Mathematics Higher Level Exploration

SIR model in epidemiology and its examples.

School name: ***************************

Candidate name: ****************
Candidate number: ******-0006

Introduction
Reading an article Should All children be immunised against influenza? in British
Medical Journal written by Dr Valtyr Thors1 challenged me to take a deeper insight into
statistical figures and methods used to calculate them. I started to think about mathematical
significance in epidemic modelling and how the spread of the disease is predicted using
mathematics. Searching for more information, I decided to focus my investigation on one type
of model, SIR model. During my research I got many different sources with vulnerable
information, however I could not find one in which all the mathematical steps to obtain SIR
model would be sufficiently explained. This broaden my project, because in order to develop
a framework of a disease, which was my main goal, first I had to understand how all those
formulas presented in books were obtained.
The first mathematical model describing the spread of disease was invented by Daniel
Bernoulli in 17602. It was based on differentiatial equations in which variables were
characterising properties of epidemic. Since that day, a tremendous development has been
made and many models have been invented. SIR model was formulated by Kermack and
McKendrick in 1927.3 It is very important mathematical tool, used to predict outbreak of a
disease. It helps scientists not only to understand the spread of a epidemic, but also to
effectively evaluate methods of disease control.
Basically, SIR model describes the spread of disease in a closed system, where the
number of population is constant and there is no migration.4 What is more, we assume that
there is no latency and homogenous mixing occur, which means that every healthy individual
has the equal probability of getting infected. In the SIR model, there is an universal set, let
call it N, which is divided into three classes5:
Susceptible (S) those who is presently healthy, but may get ill.
Infected (I) ill individuals.
Recovered (R) people who were ill and are immune, so cannot get the
disease.
1

British Medical Journal, volume 98, Issue 11, p. 846-849.

http://statprob.com/encyclopedia/DanielBERNOULLI.html
3
http://mathworld.wolfram.com/Kermack-McKendrickModel.html
4
http://mathworld.wolfram.com/Kermack-McKendrickModel.html
5
http://mathworld.wolfram.com/Kermack-McKendrickModel.html
2

The model is described as a system of three differential equations:

where

(1)

(2)

(3)

in (1) & (2) is transition rate of a disease with a proportionally coefficient

and

in (2) & (3) represents recovery rate.

Adding all equations, we can notice a very interesting property :
, which verifies that the total population is constant
.

(4)

In this portfolio, I will consider cases of an epidemic outbreaks where:

(5)

(6)

(7)

Graph 1.
Epidemic Trajectory6

Graph 1, represents the relation between the number of susceptible and the number of
infected individuals.

We can conclude that if

epidemic does not appear. Hence for epidemic to begin

then

and it means that

This condition will result

Picture taken from Lectures on mathematical modelling of Biological Systems G. Bastin,

17.02.2012.

in the increase in infected class, till the fraction of susceptible will decrease to

and the

number of infected individuals will decrease.

In order to fully understand SIR model one more quantity must be introduced:
(8)

is the basic reproductive ratio and is defined as the average number of secondary cases
arising from an average primary case in an entirely susceptible population.7

First Case
After looking through available sources about SIR modelling, nowhere I could find a
book in which all theory together with mathematical calculations would be explained in
details. As I was really interested in the topic, to ensure myself that I understand the
methodology, I decided to analyze my personal example and try to present all calculation
needed to move from one step to another, to understand how the formulas work.
Let assume that there is a group of 900 children aged 13-15 on the summer camp. One
girl arrives with a running nose and in the evening she has all flu symptoms. Teachers send
her to bed, but she manages to give infection to other kids. Assuming that camp lasts for 21
days,

and the recovery rate

, predict the spread of flu, if 641 children

get sick.
First of all we have to check, if epidemic will start at all. The condition needed is
Substituting the values into the formula stated in the introduction, we obtain:

This satisfies the condition and means that there will be flu epidemic on the summer camp.
Next to make a graph of the spread of illness, we need to calculate fraction of infected
children for each day of summer camp. To do that, we divide equation (2) by (1):

Medlock J. (1999). The effect of Scholastic Migration on an SIR model for the transition of HIV.
Georgia Institute of Technology, USA.

(9)

Next we integrate this equation to obtain the formula for I.

To calculate we will calculate the value of

for

. We know from (6) that

.

, because

(from (5)).
.
).

(10)

Now having this formula, we are able to calculate the number of infected children for each
day. We will do it later, together with daily calculations of susceptible and recovered children.
What is more, the above formula gives us opportunity to calculate

. We know that

and
.
Now we can substitute it into the formula for

In my particular case:
4

We can reduce how many susceptible children will remain after 21st day and will
come back home healthy. We start by dividing equation (1) by (3) and next as in case of
infected, we integrate our solution.

Having in mind our initial assumptions (6) and (7), we will calculate the value of
.

