Content uploaded by Abel Lucido

Author content

All content in this area was uploaded by Abel Lucido on Nov 16, 2020

Content may be subject to copyright.

Content uploaded by Abel Lucido

Author content

All content in this area was uploaded by Abel Lucido on Nov 03, 2020

Content may be subject to copyright.

Philippine Science Letters Vol. 13 | No. 02 | 2020

158

Prevention of H5N6 outbreaks in the

Philippines using optimal control

Abel G. Lucido1,2,3, Robert Smith?4, and Angelyn R. Lao2,3

1Department of Science & Technology - Science Education Institute, Bicutan, Taguig, Philippines

2Mathematics and Statistics Department, De La Salle University, 2401 Taft Avenue, 0922 Manila,

Philippines

3Center for Complexity and Emerging Technologies, De La Salle University, 2401 Taft Avenue, 0922

Manila,Philippines

4Department of Mathematics and Faculty of Medicine, University of Ottawa, 150 Louis-Pasteur Pvt

Ottawa, ON K1N 6N5, Canada

ighly Pathogenic Avian Influenza A (H5N6) is a

mutated virus of Influenza A (H5N1) and a new

emerging infection that recently caused an outbreak

in the Philippines. The 2017 H5N6 outbreak

resulted in a depopulation of 667,184 domestic

birds. We incorporate half-saturated incidence and optimal

control in our mathematical models in order to investigate three

intervention strategies against H5N6: isolation with treatment,

vaccination, and modified culling. We determine the direction

of the bifurcation when R0=1 and show that all the models

exhibit forward bifurcation. We apply the theory of optimal

control and perform numerical simulations to compare the

consequences and implementation cost of utilizing different

intervention strategies in the poultry population. Despite the

challenges of applying each control strategy, we show that

culling both infected and susceptible birds is an effective control

strategy in limiting an outbreak, with a consequence of losing a

large number of birds; the isolation-treatment strategy has the

potential to prevent an outbreak, but it highly depends on rapid

isolation and successful treatment used; while vaccination alone

is not enough to control the outbreak.

KEYWORDS

Influenza A (H5N6), half-saturated incidence, isolation-

treatment, culling, vaccination, bifurcation, optimal control

1. INRODUCTION

Avian influenza is a highly contagious disease of birds caused

by infection with influenza A viruses that circulate in domestic

and wild birds (WHO 2020). Some avian influenza virus

subtypes are H5N1, H7N9 and H5N6, which are classified

according to combinations of different virus surface proteins

hemagglutinin (HA) and neuraminidase (NA). This disease is

categorized as either Highly Pathogenic Avian Influenza (HPAI),

which causes severe disease in poultry and results in high death

rates, or Low Pathogenic Avian Influenza (LPAI), which causes

mild disease in poultry (WHO 2020).

As reported by the World Health Organization (WHO), H5N1

has been detected in poultry, wild birds and other animals in over

30 countries and has caused 861 human cases in 16 of these

countries and 455 deaths. H5N6 was reported emerging from

China in early May 2014 (Joob and Viroj 2015). H5N6 has

replaced H5N1 as one of the dominant avian influenza virus

subtypes in southern China (Bi et al. 2016). In August 2017,

cases of H5N6 in the Philippines resulted in the culling of

667,184 chicken, ducks and quails (OIE 2020).

Due to the potential of avian influenza virus to cause a pandemic,

several mathematical models have been developed in order to

test control strategies. Several studies included saturation

H

ARTICLE

*Corresponding author

Email Address: abel_lucido@dlsu.edu.ph

Date received: February 28, 2020

Date revised: August 27, 2020

Date accepted: October 3, 2020

Vol. 13 | No. 02 | 2020 Philippine Science Letters

159

incidence, where the rate of infection will eventually saturate,

showing that protective measures have been put into place as the

number of infected birds increases (Capasso and Serio 1978).

With half-saturated incidence, it includes the half-saturation

constant, which pertains to the density of infected individuals

that yields 50% chance of contracting the disease (Shi et al.

2019). When half-saturated incidence is included, the effect is a

significantly lower peak of the total number of infected humans

compared to the case when half-saturated incidence is not

included (Chong et al. 2013). However, when half-saturation is

included, the disease takes longer to die off. We are thus using

half-saturated incidence to investigate the effects of outbreaks

that may have a long tail. Some intervention strategies employed

to protect against avian influenza are biosecurity, quarantine,

control in live markets, vaccination and culling.

