Supporting Information

Mukandavire et al. 10.1073/pnas.1019712108

SI Methods
Cholera Model. Putting the formulations described in Methods
together gives the following system of differential equations
describing the cholera model:
dS
dt
N
e
S
B
B

h
SI S;
dI
dt

e
S
B
B

h
SI I;
dR
dt
I R;
dB
dt
I B:
[S1]
Parameter Values. Table S1 shows cholera model parameters and
values and Table S2 shows estimated values of
e
and
h
and
95 % condence intervals.
Basic Reproductive Number and Stability. Stability of the disease-free
equilibrium. The disease-free equilibrium for the cholera model is
given by

0
N; 0; 0; 0: [S2]
We compute the basic reproductive number for the cholera model
using the method in van den Driessche and Watmough (3). Here,
the associated next generation matrices are given by
F
_
_
N
h
N
e

0 0
_
_
and V
_
_
0

_
_
:
The expression of the basic reproductive number is given by
R
0

N

e

h
R
e
R
h
; [S3]
where R
e

e
N= and R
h

h
N= are partial
reproductive numbers due to environment-to-human transmis-
sion and human-to-human transmission, respectively. We note
that for
e
= 0, R
0
= R
h
and when
h
= 0, R
0
= R
e
, suggesting
that the two modes of cholera transmission can independently
or jointly start an epidemic depending on conditions. Thus, pre-
cisely speaking R
0
measures the number of secondary cholera
infections generated in a wholly susceptible community when
a sufcient concentration of vibrios contaminates the aquatic en-
vironment and/or when a cholera-infected individual is intro-
duced into the community. In R
e
, 1= is the expected
time humans will be infected, = is the average amount
of V. cholerae shed per infected individual, 1= is the lifetime of
the vibrios in the environment, and
e
= is the number of new
cases generated in terms of vibrios per unit time, measured by the
ID
50
concentration. In R
h
;
h
= is the average amount of
hyperinfectious V. cholerae ingested by an infected individual.
It follows from theorem 2 in ref. 3 that the disease-free equi-
librium is locally asymptotically stable, when R
0
< 1. We establish
the global stability of the disease-free equilibrium using Theorem 1.
Theorem 1. (Castillo-Chavez et al., ref. 4) If a model system can
be written in the form
dX
dt
FX; Z;
dZ
dt
GX; Z; GX; 0 0;
[S4]
where X
m
denotes (its components) the number of uninfected
individuals and Z
n
denotes (its components) the number of
infected individuals including latent, infectious, etc. U
0
= (X*, 0)
denotes the disease-free equilibrium of the system. And assume
that (i ) For dX=dt FX; 0; X

is globally asymptotically stable

and (ii) GX; Z AZ

GX; Z;

GX; Z 0 for X; Z ;
where A = D
Z
G(X*, 0) is an M-matrix (the off diagonal elements
of A are nonnegative) and is the region where the model makes
biological sense. Then the xed point U
0
= (X*, 0) is a globally
asymptotic stable equilibrium of cholera model system Eq. S1 provided
that R
0
< 1.
We begin by showing condition i as
FX; 0
_
N S
R
_
;
and solving these two ordinary differential equations gives
Rt R0e
t
and St N N S0e
t
:
Thus, R(t) 0 and S(t) N as t , regardless of the values of
R(0) and S(0). Thus,
0
is globally asymptotically stable.
Next, applying Theorem 1 to cholera model system Eq. S1 gives
GX; Z
_

e
SB
B

h
SI I
I B
_
;
and
A
_

h
N

e
N

_
;
which is clearly an M-matrix. Meanwhile, we nd

GX; Z
_

h
N S I

e
NB

e
SB
B
0
_
:
Because 0 S N, it follows that

GX; Z 0: We summarize
the result in Lemma 1.
Lemma 1. The disease-free equilibrium of the model system,
given by
0
, is globally asymptotically stable whenever R
0
< 1.
Local stability of the endemic equilibrium. The endemic equilibrium of
the cholera model is given by * = (S*, I*, R*, B*), where
S

N
I

; [S5]
I

e
S

h
S

; [S6]
R

; [S7]
Mukandavire et al. www.pnas.org/cgi/content/short/1019712108 1 of 6
We establish the following theorem.
Lemma 2. If R
0
> 1, a unique endemic equilibrium exists and is
locally asymptotically stable.
Proof. By solving Eqs. S5S8, we obtain
I

AI
2
BI

C 0; [S9]
where
A
h
;
B
h
N
e

h
;
C
e
N
h
N:
From Eq. S9, we have I* = 0, which corresponds to the disease-
free equilibrium and a quadratic equation given by
AI
2
BI

