SIR

without demography

The SIR model without demography (fixed population) reads

!" 𝑆(𝑡 = 0) = 𝑆(0)
= −𝛽𝑆𝐼 [1]
!#

!$ 𝐼(𝑡 = 0) = 𝐼(0)
= +𝛽𝑆𝐼 − 𝛾𝐼 [2]
!#

!%
= +𝛾𝐼 [3] 𝑅(𝑡 = 0) = 𝑅(0)
!#

Recall S=X/N, I=Y/N and R=Z/N, i.e., fraction of susceptible, infected and removed.
!" !$ !%
Moreover, !# + !# + !# = 0 (Fixed population, people only move between compartments)
and 𝑆 + 𝐼 + 𝑅 = 1.

(1) Threshold phenomena

Consider the initial stages after I(0) infectives are introduced into a population consisting
of S(0) susceptibles. What determines if we have an epidemic or not?
Rewrite [2]:
!$
!#
= 𝐼(𝛽𝑆 − 𝛾) [4]

!$
Investigate, !# = 0, so 𝐼(𝛽𝑆 − 𝛾) = 0 leading to 𝐼 = 0 𝑜𝑟 𝑆 = 𝛾⁄𝛽. The first solution, 𝐼 = 0
is a disease-free state (hence no new infections).

!$ &
When 𝑆 = 𝛾 ⁄𝛽 then [4] is !# = 𝐼 2𝛽 ' − 𝛾3 = 0

Define the basic reproductive rate as 𝑹𝟎 = 𝜷⁄𝜸. So we find

𝑑𝐼 𝑑𝐼
> 0 < 0
𝑑𝑡 𝑑𝑡
𝑆 > 19𝑅 𝑆 < 19𝑅
) )

If, at 𝑡 = 0, we have 1 infected individual 𝐼(0) = 1/𝑁 and 𝑆(0) = 1 − 1/𝑁 (assume no
removed, herd/existing immunity). The infection will spread if 𝑆(0) > 1⁄𝑅) , which in the
limit of large 𝑁 leads to 𝑆(0) = 1 (everyone initially susceptible), 1 > 1⁄𝑅) ⇒ 𝑅) > 1 .
So, the basic reproductive ratio 𝑅) , “the average number of secondary cases arising
from a single primary case in an entirely susceptible population”, should be larger
than 1 to spread the disease, as expected.

(2) Epidemic Burnout
Look at the long-term (“asymptotic”) state

2.1 Fixed points

𝑑𝑆 ∗ −𝛽𝑆 ∗ 𝐼 ∗ = 0
= 0
𝑑𝑡
𝑑𝐼 ∗ +𝛽𝑆 ∗ 𝐼 ∗ − 𝛾𝐼 ∗ = 0
= 0
𝑑𝑡 +𝛾𝐼 ∗ = 0

 𝑑𝑅∗
= 0
𝑑𝑡

!" !$ !%
So, 𝐼 ∗ = 0 is when the system is stable, given 𝐼 ∗ = 0 and !# + !# + !# = 0, 𝑆 ∗ + 𝑅∗ = 1
leading to the fixed point (𝑆 ∗ , 0 ,1 − 𝑆 ∗ ). First, this is a disease free state, so the SIR
without demography does not have a endemic state. Second, there are many fixed points,
𝑆 ∗ depends on the initial conditions, so on 𝑆(0), 𝐼(0), 𝑅(0).

2.2 What happens at 𝒕 → ∞, given initial conditions?

Look at change in susceptible as a function change in removed:

𝑑𝑆 𝑑𝑆⁄𝑑𝑡 𝛽𝑆
= =− = −𝑅) 𝑆
𝑑𝑅 𝑑𝑅 𝑑𝑡 ⁄ 𝛾

Integration with respect to R gives:

𝑆(𝑡) = 𝑆(0)𝑒 +%! %(#)

The first important implication is that

𝑒 +%! %(#) > 0 𝑠𝑜 𝑆(𝑡) > 0

And

0 ≤ 𝑅(𝑡) ≤ 1 ⇒ 𝑆(0)𝑒 +%! ≤ 𝑆(𝑡) ≤ 𝑆(0),

so,

• Always some susceptible left;
• The chain of transmission eventually breaks due to the decline in infectives, not
due to a complete lack of susceptibles.

Moreover,
𝑆(∞) = 1 − 𝑅(∞) = 𝑆(0)𝑒 +%! %(.)
𝑆 + 𝐼 + 𝑅 = 1, 𝐼(∞) = 0, and 𝑆(∞) = 1 − 𝑅(∞)
⇒ 1 − 𝑅(∞) − 𝑆(0)𝑒 +%! %(.) = 0 [5]

This equation can be used to find (no exact solution) 𝑅(∞) , so the total number of
infecteds after the epidemic, as a function of 𝑆(0) and 𝑅) . But note that when 𝑅(∞) = 0
[5] is positive, when 𝑅(∞) = 1 it’s negative so in between there must be a solution.

(3) Epidemic Curve

Let us finally derive an approximate equation (non-linear term 𝛽𝑆𝐼 makes exact not
possible) for the number of new cases per time interval. For this we will explore

𝑑𝑅
= 𝛾𝐼 = 𝛾(1 − 𝑆 − 𝑅) = 𝛾E1 − 𝑆(0)𝑒 +%! %(#) − 𝑅F
𝑑𝑡

Assuming 𝑅) 𝑅 is small, we can Taylor expand the exponential to 2nd order without large
error (note if 𝑅) is large this isn’t possible, so this approximation will not work for highly
infectious diseases).

Recall Taylor expansion:
𝑥# 𝑥 $ 𝑥 % 𝑥 &
!𝑒 !" = 1 − + − + + ⋯ -
1! 2! 3! 4!

Taylor expand the exponential to 2nd order
1
𝑒 +%! % = 1 − 𝑅) 𝑅 + (𝑅) 𝑅)/
2
𝑑𝑅 1
= 𝛾 H1 − 𝑆(0)[1 − 𝑅) 𝑅 + (𝑅) 𝑅)/ ] − 𝑅K
𝑑𝑡 2
1
= 𝛾 H1 − 𝑆(0) + (𝑆(0)𝑅) − 1)𝑅 + 𝑆(0)𝑅)/ 𝑅/ K
2

It’s messy but the equation can be solved (see e.g. the paper by Cormack and McKendrick),
we just state the result

1 1
𝑅(𝑡) = / H𝑆(0)𝑅) − 1 + 𝛼 tanh H 𝛼𝛾𝑡 − 𝜙KK
𝑅) 𝑆(0) 2
0
𝛼 = [(𝑆(0)𝑅) − 1)/ + 2𝑆(0)𝐼(0)𝑅)/ ]/
1
𝜙 = tan+0 R (𝑆(0)𝑅) − 1)S
𝛼

and finally

𝑑𝑅 𝛾𝛼 / 1
= / sech/ H 𝛼𝛾𝑡 − 𝜙K
𝑑𝑡 2𝑅) 𝑆(0) 2