Emergency vaccination, prophylactic or preventive vaccination,

and routine vaccination are the three vaccination strategies

mentioned by the United Nations Food and Agriculture

Organization (UNFAO). In China, A(H5N1) influenza infection

caused severe economic damage for the poultry industry, and

vaccination served a significant role in controlling the spread of

this infection since 2004 (Chen 2009). UNFAO and Office

International des Epizooties (OIE) of the World Organization for

Animal Health suggested vaccination of flocks should replace

mass culling of poultry as the primary control strategy during

outbreak (Butler 2005). For this reason, many mathematical

models focus on how vaccination could prohibit the spread of

infection.

Culling is a widely used control strategy during an outbreak of

avian influenza virus (AIV). During the 2017 outbreak of H5N6

in the Philippines, mass culling was implemented to control the

spread of AIV. Gulbudak and Martcheva (2013) suggested a

modified culling strategy, which involves culling only the

infected birds and high-risk in-contact birds. They utilized a

function to represent the culling rate considering both HPAI and

LPAI. Gulbudak et al. (2014) used half-saturated incidence to

describe the culling of infected birds. The two-host model of Liu

and Fang (2015) showed that screening and culling of infected

poultry is a critical measure for preventing human A(H7N9)

infections in the long term. There is a limited understanding of

the effects of isolation with treatment as a control strategy to

counter the spread of avian influenza. Isolation is also used when

adding a new flock of birds to the poultry farm in order to

prevent the possible transmission of disease to the current flock.

The importance of optimal-control theory in modeling infectious

diseases has been highlighted by several recent studies. Agusto

(2013) used optimal control and cost-effective analysis in a two-

strain avian influenza model. Jung et al. (2009) used optimal

control in modeling H5N1 to prevent an influenza pandemic.

Kim et al. (2018) utilized an optimal-control approach in

modeling tuberculosis (TB) in the Philippines. Okosun and

Smith? (2017) used optimal control to examine strategies for

malaria–schistosomiasis coinfection.

To the best of our knowledge, optimal-control theory has not

been applied to the spread of infectious diseases with

transmission represented by half-saturated incidence. In this

study, we adapt the vaccination model and modify the isolation

model of Lee and Lao (2017). We modify the isolation model by

partitioning the outflow of birds from isolation into two

compartments. A proportion of birds will transfer to the

recovered population, while the remainder will return to the

infected population. We focus on the poultry population and use

half-saturated incidence to describe the transmission of AIV. We

include a modified culling strategy as one of our control

strategies and use half-saturated incidence to depict the modified

culling of susceptible and infected birds. We apply optimal-

control theory to our three strategies — isolation-treatment,

vaccination, and culling — and determine which among these

strategies can inhibit the occurrence of an AIV outbreak.

2. THE MODELS

We examine three control strategies: isolation-treatment, culling,

and vaccination. Our mathematical models are in the form of

half-saturated incidence (HSI), which takes into consideration

the density of infected individuals in the population that yields

50% chance of contracting avian influenza. Mathematical

models with half-saturated incidence are more realistic

compared to models with bilinear incidence (Chong et al. 2013,

Lee and Lao 2018). We present four mathematical models: a

model without control, which describes the transmission

dynamics of avian influenza in bird population (i.e., the AIV

model), and three models obtained by applying the following

intervention strategies: isolation with treatment, vaccination,

and culling. Description of variables and parameters used in the

models are listed in the table in Appendix A.

2.1. AIV model without intervention

Figure 1: Schematic diagram of the AIV model with half-saturated

incidence.

In the AIV model without intervention (shown in Figure 1) the

bird population is divided into subpopulations (represented by

compartments): susceptible birds () and infected birds (). The

total bird population is represented by () at time , where

() = () + (). The number of susceptible birds increases

through the birth rate Λ and reduces through the natural death

rate of birds () which are both constant parameter values.

Infected birds additionally decrease through the disease-specific

death rate ().

The number of susceptible birds who become infected through

direct contact is represented by

, which denotes the transfer

of the susceptible bird population to the infected bird population.

Note that is the rate of transmitting AIV and is the half-

saturation constant, indicating the density of infected individuals

in the population that yields 50% possibility of contracting

avian influenza (Chong et al. 2013). The saturation effect of the

infected bird population indicates that a very large number of

infected may tend to reduce the number of contacts per unit of

time due to awareness of farmers to the disease (Capasso and

Serio 1978). In Figure 1, the dashed directional arrow from to

the arrow from to indicates that

is regulated by .

Based on the AIV model described above, we have the following

system of nonlinear ordinary differential equations (ODEs):

=

+,

=

+(+).

2.2. Confinement with treatment strategy for infected poultry

(isolation-treatment model)

(1)

Philippine Science Letters Vol. 13 | No. 02 | 2020

160

Figure 2: Schematic diagram of isolation-treatment model with

HSI.