C 0: [S10]
The roots of this quadratic equation must satisfy
I

1
I

2

C
A
[S11]
and
I

1
I

2

B
A
: [S12]
With
R
0

N
e

h

and defining S
c

e

h
gives N > S
c
when R
0
> 1. It follows that A < 0 and C > 0, and
the right-hand side of Eq. S11 is <0. Thus, the quadratic Eq. S10
has a unique positive root I*. On the other hand, if R
0
< 1, which
yields N < S
c
, we can obtain C < 0 so that the right-hand side
of Eq. S11 >0, and the right-hand side of Eq. S12 <0 if B < 0.
In fact, we have ( + ) > N
h
for N < S
c
, and thus we obtain
( + ) > N
h
, which gives B < 0. In this case, Eq. S10 has
two negative roots and the positive endemic equilibrium does
not exist. Hence it is not biologically feasible.
To deduce the local stability of the endemic equilibrium, we use
the Jacobian of cholera model system Eq. S1. For simplicity we
set
P
e
B

= B

h
I

; Q
e
S

= B

2
; and the Jaco-
bian matrix becomes
J

B

_

_
P
h
S 0 Q
P
h
S 0 Q
0 0
0 0
_

_
:
The characteristic equation of the matrix J

B
is given by
Det I J

B

h
S

P Q 0: [S13]
Clearly, Eq. 13 has a negative root = . Expanding Eq. 13
gives ( + )(
h
S* + + )( + ) + P( + + )( + )
Q( + ), which we write as
a
0

3
a
1

2
a
2
a
3
0; [S14]
where
a
0
1;
a
1
P 2
h
S

;
a
2

2
P P P Q 2
h
S

h
S

;
a
3

2
P P Q S

h
:
Rouths stability criterion (5) requires
a
1
>0; a
2
>0; a
3
>0 and a
1
a
2
a
0
a
3
>0
as the necessary and sufcient conditions for stability.
From Eq. S6, we have the expression of ( + ) at the positive
endemic equilibrium,

S

e
S

2
I

2

h
S

:
Note that Q
2

e
S

= I

2
; and we have two con-
ditions that we use to prove the necessary and sufcient con-
ditions, which are
>
h
S

[S15]
and
>Q
h
S

: [S16]
First, we prove that a
1
> 0 using condition Eq. S15 and P > 0. We
have
a
1
P
h
S

>0: [S17]
Second, using the two inequalities ( + ) > Q +
h
S* and
( + ) >
h
S*, it can be shown that a
2
> 0. In addition, it can
also be shown that a
3
> 0 by using ( + ) > Q +
h
S*.
We show that a
1
a
2
a
0
a
3
> 0 in the following:
a
2
a
3

2

2
Q S

2

3

2
S

2
S

P
2
P
2

2

2
P P:
Using conditions Eqs. S15 and S16, we have

2

2
>Q S

2
and

2

3

2
>S

2
S

h
:
Thus, a
2
( + ) > a
3
holds. Because a
1
> ( + ), on the basis of
Eq. S17, we obtain a
1
a
2
> a
3
. Thus, we have shown that when
R
0
> 1, a unique endemic equilibrium is locally asymptotically
stable.
Global stability. By considering the following domain that is a result
of a nondimensionalized model system Eq. S1,
D fS; I; BjS 0; I 0; S I 1; B 0g;
we construct the Lyapunov function
V w
1
S S

2
w
2
I I

2
w
3
BB

2
[S18]
with w
1
> 0, w
2
> 0, and w
3
> 0.
Note that for the endemic equilibrium * = (S*, I*, B*), we have
the following three equations for the nondimensionalized system:

e
S

h
S

0; [S19]

e
S

h
S

0; [S20]
B

: [S8]
Mukandavire et al. www.pnas.org/cgi/content/short/1019712108 2 of 6
I

0: [S21]
From Eqs. S20 and S21, we have

h
S

e
S

>
h
S

e
S

B
:
Using Eqs. S19S21, we obtain
where Y = [S S*, I I*, B B*], W = diag(w
1
, w
2
, w
3
), and
A

e
B
B

h
I
h
S

e
S

e
B
B

h
I
h
S

e
S

B
0
_

_
_

_
:
[S24]
The global stability of * will be established if we can show the
matrix A dened in Eq. S24 is VolterraLyapunov stable (6); i.e.,
a positive diagonal matrix W exists such that WA + A
T
W is
negative denite.
Theorem 2. Let A be a 2 2 matrix (7, 8). Then
A
_
a
11
a
12
a
21
a
22
_
is VolterraLyapunov stable if and only if a
11
< 0, a
22
< 0, and
a
11
a
22
a
12
a
21
> 0.
Theorem 3. Let A be a nonsingular n n matrix, where
n 2, with inverse A
1
= B and W a positive diagonal n n
matrix (6). Let A*, B*, and W* denote the (n 1) (n 1)
matrices obtained from A, B, and W, respectively, by deleting
the last row and column. Then (i ) if WA + (WA)
T
> 0, we
must have a
nn
> 0, W*A* + (W*A*)
T
> 0, and W*B* +
(W*B*)
T
> 0; and (ii ) if a
nn
> 0, W*A* + (W*A*)
T
> 0, and
W*B* + (W*B*)
T
> 0, it is possible to choose w
n
> 0 such that
WA + (WA)
T
> 0.
From Eq. S24, we obtain
dV
dt
2w
1
S S