Hence we finally have:

Unfortunately we do not know how many children recovered before the summer camp
was finished, so right now we are not able to calculate

, but do not worry, we have not

done all those calculations for nothing, this will help us to obtain formula for :

To calculate

, so the fraction of the population that recovered, first we have to

notice one thing. As the epidemic develops, the number of infected individuals increase,
however after reaching threshold moment, people recover and it decreases, so

We know that

and from (5) that

, therefore we can write:

(12)
Here in our particular case:

I was very surprised by this result, because 1.75 is our value of

and how it was possible to

obtain it? Then I found the explanation for this result, which I will describe at the beginning
of the next chapter.
As I am very interested in looking for the answer using various method I decided to
finally plot my epidemic model using Euler equations and Excel.

Graph 2.
The representation
of epidemic spread
on summer camp
for 21 days.

Second case
Firstly, let start from the explanation of a case. This time it is real situation of
influenza epidemic in boarding school in 1978 described by British Medical Journal8. During
two weeks, the majority (512) boys out of 763 were infected. It started from one boy coming
back from winter break.

Graph 3.
Graph taken from British Medical Journal
presenting numbers of infected boys during
influenza epidemic in 1978 in boarding school.

The biggest difficulty for me was to get an idea how to calculate and

I even did not

have detailed data, just only a graph. I decided that the easiest way to calculate and

is to

write a computer command using Scilab and by guessing put random numbers. This way I
would obtain the best fit model. On the other hand, I understood limitations of this method,
the values of and

are very small, I might easily make a mistake.

I was very curious and determined to obtain the solution in mathematical way. I was
thinking about this for a very long period of time and finally I made my own observation,
which really surprised me. As we know, spread of disease is not always (rather rarely) directly
proportional. It depends on many factors, such as genetic predisposition, individuals
immunity, age, nourishment, etc. Therefore the values of and
easily produce a graph. Assuming that at the end of epidemic

are simplified in order to

, because at the end

every infected person eventually recovers and get immunity, we can modify our previous
equation (11).

I checked this equation on many already solved cases and in every it seemed to work, because
I got correct answers. I decided to apply this into my case.
8

British Medical Journal, volume 98, Issue 11, p. 846-849.

By reading values from the graph, I approximately estimated the number of infected
boys for each day. This moved me one step closer to the solution. Knowing the number of
infected for each day, I decided to solve the equation (1) using Euler method and obtained
:

, where

is always equal to

, because it is always 1 day interval

between the measurement of infected boys. Using Excel I obtained the values of
day. Then from the equation

, I calculated

for each

values for each day, the table with

data for each day is present in the Appendix.

I had data for everyday, but I still needed one more thing, the value of . To do this I
used the least square method in Excel and finally obtained the value of

. Again I

used Scilab and this time I produce the most accurate graph (The command I wrote for Scilab
is in the Appendix).

Graph 4.
The model of influenza
epidemic in boarding school
in England in 1978 using
Scilab.

Happy that I managed to solve the problem, I started to consider different situations
which may occur in natural environment. As I was reading different articles about epidemic
cases in the word, I stared to think about the properties of SIR model. We used there
definitely too many assumptions, which will give us unrealistic outcomes. This method could
work, but just only for short period of time cases and in a closed universal group. However,

people are born and die from different causes than studied illness. This lead me to third
model: SIR model with demography.

Third Case with Demography.

Let consider the situation, where we have change in the number of population due to
migration, births and deaths. This will change a system of three differential equations
describing potential situation9:

where

(13)

(14)

(15)

in (a) & (b) is transition rate of a disease with a proportionally coefficient

in (b) & (c) represents recovery rate.

and

is the rate of change in the number of society:

migrations, births and deaths.

Adding all the differential equations, we can notice a very interesting property :

and we have verified that also in this case, the total population is constant

This model is very useful in establishing the prevalence of disease at equilibrium,

determining stability properties for endemic equilibrium and oscillatory level, as well as
establishing the threshold level of a disease. I would like to consider all the cases, however
due to time limit, let focus on the first two points. The system of differential equations has
two equlibria which are the possible constant answers:
1. Disease-free equilibrium disease finally ends up. Every individual is
susceptible. It occurs when:

Lectures on mathematical modelling of Biological Systems G. Bastin,

2. Disease endemic equilibrium disease never dies, its always present in

population. To establish when a disease endemic equilibrium is present we
solve the second differential equation (11), by equalling it to zero.

We have two solutions

establish the value of

Calculating

, which is disease-free equilibrium, or

. Now, let

using the equation (b).

is a piece of cake:

Therefore the endemic equilibrium occurs when:

Now, we want to know how these two equilibrium are likely to occur. To evaluate that
stability properties of both cases should be calculated. We can predict, having knowledge
from first and the second model, that it is highly possible for the disease-free equilibrium to
be present when

because just only if

, epidemic can begin and this is probably

the condition for endemic equilibrium. So, let us check that.