Here, we employ the strategy of confining and treating the

infected poultry population (which will be referred as the

isolation-treatment strategy). Several studies concluded that

reducing the contact rate is an effective measure in preventing

the spread of infection into the population (Lee and Lao 2018,

Teng et al. 2018). For the isolation-treatment model (shown in

Figure 2), we have included the compartment representing the

population of isolated birds that undergoes treatment (T) and the

compartment representing the population of recovered birds (R).

We denote the isolation rate of identified infected birds by ψ and

the release rate of birds from isolation by γ.

During isolation, we apply treatment then release birds

afterward. These birds will either recover successfully (transfer

to recovered population) or remain infected (return to the

infected population) depending on the effectiveness of treatment.

The proportion of isolated birds that have recovered is

represented by f, while the proportion of isolated birds that have

not recovered (and so remained infected) are represented by (1-

f). We did not consider natural recovery of poultry in our model,

due to the high mortality rate of HPAI virus infection.

The system of ODEs for the isolation-treatment model is

=

+,

=

++(1)(++),

=(++),

=.

2.3. Immunization of the poultry population (vaccination model)

Figure 3: Schematic diagram of preventi ve vaccination model

with HSI.

We modified the vaccination model of Lee and Lao (2018) by

splitting the birth rate () depending on the proportion of

vaccinated population (), as shown in Figure 3. The poultry

population prone to H5N6 is divided into two compartments:

vaccinated birds represented by and susceptible, unvaccinated

birds denoted by . In our vaccination model, we differentiate

the immunized group (vaccinated) from non-immunized group

(unvaccinated).

We investigate the effectiveness of the vaccine not only through

its reported efficacy (denoted by ) but also based on the waning

rate of the vaccine (denoted by ). To represent the acquired

immunity of the vaccinated group, the infectivity of vaccinated

birds is reduced by a factor

(1). The system of ODEs

representing the vaccination model is

=(1)+

+,

=(1)

+(+),

=

++(1)

+(+).

2.4. Depopulation of susceptible and infected birds (culling

model)

Figure 4: Schematic diagram of depopulation or culling model

with HSI.

We modified the culling model of Gulbudak et al. (2014) by

incorporating the dynamics of half-saturated incidence on the

transmission of infection and on the culling rate for infected

birds and for susceptible birds that are at high risk of infection.

We define the culling function of the infected and susceptible

birds as =

and =

, respectively. The culling

frequency is represented by for susceptible birds and for

infected birds. The following system of ODEs represents the

culling model:

=

+(),

=

+()(+).

3. STABILITY AND BIFURCATION ANALYSIS

We first analyze the AIV model without intervention. The

disease-free equilibrium (DFE) of the AIV model (1) is

=(,)=

, 0.

We denote the basic reproduction number as for the AIV

model and obtain

=

(+).

The DFE

of the AIV model is locally asymptotically stable

if < 1 and unstable if > 1.

The endemic equilibrium for the AIV model is represented by

=(, )=()

,()

()().

(2)

(3)

(4)

(5)

(6)

Vol. 13 | No. 02 | 2020 Philippine Science Letters

161

We can rewrite as

=

+(1).

When 1, it follows that 0, so there is no biologically

feasible endemic equilibrium. For > 1, we have > 0, so

we have an endemic equilibrium. We conclude that the AIV

model has no endemic equilibrium when 1, and has an

endemic equilibrium when > 1. It follows that reducing the

basic reproduction number below one is sufficient to

eliminate avian influenza from the poultry population.

As exhibited in Figure 5A, we have a bifurcation plot between

the infected population and the basic reproduction number .

When the basic reproduction number is below one and the DFE

and the endemic equilibrium coexist, then we have a backward

bifurcation. A forward bifurcation is when the basic

reproduction number crosses one from below and the DFE

becomes unstable while the endemic equilibrium becomes stable.

Clearly, we have a forward bifurcation for the AIV model,

showing that when the basic reproduction number crosses unity,

an endemic equilibrium appears and the DFE continues to exist

but loses its stability.

We continue by investigating different strategies that can reduce

or stop the spread of AIV. From the isolation-treatment model

(2), the DFE is given by

=(,,,)=

, 0,0,0.

The corresponding basic reproduction number

() with

respect to (2) is represented by

=(++)

[(++)(++)(1)].

The DFE (

) of the isolation-treatment model is locally

asymptotically stable if < 1 and unstable if > 1 .

Consequently, we can identify some conditions on how

confinement of infected birds affects the stability of

. The

DFE (

) is locally asymptotically stable whenever

(++)(+)(++)

(++)<.

Figure 5: Bifurcation diagram for the basic reproduction number

for AIV, considering no control (A), isolation-treatment (B),

vaccination (C) and culling (D). Only forward bifurcations occur.