e
B
SB
h
SI S

e
B

h
S

_
2w
2
I I

_

e
B
SB
h
SI I

e
B

h
S

I I

_
2w
3
BB

I BI

2w
1
S S

_

e
B
SB

e
B
S

B

e
B
S

B

e
B

h
SI S

I S

I S

2w
1
S S

2w
2
I I

e
B
SB

e
B
S

B

e
B
S

B

e
B

h
SI S

I S

I S

2w
2
I I

2
2w
3
BB

I I

BB

2w
1

e
B
B
h
I S S

2
2w
1

h
S

S S

I I

2w
1

e
S

B
S S

BB

2w
2

e
B
B
h
II I

S S

2w
2

h
S

I I

2
2w
2

e
S

B
I I

BB

2w
3
BB

I I

2w
3
BB

2
YWA A
T
WY
T
;
[S23]
A
1

1
det A

h
S

e
S

B

h
S

e
S

B


e
S

h
I

e
B
B

h
I

e
B
B


e
S

h
I

e
B
B

h
I

e
B
B

h
S

e
B
B

h
I
_

_
_

_
;
Mukandavire et al. www.pnas.org/cgi/content/short/1019712108 3 of 6
where
It can easily be shown that det A < 0 because

h
S

e
S

= B

B >0 (Eq. S22). On

the basis of Theorem 2, it is straightforward to verify that (A
1
)*
is VolterraLyapunov stable. Hence, a 2 2 positive diagonal
matrix W* = diag(w
1
, w
2
) exists such that W*(A
1
)* + (W*(A
1
)*)
T
< 0. Setting M = (A)
1
, we have W*M* + (W*M*)
T
> 0.
After some calculation, we obtain the matrices W*M* +
(W*M*)
T
and W*(A)* + (W*(A)*)
T
, as
det AW

T
Q;
where
and
It can easily be shown that
Hence, the matrix W*(A)* + (W*(A)*)
T
is positive denite.
On the basis of Theorem 3, w
3
> 0 exists such that W(A) +
(A)
T
W
T
> 0; i.e., WA + A
T
W
T
< 0. We have thus established
the following result.
Theorem 4. The endemic equilibrium * is globally asymptoti-
cally stable when R
0
> 1.
We illustrate the existence of the unique globally asymptotically
stable endemic equilibrium using a phase plane portrait in Fig. S1
based on a set of varying initial conditions and parameter values
in Table S1.
Transmission Routes and Model Dynamics. In Fig. S2 we present
numerical simulation results illustrating the contribution of the
transmission routes in the dynamics of cholera for the following
scenarios: (i) R
0
> 1, R
e
< 1 < R
h
; (ii) R
0
> 1, R
e
> 1 > R
h
; (iii)
R
0
< 1, R
e
< R
h
< 1; and (iv) R
0
< 1, R
e
> R
h
< 1.
Q
2w
1
_

h
S

e
S

B
_
w
2
_

h
I

e
B
B
_
w
1
_

h
S

e
S

B
_
w
2
_

h
I

e
B
B
_
w
1
_

h
S

e
S

B
_
2w
2
_

h
I

e
B
B
_
_

_
_

_
det A
_

e
B
B

h
I
_

_

h
S

e
S

B
_

h
I

e
B
B
__

h
S

e
S

B
_
:
We show that W*(A)* + (W*(A)*)
T
> 0. In fact, because
W*M* + (W*M*)
T
is positive denite, and det A > 0, we have
W

_
2w
1
_

h
I

e
B
B
_
w
1

h
S

w
2
_

h
I

e
B
B
_
w
1

h
S

w
2
_

h
I

e
B
B
_
2w
2

h
S

_
:
1. WHO (2010) Zimbabwe. Available at http://www.who.int/countries/zwe/en/. Accessed
April 5, 2010.
2. Hartley DM, Morris JG, Jr., Smith DL (2006) Hyperinfectivity: A critical element in the
ability of V. cholerae to cause epidemics? PLoS Med 3:e7.
3. van den Driessche P, Watmough J (2002) Reproduction numbers and sub-threshold
endemic equilibria for compartmental models of disease transmission. Math Biosci 180:
2948.
4. Castillo-Chavez C, Feng Z, Huang W (2002) On the computation of R0 and its role on
global stability. Available at math.la.asu.edu/chavez/2002/JB276.pdf. Accessed April 5,
2010.
5. Katsuhiko O (1970) Modern Control Engineering (Prentice-Hall, Inc., Englewood Cliffs,
NJ), pp 252258.
6. Redheffer R (1985) Volterra multipliers ii. SIAM J Alg Disc Math 6:612623.
7. Cross GW (1978) Three types of matrix stability. Linear Algebra Appl 20:253263.
8. Goh BS (1976) Global stability in two species interactions. J Math Biol 3:313318.
detW