10

To do that we need to used Jacobian matrix. It is the matrix of all first-order partial
derivatives of a vector-valued function.10 For a SIR model with ordinary differential equations

the stability occurs if the real of all eigenvalues (here represented by

) are negative11.

Let us start with disease-free equilibrium, we calculate the determinant by solving this
equation when

and

:
(14)

We get two solutions which must be negative:

For
sides by

to be negative, such condition must be present

we obtain

Dividing both

This verifies the first part of our assumption. Now we can

move forwards to check the condition for endemic equilibrium. We substitute the values for
and

and calculate the equation (14).

1Ro=0

We again have two solution, however the first apply for the disease-free equilibrium,
so it is ignored. To calculate the exact value for

we have to notice that it is kind of

quadratic equation and we will solve it as that.

10

http://en.wikipedia.org/wiki/Jacobian_matrix_and_determinant
Medlock J. (1999). The effect of Scholastic Migration on an SIR model for the transition of HIV.
Georgia Institute of Technology, USA.
11

11

We can notice that

is very small, so we can skip it and move further:

As I mentioned above the stability occurs if the real parts of all eigenvalues are
negative, therefore endemic equilibrium is stable just only if

. This way we verified

the second part of our assumption.

As a third model I decided to consider cholera. After all this time, this pandemic still
kills hundreds of thousands every year. I was researching for the real data of cholera epidemic
and unfortunately I could not find one. I used some parameters found in work of Leah
Johnson12. I created SIR models without and with demography for the following parameters:
total susceptible human population
recovery

per year,

individuals,

per year, rate of

and time interval 20 years.

Graph 5.
SIR model of cholera
epidemic without
demography.

Graph 6.
SIR model of cholera
epidemic with
demography.
12

Johnson L., 2006. Mathematical modelling of cholera: from bacterial life histories to human
epidemics, University of California, USA

12

Conclusion
This project gave me a valuable opportunity to develop insight into the importance and

beauty of mathematics in medicine. Ive never thought of mathematical significance in

epidemic study and in this exploration I could not only research about my area of interest, but
also I could see limitations of modelling. I understood that developing the most accurate
model of epidemic is very difficult. Even if, due to time and space limit, I was able to explore
just only two cases, I found this work very fruitful. I learned that you cannot trust what you
read and always have to look at limitations of the statistical figures. In the real world there are
numerous important factors which have to be considered and people not always apply all of
them.
What is more, I checked that I am able to understand the methodology hidden by
statistical figures. I have a lot of fun creating my personal examples as well as searching for
the real case situations. I had a chance to write my own computer command, which was for a
total amateur a big challenge.
I learned a lot working on this portfolio, I am very enthusiastic about further
investigation. This investigation not only broaden my interests, but also taught me to look at
things from different perspectives. In future I would like to apply other factors to the SIR
model. What is more, I would like to consider a model of human cholera with a bacteria
reservoir.

13

Appendix

Table 1.
The number of susceptible, infected and
recovered individuals for epidemic in boarding
school in 1978 in England.

Scilab command used to calculate the differential equations for influenza epidemic in
boarding school in 1978:

function ydot=f(t, y);

b=0.002174;
g=0.4545
S=y(1);
I=y(2);
R=y(3);
dS=-b*S*I;
dI=b*S*I-g*I;
dR=g*I;
ydot=[dS,dI,dR];
endfunction
t=0:1:14;
y0=[762;1;0];
t0=0
y=ode(y0,t0,t,f);
plot(t,y)
legend('S','I','R')

14

Bibliography

1. Bastin G. (2012). Lectures on mathematical modelling of Biological Systems.

2. (1978). Influenza in a boarding school. British Medical Journal, volume 98, Issue 11,
p. 846-849
3. Johnson L. (2006). Mathematical modelling of cholera: from bacterial life histories to
human epidemics. University of California, USA.
4. Medlock J. (1999). The effect of Scholastic Migration on an SIR model for the
transition of HIV. Georgia Institute of Technology, USA.
5. Siddiqui S. (2008). Backward bifurcations in SIR endemic models.

Massey

University, New Ziland.

6. Tian P., Liao S.(2000). Dynamical analysis and control strategies in modelling
cholera. Old Dominion University.

7. Gani J., StatProb: Daniel Bernoulli. Available at:

http://statprob.com/encyclopedia/DanielBERNOULLI.html. [Accessed 10.01.2014].

8. Wikipedia: Jacobian matrix and determinant. Available at:

http://en.wikipedia.org/wiki/Jacobian_matrix_and_determinant. [Accessed
16.01.2014].

9. Wolfram Math World: Kermack-McKendrick Model. Available at:

http://mathworld.wolfram.com/Kermack-McKendrickModel.html. [Accessed
21.01.2014].

15