Note the change of scale on the vertical axis in each case. The red

dotted curve illustrates the unstable branch of the bifurcation diagram.

For the endemic equilibrium of the isolation-treatment model

(2), we indicate the presence of infection in the population by

letting

0 and solve for

,

,

, and

. Thus, we have

=(

,

,

)

=(+

)

(+

)+

,

,

++,

++,

where

=(++)[(++)]+(1)

(+)[(++)(++)(1)].

Given the basic reproduction number (7) , we rewrite the

expression

of the isolation-treatment model as

=

+(1).

From (9), it follows that when 1, we have

0 and

there is no endemic equilibrium; however, when > 1, we

have

> 0 and we have an endemic equilibrium. Thus, the

isolation-treatment model (2) has no endemic equilibrium when

1 and has an endemic equilibrium when > 1. Hence

there is no backward bifurcation for the isolation-treatment

model when < 1.

In Figure 5B, we have a forward bifurcation for the isolation-

treatment model, which supports our claim. The bifurcation plot

between the infected population

and the basic reproduction

number for the isolation-treatment model shows that

reducing below unity is enough to eliminate avian influenza

from the poultry population.

Next, we analyze the stability of the associated equilibria of the

AIV model with vaccination strategy

(3). The DFE and the

basic reproduction number are

=(,,)=(+)

(+),

+, 0

and

=(+)

(+)(+).

The DFE

of vaccination model is locally asymptotically

stable if < 1 and unstable if > 1. Moreover, we obtain

some conditions for the proportion of vaccinated poultry

()

and vaccine efficacy

(), which both range from 0 to 1.

is

locally asymptotically stable whenever

+

1(+)

< 1 .

For the endemic equilibrium of the vaccination model (3), we

obtain the following:

=(

,

,

),

where

=(+

)[(1)[(+

)(+)+(1)

]+(H +

)]

[(+

)+

][(+

)(+)+(1)

]

=(+

)

(+

)(+)+(1)

(7)

(8)

(9)

(10)

Philippine Science Letters Vol. 13 | No. 02 | 2020

162

=±4

2,

such that

=(+)[(1)+(+)(+)+(1)],

=(1)+(+)(+)(1)

(+)[(1)+(+)(+)],

=(+)(+)(1).

The vaccination model

(3) has no endemic equilibrium when

1, and has a unique endemic equilibrium when > 1.

Figure 5C illustrates a bifurcation plot between the population

of infected birds and the basic reproduction number ,

showing a forward bifurcation. This bifurcation diagram is in

line with our result in Theorem B.1 in Appendix B, so there is

no endemic equilibrium when < 1 but there is a unique

endemic equilibrium when > 1. In this case, reducing

below one is sufficient to control the disease.

Finally, we analyze the stability of equilibria of the AIV model

with culling (4). The DFE for the culling model is given by

=(,)=

, 0,

and the basic reproduction number is

=

(+).

The endemic equilibria of the culling model is determined as

=(

,

)=(+

)

+(++)

,

where C

=±24

2,

such that

=(++)(++),

=(+)(1)(+)(++),

=(+)(1).

For the culling model

(4), we have shown that a backward

bifurcation does not exist when < 1. Thus, the culling model

(4) has no endemic equilibrium when < 1, and has a unique

endemic equilibrium when > 1.

In Figure 5D, we have a bifurcation diagram showing the

infected population and the basic reproduction number

().

We have a forward bifurcation in the plot, which is similar to the

result stated in Theorem B.2, implying that, when < 1, avian

influenza will be eradicated from the poultry population.

4. OPTIMAL-CONTROL STRATEGIES

We now integrate an optimal-control approach in all our models:

isolation-treatment, vaccination, and culling.

4.1. Isolation-treatment strategy

In applying the isolation-treatment strategy, we identify infected

birds and isolate them at rate . While the birds are isolated, we

apply treatment such that a proportion will successfully

recover. Our first control involves isolating infected birds with

replacing . The second control indicates the effort of the

farmers in choosing a drug that can increase the success of

treatment with replacing . The isolation-treatment model

(2) becomes

=

+,

=

++1()++(),

=()(++),

=().

We represent the rate of isolation of infected birds by control

() that is the rate () transfers from to . The

proportion of successfully treated birds released from isolation

is denoted by ().

The problem is to minimize the objective functional defined by

(,)= ()+()+

2

()+

2

(),

which is subject to the ordinary differential equations in

(12)

and where is the final time. The objective functional includes

isolation control () and treatment control (), while

and are weight constants associated with relative costs of

applying respective control strategies. The quadratic

formulation of the objective functional (,) is popular and

useful to satisfy the convexity property of the cost function

(Agusto 2013, Jung et al. 2009, Kim et al. 2018). Given that we

have two controls () and (), we want to find the optimal

controls

() and

() such that

(

,

)=min

{(,)},

where

=(,)|:0, [,],= 1,2, is Lebesgue integrable

is the control set. We consider the best- and worst-case scenarios

of isolating infected birds and giving treatment by setting the

lower bounds to = 0 and upper bounds to = 1, for = 1,2.