4w
1
w
2

h
S

_

h
I

e
B
B
_
2w
1
w
2

h
S

h
I

e
B
B
_
w
2
2
_

h
I

e
B
B
_
2
w
2
1
S

2
h
>0:
det
_
det A
_
W

T
__
4w
1
w
2

2

h
S

_

h
I

e
B
B
_
2w
1
w
2

h
S

2
_

h
I

e
B
B
_
w
2
2

2
_

h
I

e
B
B
_
2
w
2
1
S

2
h

2
w
2
1
_

e
S

B
_
2
2w
1

h
S

e
S

B
4w
1
w
2

e
S

B
2w
1
w
2

e
S

B
_

h
I

e
B
B
_
>0:
Mukandavire et al. www.pnas.org/cgi/content/short/1019712108 4 of 6
Fig. S2. An illustration of the contribution of each transmission route for (A) R
0
> 1, R
e
< 1 < R
h
; (B) R
0
> 1, R
e
> 1 > R
h
; (C) R
0
< 1, R
e
< R
h
< 1; and (D)
R
0
< 1, R
e
> R
h
< 1 using parameter values in Table S1 and choosing
e
and
h
arbitrarily and a suitable population size. The blue line denotes contribution
from the environment, the black line denotes humanhuman transmission, and the red line is a combination of both modes of transmission.
0 10 20 30 40
0
20
40
60
80
100
Zimbabwe
Time (weeks)
C
u
m
u
l
a
t
i
v
e

C
h
o
l
e
r
a

C
a
s
e
s

(
x

1
0
0
0
)
Fig. S3. Cholera model tting for the cumulative cholera cases where the thick red line represents the model t and the circles mark the reported data for the
cumulative number of cholera for Zimbabwe using parameter values in Table S1 and population size in Table 1.
0.98 1 1.02 1.04
x 10
4
0.5
1
1.5
S
I
Fig. S1. The endemic equilibrium point of I versus S with six different initial conditions and using parameter values in Table S1, with
e
= 0.2668 and
h
=
5.8991 10
5
.
Mukandavire et al. www.pnas.org/cgi/content/short/1019712108 5 of 6
Fig. S4. Cholera model tting for the cholera cases. The red line represents the model t, and the blue dashed line and the circles mark the reported data for
cholera cases for Zimbabwe using parameter values in Table S1 and population size in Table 1.
0 20 40 60 80 100
1.145
1.15
1.155
1.16
1.165
1.17
1.175
1.18
Percentage of reported symptomatic cholera cases
B
a
s
i
c

r
e
p
r
o
d
u
c
t
i
v
e

n
u
m
b
e
r
Fig. S5. The relationship between R
0
and the percentage of symptomatic cholera cases reported using data from Zimbabwe, parameter values from Table S1,
and population size from Table 1.
Table S1. Cholera model parameters and values
Parameter Symbol Value Source
Natural human birth and death rate (43.5 y)
1
(1)
Concentration of V. cholerae in environment (ID
50
) 10
6
cells/mL (2)
Rate of recovery from cholera (5 d)
1
(2)
Rate of contribution to V. cholerae in the aquatic environment 10 cellsmL
1
d
1
per person (2)
Death rate of vibrios in the environment (30 d)
1
(2)
Table S2. Estimated values of
e
and
h
and 95% condence intervals

e

e
95% CI
h

h
95% CI
Harare 2.1 (1.312.88) 0.00043 (0.00040.005)
Bulawayo 0.94 (0.461.42) 0.0024 (0.00200.0027)
Mashonaland Central 0.87 (0.471.27) 0.0016 (0.00150.0017)
Mashonaland East 1.75 (1.222.28) 0.00077 (0.00070.0009)
Mashonaland West 1.13 (0.0382.22) 0.0017 (0.00140.0019)
Midlands 0.23 (0.0350.43) 0.0012 (0.00110.0013)
Manicaland 0.55 (0.280.82) 0.0016 (0.00140.0017)
Matebeleland South 9.62 (4.5714.8) 0.0026 (0.0010.0041)
Matebeleland North 0.85 (0.231.48) 0.003 (0.00260.0033)
Masvingo 0.85 (0.161.54) 0.0014 (0.00110.0016)
Zimbabwe 0.075 (0.0550.094) 0.00011 (0.0001050.000111)
Mukandavire et al. www.pnas.org/cgi/content/short/1019712108 6 of 6