertical transmission.

5 Mathematical Analysis of Epidemic Models

5.1 Qualitative Analysis
This includes, but not limited to,

(a) Positivity of solutions.

(b) Invariant region (i.e. which region does the solution lie).

(c) Existence and stability of equilibria (local and global stability).

(d) Reproduction number (using next-generation matrix).

(e) Existence and non-existence of solutions or periodic solutions.

(f) Uniqueness of solutions.

(g) Sensitivity analysis using either partial rank correlation coefficient (PRCC) method or the normalized
forward sensitivity index.

(h) Parameter estimation

5.2 Numerical Solutions/Simulations

(a) Implementing the model in computing languages such as R, MATLAB, Python, Mathematica, etc.

(b) Model fitting to data (if available)

(c) Simulation of the model using parameter values in literature

M. Maliyoni, PhD (Appl. Math − Math Bio)

22 Mathematical Modelling: Introduction and Applications

References

1. Anderson RM, May RM (1991). Infectious diseases of humans: Dynamics and Control, Oxford
University Press.

2. Boyce WE, DiPrima RC (2012). Elementary Differential Equations and Boundary Value Problems,
10th edition, New York: John Wiley

3. Hethcote HW (2000) The mathematics of infectious diseases. SIAM Rev 42(4):599653

4. Kraemer A, et al. (2010) Morden infectious diseases epidemiology, Springer-Verlag.

5. Zill, D.G., (2012). A First Course in Differential Equations with Modeling Applications, 10th Edi-
tion, Boston: PWS-Kent Publishing.

M. Maliyoni, PhD (Appl. Math − Math Bio)