4.1.1. Characterization of optimal control for isolation-treatment

strategy

We generate the necessary conditions of this optimal control

using Pontryagin's Maximum Principle (Pontryagin et al. 1986).

The Hamiltonian is

=()+()+

2

()+

2

()+

+

+

++1()[++()]

+(()(++))+(()),

where ,,, are the associated adjoints for the states

,,,. We obtain the system of adjoint equations by using the

partial derivatives of the Hamiltonian (13) with respect to each

state variable.

Theorem 4.1. There exist optimal controls

()and

() and

solutions ,,, of the corresponding state system

(12)

that minimizes the objective functional (,) over .

Since these optimal solutions exist, there exists adjoint variables

,, and satisfying

=+

+

+,

=1 +

(H + I)

(H + I)+[++()]

(),

=1 [1()]+(++)(),

=,

(11)

(12)

(13)

Vol. 13 | No. 02 | 2020 Philippine Science Letters

163

with transversality conditions = 0 , for = 1, 2, 3, 4 .

Furthermore,

=min ,max ,

and

=min ,max ,

.

Proof. The existence of optimal control (

,

) is given by the

result of Fleming and Rishel (1975). Boundedness of the

solution of our system (2) shows the existence of a solution for

the system. We have nonnegative values for the controls and

state variables. In our minimizing problem, we have a convex

integrand for with respect to

(,). By definition, the

control set is closed, convex and compact, which shows the

existence of optimality solutions in our optimal system. By

Pontryagin's Maximum Principle, we obtain the adjoint

equations and transversality conditions. We differentiate the

Hamiltonian (13) with respect to the corresponding state

variables as follows:

=

,

=

,

=

,

=

,

with = 0 where = 1, 2, 3, 4 . We consider the

optimality condition

=()+= 0 and

=()+= 0,

to derive the optimal controls in (14). We consider the bounds

of the controls and obtain the characterization for optimal

controls as follows:

=min 1, max 0,

and

=min 1, max 0,

.

4.2. Vaccination

For vaccination, the first control represents the effort of the

farmers to increase vaccinated birds, while the other control

describes the efficacy of the vaccine in providing immunity

against H5N6. () and () replace and , respectively,

in the vaccination model (3) to obtain

=1()+

+,

=(t)1()

+(+),

=

++1()

+(+).

We describe the proportion of birds that are vaccinated by the

control () and the immunity of the vaccinated population

against acquiring the disease by (). We have the objective

functional

(,)= ()+

2

()+

2

(),

which is subject to

(15). This objective functional involves

increased vaccination () and the vaccine-efficacy control

(), where and are the weight constants representing

the relative cost of implementing each respective control. We

need to find the optimal controls

() and

() such that

(

,

)=min

{(,)},

where

=(,)|:0, [,],

= 3 ,4, is Lebesgue integrable

is the control set. We consider the lower bound = 0 and upper

bounds = 1, for = 3 , 4.

4.2.1. Characterization of optimal control for vaccination

strategy

In this case, the Hamiltonian is

=()+

2

()+

2

()

+1()+

+

+()(+)1()

+

+

++1()

+(+).

Theorem 4.2. There exist optimal controls

() and

() and

solutions ,, of the corresponding state system

(15) that

minimize the objective functional (,) over . Since

these are optimal solutions, there exists adjoint variables ,

and satisfying

=+

+

+,

=+++1()

+

1()

+,

=1 +

(+)+1()

(+)

(+)+1()

(+)

(+),

with transversality conditions = 0 , for = 1, 2, 3 .

Furthermore,

=min ,max ,

and

=min ,max ,

(+).

The proof is similar to the proof of Theorem 4.1 and can be

found in Appendix C.

4.3. Culling

Finally, we administer optimal control to the culling model (4).

Thus, we have

=

+()

+,

=

+()

+(+).

We represent the frequency of culling the susceptible population

by () and frequency of culling the infected population by

(). We have the objective functional

(15)

(16)

(17)

(18)

Philippine Science Letters Vol. 13 | No. 02 | 2020

164

(,)= ()+

2

()+

2

(),

which is subject to (18). The objective functional includes the

susceptible and infected culling control denoted by () and

(), respectively, with and as the weight constants

representing the relative cost of implementing each respective

control. Hence we have to find the optimal controls

and

such that

(

,

)=min

{(,)},

where

=(,)|:0, [,],

= 5 ,6, is Lebesgue integrable

is the control set. We consider the lower bound = 0 and upper

bounds = 1, for = 5, 6.

