Developing a model from the ground up :

Case study of the spread of an infection

By: Asma Alweldi ,Waad Alnajjari, Nouf Alyoubi

By Supervision: Dr Noura Alharthi
6.1Deriving a model for the spread of an infection
• Assume that the population of interest is “well mixed,” which means that any individual
is equally likely to come into contact with any other individual. Let us subdivide the
population into two classes: those that are healthy, but susceptible to the disease, and
those that are infected and able to spread the disease through contact. Figure 6.1
illustrates one view

Figure 6.1. A compartmental diagram for the susceptible-infected classes in a simple model for the spread of an
infection.
6.2Dimensional analysis applied to
the model
• The ideas of Chapter 4 will now be used in order to reduce the number of
parameters and make the model dimensionless.We will thereby identify an
important parameter grouping (renamed R0) that affects the qualitative behavior
of the solutions.
Example 6.4 (units). Suppose that we agree to measure time in units of days. Use
the differential equations (6.2) or (6.5) to determine the units carried by each of the
variables and parameters in the model.
Solution. We easily see that the left-hand sides of Eqs. 𝑑𝑠
𝑑𝑡
= 𝜇𝐼 − 𝛽𝑆𝐼 𝑎𝑛𝑑
𝑑𝐼
𝑑𝑡
= 𝛽𝑆𝐼 − 𝜇𝐼
have dimensions of [number of people]/[time]. This then has to be true of every
term in the equations. It follows that µ must have units of 1/time (in this case,
days−1) and β should have units of per person per unit time
Example 6.5 (a convenient unit for time). Which parameter (or
combination of parameters) carries units of time? We will use such a quantity as a
reference scale for time. Explain what that time scale means.

Solution. Since µ has units of 1/time, it follows that 1/µ carries units of time (in this
case, days). This time unit is a typical time associated with recovery from infection.

Example 6.6 (scaling the model). Define x∗(t) as the fraction of the population
in the infected class and y∗(t) as the fraction of the total population in the susceptible
class. Argue that both of these are dimensionless, and that x∗ + y∗ = 1. Find a
dimensionless time variable using the results of Example 6.5. Then rewrite the model
equations in dimensionless form.

Solution. It is convenient to define dimensionless variables

We have identified I as being the (slightly more) interesting variable that deserves the
first label (x∗). Later we shall see that this proves convenient. However, to a large
extent these choices are arbitrary.
Since S(t)+I (t) = N= constant, it follows that

We will rewrite the relationships so obtained in a form that can be used to substitute directly
into the model equations, namely, S = y∗N, I = x∗N, t = t∗/µ. Substituting these in for the
𝑑𝑠 𝑑𝐼
variables in Eqs 𝑑𝑡 = 𝜇𝐼 − 𝛽𝑆𝐼, 𝑑𝑡 = 𝛽𝑆𝐼 − 𝜇𝐼 yields
Cancel the constant common factors of N and µ from both sides to
arrive at

Let us note that the equations contain a single remaining ratio of parameters that we will
denote by the notation R0 ≡ βN/µ. This proves to be an important quantity, and in fact the only
parameter combination that governs qualitative behavior. Rewriting the equations in terms of R0
and dropping the stars leads to
In this example, we illustrated the process of scaling on the original full model. As the
astute reader likely realizes, this was not entirely needed, as we could have opted to
scale the reduced single ODE version of the model.
Let us reconsider this process of scaling from the perspective of the reduced model,
𝑑𝐼
Eqn. 𝑑𝑡
= 𝛽 𝑁 − 𝐼 𝐼 − 𝜇𝐼 , where conservation was used. Then conservation means that y∗ +x∗ = 1
𝑑𝐼
and Eqn 𝑑𝑡 = 𝛽 𝑁 − 𝐼 𝐼 − 𝜇𝐼 can be similarly treated by the substitutions I = x∗N and t = t∗/µ:

Simplification (canceling factors of N and µ from both sides, and dropping the stars as before)
leads to

which, together with y = 1 − x, completely specifies the problem. The same dimensionless
parameter ratio, R0 = βN/µ has occurred in this equation so that we can rewrite the equation in
the form
6.3 analysis
Steady states :
Let us write a single ODE for the infected fraction of the population in several suggestive
forms, namely,

Example 6.8 (finding steady states). Find steady states of the model. Do
those steady states always exist? Interpret your results.

