Uncertainty and Sensitivity Analysis of Ebola Virus Model

Syed Touqeer Hussain Shah : 2014-07-0003

December 9, 2014

1 Introduction
This part of the project consists of two parts. The first one comprises of the calculations of R0
and steady states, whereas, the second one is about Sensitivity and Uncertainty Analysis. Sen-
sitivity Analysis helps us determine the effect of several parameters on the value of R0 and the
Uncertainty analysis depicts the average value of all the parameters, their standard deviations
and their impact on determining R0 . Further, it helps in determining whether the disease is
going to be epidemic under given conditions or not.
Statistical tools of MATLAB are used in order to generated a normally distributed sample of the
parameters in a specified range (whose values are provided in detail in the previous part of the
project). In the analysis part, the next generation operator method is used in order to evaluate
R0 , The disease free equilibrium state and endemic equilibrium state is calculated.

2 The Vaccination Model

dSL
= Π(1 − p) − γ1 SL − µSL − λSL (1)
dt
dSH
= Πp − γ2 SH − ψH λSH − µSH , ψH > 1 (2)
dt
dV
= γ1 SL + γ2 SH − (1 − )λV − µV (3)
dt
dE
= λ(SL + ψH SH ) + (1 − )λV − (α + µ)E (4)
dt
dI
= αE − (τ + θI + δI + µ)I (5)
dt
dH
= τ I − (θH + δH + µ)H (6)
dt
dR
= θI I + θH H − µR (7)
dt
where
β(I + ηH)
λ=
N

3 Basic Properties
In this section he basic properties of the above model will be studied. We have the following
results.

1
Theorem 1 For any nonnegative initial data, there exists a unique solution
SL , SH , V, E, I, H, R respectively, defined on [0, ∞) to model (3.1). The solutions are nonnegative
and uniformly eventually bounded.
Proof.The system can be written as x0 = F (x), with
x = (SL , SH , V, E, I, H, R) and F (x) = (f1 , f2 , ..., f7 ). we want to apply theorem A.4 of “Math-
ematics in Population Biology” by Horst R. Thieme. It is easy to see that for every j = 1, ..., 7,
Fj (xj ) ≥ 0 of x ∈ [0, ∞)6 and xj = 0.
Now let x be a solution with values in Rn+ defined on some interval [0, b). It is sufficient to
show that x is bounded on [0, b). Adding the right hand side of model (3.1) we have

dN
= Π − µN − δH H − δI I ≤ Π − µN (8)
dt
Hence,

Π
N (t) ≤ N0 e−µt + 1 − e−µt


(9)
µ
and thus the solutions are bounded on every finite interval. Thus b = ∞. Thus solutions exists
for all nonnegative time. Moreover, from equation (9) it is clear that the solutions are uniformly
eventually bounded. 

Lemma 1 The closed set:

7 Π
B = {(SL , SH , V, E, I, H, R) ∈ R+ : SL + SH + V + E + I + H + R ≤ }
µ
is positively invariant and attracts all positive orbits in B.

Proof. Since dN dN Π
dt ≤ Π − µN and it follows that dt < 0 if N > µ . Thus from equation (9) of
above theorem we have, N (t) ≤ Π Π Π
µ if N0 ≤ µ . Thus, B is positively-invariant. Now if N (t) ≥ µ ,
then the solutions enter B in finite time, or that N (t) approaches Πµ . Also the infected classes,
which are E, I, H, approach 0. Therefore, the set B is attracting. 

4 Existence and Stability of Equilibria

4.1 Local stability of disease-free equilibrium (DFE)
The model (3.1) has a DFE, given by

E0 = (SL∗ , SH
∗
, V ∗ , E ∗ , I ∗ , H ∗ , R∗ ) = (SL∗ , SH
∗
, V ∗ , 0, 0, 0, 0, ) ,

with,
Π(1 − p) Πp γ1 SL∗ + γ2 SH
∗
SL∗ = ∗
, SH = , and V ∗ = .
µ µ µ
Following “Reproduction numbers and sub-threshold endemic equilibria for compartmental mod-
els of disease transmission” by P. van den Driessche and James Watmough (2002), the linear
stability of E0 can be established using the next generation operator method on system (3.1).
The matrices F (for the new infection terms) and V (of the transition terms) are given,
respectively, by
  
0 β(Ω1 + Ω2 ) βη(Ω1 + Ω2 )
F = 0 0 0 
0 0 0
 
K1 0 0
V = −α K2 0 ,
0 −τ K3

S ∗ + ψH SH
∗
(1 − )V ∗
where, Ω1 = L , Ω 2 = , K1 = α+µ, K2 = τ +θI +δI +µ, K3 = θH +µ+δH ,
N∗ N∗
Reproduction number Rc is given as

αβ(Ω1 + Ω2 ) αβητ (Ω1 Ω2 )

Rc = ρ(F V −1 ) = +
K1 K2 K1 K2 K3
where, ρ denotes the spectral radius.

5 Endemic Equilibrium
∗∗
The endemic equilibrium is given by E ∗∗ = (SL∗∗ , SH , V ∗∗ , E ∗∗ , I ∗∗ , H ∗∗ , R∗∗ ),

where
Π(1 − p)
SL∗∗ = ,
(λ∗∗ + γ1 + µ)

∗∗ Πp
SH = ,
(ψH λ∗∗ + γ2 + µ)
(1 − p)Πγ1 pΠγ2
V ∗∗ = +
J1 J3 J2 J3
λ∗∗ Π(1 − p) λ∗∗ (−1 + )Π(p(λ∗∗ + µ) + γ1 ) λ∗∗ pµΠ(−1 +  + ψH )
E ∗∗ = + −
K1 J1 K1 J1 J3 J2 J3 K1
αλ∗∗ Π(1 − p) αλ∗∗ (−1 + )Π(p(λ∗∗ + µ) + γ1 ) τ αλ∗∗ pµΠ(−1 +  + ψH )
I ∗∗ = + −
K1 J1 K2 K1 J1 J3 K2 J2 J3 K1 K2
τ αλ∗∗ Π(1 − p) τ αλ∗∗ (−1 + )Π(p(λ∗∗ + µ) + γ1 ) τ αλ∗∗ pµΠ(−1 +  + ψH )
H ∗∗ = + −
K1 J1 K2 K3 K1 J1 J3 K2 K3 J2 J3 K1 K2 K3
ΘI I ∗∗ + ΘH H ∗∗
R∗∗ = ,
µ
β(I ∗∗ +ηH ∗∗ )
where J1 = λ∗∗ + γ1 + µ, J2 = ψH λ∗∗ + γ2 + µ, J3 = λ∗∗ (1 − ) + µ, and λ∗∗ = N ∗∗
6 Sensitivity and Uncertainty
6.1 Sensitivity
THEORY: Mohsin’s part.
6.2 Uncertainty
THEORY: Mohsin’s part.
7 Conclusion
This part of the project is concerned with the working on the basic properties of the Vaccination
model as well as working on the Sensitivity and Uncertainty Analysis. The original model,
which allowed transmission of Ebola virus via human contact, was extended to include imperfect
vaccination. After the qualitative analysis of the vaccinated form of the Ebola model, numerical
techniques on MATLAB as a platform, were formulated and analyzed in detail in order to
evaluate how sensitive the value of R0 is, to the given parameters.
Since the parameter values estimation for simulations and numerical solutions was performed in
the previous part, it was easy to use all those mentioned parameters to work on the Sensitivity
and Uncertainty part. Several papers (given to us in classroom) were consulted in order to cook
up some decent and reasonable parameter values. These values were then used in order to get
some very promising results.