4.3.1. Characterization of optimal control for culling strategy

In this case, the Hamiltonian is

=()+

2

()+

2

()

+()

+

+

+

+(+)()

+.

Theorem 4.3. There exists optimal controls

() and

()

and solutions , of the corresponding state system (18) that

minimize the objective functional (,) over . Since

these optimal solutions, there exists adjoint variables and

satisfying

=+()

++

+

+,

=1 + ()

(+)+

(+)

(+)(+)

(2+)()

(+),

with transversality conditions , for = 1, 2. Furthermore,

=min ,max ,

() and

=min ,max ,

(+).

The proof can be found in Appendix D.

5. NUMERICAL RESULTS

The parameter values applied to generate our simulations are

listed in the table in Appendix A. The initial conditions of the

simulations are based on the Philippines' H5N6 outbreak report

(OIE 2020). We set (0)=407 837, (0)=73 360, (0)=

0, (0), and the total population of birds (0)=481 197.

Figure 6: Simulation results showing the transmission dynamics

of H5N6 in the Philippines with no intervention strategy. We use

initial conditions and parameter values as follows: (0)=407 837,

(0)=73 360, =

, = 3.4246 ×10, = 0.025 , =180 000,

= 4 × 10.

Previous studies suggested that the basic reproduction number

for the presence of avian influenza without applying any

intervention strategy was = 3 (Mills et al. 2004, Ward et al.

2009). We consider density-dependent transmission, where the

contact between birds increases as the poultry population

increases (Roche et al. 2009). We have calculated the

transmissibility of the disease (= 0.025) based on

(5) with

= 3, and fixed values of (birth rate), (natural death rate),

(disease induced death rate) and (half-saturation constant).

Without any control strategy, avian influenza will become

endemic in the poultry population as shown in Figure 6. After

50 days, the population of the infected poultry exceeds that of

susceptible poultry, with all birds eventually infected or dead.

5.1 Confinement with treatment strategy

Isolation of infected birds and application of treatment is a

potential strategy to hinder an outbreak and reduce further

spread of infection in the population. Figures 7–9 illustrate the

effects of applying optimal control to the isolation-treatment

strategy under different approaches. In Figure 7, we investigate

the effects of varying the weight constant and , which

represents the relative cost of implementing isolation and

treatment controls, respectively. Figure 8 portrays the difference

between using a constant parameter and optimal control in

describing the spread of infection using the isolation-treatment

strategy. Figure 9 shows the disparity of using both isolation and

treatment to using only one control measure.

As the relative cost of implementing isolation control and

treatment control increases, slightly lower isolation and

treatment rates are utilized, as illustrated in Figure 7. As we

increase and , the population of the susceptible birds

decreases (see Figure 7A) while the population of the infected

birds increases (as shown in Figure 7B). Isolated birds increase

significantly in the first six days, then decline afterward due to

treatment, as portrayed in Figure 7C. Increasing the cost of

treatment leads to slower increase of recovered birds (Figure

7D) and slower decline of isolated birds (Figure 7C). We can

observe that when we have lower values for and , the

susceptible population has a slower decline, there are fewer

infected and isolated birds, and recovered birds increase faster.

Thus, we consider ,=500,000 . Moreover, it can be

observed that cheaper cost controls and (Figures 7A–D)

should be administered at higher rates of and (shown in

Figures 7E–F).

(19)

(20)

Vol. 13 | No. 02 | 2020 Philippine Science Letters

165

Figure 7: Application of isolation-treatment strategy with optimal

control to the population of susceptible (A), infected (B), isolated

(C) and recovered (D) birds along with isolation control (E) and

treatment control (F) for varying values of , for =,, fro m

to birds.

With optimal control, we can possibly prevent the spread of

H5N6 in the poultry population, as demonstrated in Figure 8.

The red dashed line (without optimal control) is a simulation of

the isolation-treatment model (2) where we represent the

isolation rate and the proportion of successfully recovered birds

by a constant parameter. The blue solid line (with optimal

control) is a simulation of isolation-treatment model (12) where

control parameters () and () are included. In Figure 8A,

the susceptible population declines slower under optimal control

compared to using a constant parameter. This is due to rapid

isolation of infected birds triggering the surge in Figure 8C with

78% isolation at the beginning, as seen in Figure 8E. It also has

a faster increase in the recovered population, with 843,600 birds

compared to 73,340 birds without optimal control within 100

days, as portrayed in Figure 8D. Application of optimal controls

() and

() in the susceptible, infected, isolated and

recovered population is clearly better than using constant

parameter (Figure 8). We can observe a slower decline of

susceptible birds, an initial reduction in infected birds and a

delayed increase in infection. More infected birds are isolated

(Figure 8C), and we have a higher number of birds that will

Figure 8: Applying isolation-treatment strategy with optimal

control (blue solid line) and without optimal control or using

constant parameter (red dashed line) in the population of

susceptible (A), infected (B), isolated (C) and recovered (D) birds.