Solution. Steady states satisfy dx/dt = 0. From (6.12), this means

Clearly, there are two possible solutions. Either x = 0 and then (by conservation) y = 1.
This corresponds to a population that has no infected individuals, I (t) = 0,S(t) = N. We
will refer to this as the disease-free equilibrium. A second possibility is that x[(R0 −1)−
R0x] = 0 so x = (R0 −1)/R0 = 1−(1/R0). Using conservation once more, we find that in
this case y = 1/R0. This steady state has some proportion of the population in the
infected class, and is denoted the disease endemic state. However, we observe that
this steady state is biologically feasible only if x > 0, which means that R0 > 1. When
this steady state exists (which implies that it is positive), we say that the disease can
become endemic, which means that it can take hold of some constant fraction of the
population. See Exercise 6.2.
To summarize, we have found two possible steady states of the model, namely

Clearly, to convert these findings to results for the unit-carrying variables, we simply multiply
each quantity by N, so that S = yN, I = xN, and thus
 𝑑𝑥
= 0 → x[(𝑅0 -1)-𝑅0 x] = 0
𝑑𝑡
↓
x = 0 or -𝑅0 x = -(𝑅0 -1) → x = (𝑅0 -1)/𝑅0 = 1-1/𝑅0
The first case when x=0 by conservation y=1 → I(t)=0, S(t)=N This
corresponds to a population that has no infected individuals, this case
called Disease free.
The second case when x = 1 − 1/𝑅0 by conservation y = 1/𝑅0
This steady state has some proportion of the population in the infected
class and is denoted the Disease endemic state.
we observe that this steady state is biologically feasible only if x > 0,
which means that 𝑅0 > 1.
Now we have to cases.
1
1- Disease free: (𝑥0 = 0 , 𝑦0 = 1) 2- Disease endemic: 𝑥1 = 1 − 𝑦1 =
𝑅0
1
𝑅0

to convert these findings to results for the unit- variables, we simply

multiply each quantity by N, so that S = yN, I = xN, and thus
Disease free: 𝐼0 = 𝑥0 𝑁 = 0N = 0 , 𝑆0 = 𝑦0 𝑁 = 1N = N
1 1
Disease endemic: 𝐼1 = 𝑥1 𝑁 = 𝑁(1 − ) , 𝑆1 = 𝑦1 𝑁 = 𝑁( )
𝑅0 𝑅0

we see that the endemic steady state makes sense (is positive valued)
only when 𝑅0 > 1.
From either (6.14) or (6.15), we see that the endemic steady state makes sense (is
positive valued) only when R0 > 1. We summarize this important result by the following
threshold theorem.

Theorem 2. In a simple SI disease-dynamics model, the disease can become endemic

only if R0 > 1, where R0 (“R naught”) is the reproductive number of the disease, and R0
= βN/µ.

We will return to this result shortly. On a brief historical note, the notation R0 and its
name “R naught” originate from work by Robert May, who identified this quantity as the
important indicator of the fate of a communicable disease.
 Qualitative behavior:
Following once more the methods of Chapter 5, we now apply qualitative techniques to
understanding the behavior of the model. Since we are interested in the growth of the
infection, we opt to keep the equation for the fraction of infected individuals, (6.12).
However, we rewrite it in a slightly more convenient form,

Note that the expression 1 − (1/R0) is a steady state value obtained in (6.14), a constant.
We next consider the function f (x) on the RHS of the ODE (6.16).

Example 6.9 (qualitative flow diagram). Sketch a graph of the function f (x)
for two cases, (i) R0 > 1 and (ii) R0 < 1. Use your diagram to identify steady states, and
draw arrows along the x-axis to show the flow towards or away from those steady states.
Solution. We show the flow diagram in Fig. 6.2. From Fig. 6.2,
we can see that in the case R0 > 1, the disease will progress towards the endemic
steady state, x1 = 1−(1/R0). When R0 < 1, the only stable steady state is at x0 = 0, so
the disease is eradicated. The stability of the steady states can be seen directly

Figure 6.2. A sketch of the function f (x) for Eqn. (6.16) with arrows showing the direction
of change of x(t) from various initial values. (a) R0 > 1, so that a disease-endemic steady
state, x = 1−1/R0 (dark circle), exists. (b) R0 < 1. Only the steady state with no disease is
viable. The gray region is biologically irrelevant as x < 0.

Stable steady states (black dots) have arrows directed towards them, and unstable steady
states (open dots) have arrows directed away from them.