Optimal-control values for isolation control (E) and treatment

control (F) over 100 days.

recover after going through isolation (Figure 8D). Thus, using

optimal control illustrated a more appropriate representation of

implementing isolation-treatment strategy in controlling an

outbreak.

Figure 9: Isolation-treatment strategy with the optimal approach

and with consideration of using both isolation and treatment

control (blue solid line), using isolation control only (red dashed

line), and using treatment only (green dashed line) to the

population of susceptible (A), infected (B), isolated (C) and

recovered (D) birds.

It is evident that using isolation together with treatment showed

better results in all populations compared to implementing

isolation alone or treatment alone, as depicted in Figure 9. In

applying both controls, the susceptible populations decrease

slowly; infected birds are eliminated from the poultry

population; and isolated birds increase within 5 days, and then

decrease afterward, which is due to releasing of birds from

isolation. In reality, treatment can only be applied to birds that

have been identified as infected. In the isolation-treatment

Philippine Science Letters Vol. 13 | No. 02 | 2020

166

Figure 10: Application of vaccination strategy with optimal control to the population of susceptible (A), vaccinated (B) and infected (C) birds

and the increased vaccination coverage (D) and the vaccine-efficacy control (E) with varying values of , for = ,, from to

birds.

Figure 11: Applying the vaccination strategy with optimal control (blue solid line) and without optimal control or using constant parameter

(red dashed line) in the population of susceptible (A), vaccinated (B) and infected (C) birds. Optimal-control values for vaccination p revalence

control (D) and vaccine efficacy control (E) over 300 days.

model, treatment cannot be performed without isolation. Hence,

the continuous increase in the infected population if = 0 and

0 (represented by the green line in Figure 9). Isolated birds

will transfer to either the infected or recovered population,

depending on the effect of treatment. Without treatment (0

and = 0), isolated birds increase continuously then decrease

after 85 days where the birds transfer to the infected population,

as illustrated by the red line in Figures 9B–C. Applying isolation

alone will reduce the infected population and prevent possible

transmission of the disease to the susceptible population.

However, due to the absence of treatment, birds will be released

from isolation even though they are still infectious. This results

in a rapid increase of the infected population after 85 days, as

represented by the red line in Figure 9B. Our result suggests that

isolation of infected birds without applying treatment is not

sufficient to prevent the spread of H5N6 in the population.

5.2 Immunization strategy

Next, we consider immunizing the poultry population via a

vaccine. Figure 10 illustrates the outcome of varying the relative

cost of performing vaccination implementation control and

vaccine efficacy control . In Figure 11, we portray the

comparison using fixed control (red dashed line) and optimal

control (blue solid line).

In Figure 10, we observe that varying the relative costs ( and

) of implementing the controls ( and ) significantly

Vol. 13 | No. 02 | 2020 Philippine Science Letters

167

affects the spread of H5N6 in the vaccinated population. As we

increase the relative costs, the vaccine efficacy decreases

(Figure 10E), and this makes the vaccinated population

vulnerable to acquiring H5N6. As shown in Figure 10D, the

effects of varying the relative costs to vaccination control is very

close to zero (the control ranges from 0 to 0.04), and it has a

minimal effect in the spread of the virus in the population. We

can observe that the changes in the vaccine efficacy (Figure 10E)

greatly affect the curves in the vaccinated and infected

population (Figures 10B–C). As the relative cost of the vaccine

efficacy increases, the value of is lowered. Lower vaccine

efficacy leads to rapid decline in the number of vaccinated birds

and hence an increase in the infected population.

Through the application of optimal control, we can observe that

the diminishing effectiveness of the vaccine results in the spread

of infection in the vaccinated population, as depicted in Figure

11. After 120 days, the vaccine efficacy starts to decline, causing

vaccinated birds to acquire the disease. Simulations shown in

Figures 10–11 contribute to our understanding that immunizing

the poultry population is not sufficient to prevent an outbreak.

In using an optimal-control approach, we see that a successful

immunization strategy highly depends on developing an

effective vaccine. Note that, for the vaccination strategy, the

cheapest vaccination is administered at a higher rate of vaccine

efficacy control (shown in Figure 11).

5.3 Depopulation strategy

We obtain simulations for applying a modified culling strategy

that targets infected birds as well as high-risk susceptible birds

that are in contact with infected birds. Figure 12 compares the

difference in outcomes of applying optimal control versus fixed

control. Figure 13 depicts the effect of changing the relative cost

of implementing the culling strategy for susceptible and infected

populations. In Figure 14, we investigate the discrepancies in

applying the modified culling strategy for culling both

susceptible and infected birds, culling only susceptible birds and

culling only the infected birds.

Integrating optimal control into a culling strategy results in a

lower number of susceptible and infected birds compared to

using a constant value, as portrayed in Figure 12. With optimal

control, intensive culling occurred during the first 30 days of

outbreak then slowed down over time. The decline in the

numbers of both susceptible and infected birds occurs faster

when optimal control is applied. In Figures 12A–B, 88% of

susceptible birds and 63% of infected birds were culled within

30 days to prevent the spread of H5N6 avian influenza virus.

After 100 days, there are only 4% susceptible birds and 11%

infected birds left. Our optimal-control results suggest that

culling of susceptible and infected birds must be implemented

rigorously in the first 30 days of the outbreak to prevent further

spread of the infection.

Even though the relative cost of culling increases for both

susceptible and infected populations, we were able to control the

outbreak and prevent further increase in the number of infected

birds, as illustrated in Figure 13. We have lower values of culling

controls for susceptible and infected populations ( and ,

respectively) when the relative cost of implementation increases,

as depicted in Figures 13C–D. Thus, the higher cost of

implementation of culling will result a higher number of

susceptible birds but also more infected birds. Hence, varying

the relative cost and from 100,000 to 900,000 will not

affect the effectiveness of culling in preventing the spread of the

H5N6 in the poultry population.

Figure 12: Implementing the culling strategy with optimal control

(blue solid line) and without optimal control or using constant

parameter (red dashed line) in the population of susceptible (A)

and infected (B) birds. Optimal-control values of culling

frequency control for susceptible (C) and infected (D) birds over

300 days.

Figure 13: Application of culling strategy with optimal control to

the population of susceptible (A) and infected (B) birds and

susceptible-culling control (C) and infected-culling control (D),

with varying values of , for =,, from to

birds.

Administering a culling strategy for both susceptible and

infected birds is more effective than culling only the infected

birds, as indicated in Figure 14. Looking at the blue dashed line

of Figure 14A, we have more susceptible birds if we cull only

the infected population, but, as shown in Figure 14B, the

infected population increases afterward. This implies that

culling only the infected population is not enough to stop the

spread of infection. We can infer that culling only the infected

population can be successful if we can entirely eradicate the

infected population. Currently, we cannot easily identify

infected birds from the poultry population. Culling both

susceptible and infected birds may lead to near eradication of the

infected population, and due to the low number of susceptible

birds, further spread of H5N6 would not be possible. Thus,

culling both susceptible and infected birds is necessary to

eliminate the spread of infection in the poultry population.

In the 2017 Central Luzon H5N6 outbreak, it cost the country's

poultry industry 2.3 billion pesos with around 160,000 infected

poultry (Simeon, 2017). There is insufficient data for the actual

cost of implementation of each strategy per poultry. Henceforth,

Philippine Science Letters Vol. 13 | No. 02 | 2020

168

Figure 14: Simulation of culling strategy with the optimal

approach and with consideration of using both susceptible

culling control ()and infected culling control () (black solid

line), using susceptible culling control () only (red dotted-

dashed line) and using infected culling control () to the

population of susceptible (A) and infected (B) birds.

in this study, we can only present an abstract concept of the cost

(based on the number of infected birds) and compare the cost

from each strategy. Among the three strategies, we concluded

that the modified culling strategy is the cheapest with the least

number of infected birds after 100 days. For future work,

collaborations with engineers can be established to build the

actual facilities and compute the cost per unit of poultry.

Table 1: Total cost of implementation and the number of

infected birds after 100 days for each strategy.

Strategies Total Cost

Infected birds after

100 days

Isolation-

treatment

5.0x104

3.7x104 (reduced by

50%)

Vaccination 2.8x105

8.9x104 (increased

by 22%)

Modified culling 8.1x103

1.2x104 (reduced by

84%)

6. DISCUSSION

Understanding and learning to control avian influenza is a

crucial issue for many countries, especially in Asia. Avian

influenza virus A (H5N6) is an emerging infectious disease that

was reported in China in early May 2014 (Joob and Viroj 2015).

In 2017, the Philippines reported an outbreak of H5N6 which

resulted in a mass culling of 667,184 birds. After more than two

years H5N6 reemerged, causing the depopulation of 12,000

quails (OIE 2020). Lee and Lao (2017) proposed intervention

strategies against the spread H5N6 virus in the Philippines. They

suggested poultry isolation strategy over vaccination strategy in

reducing the number of infected birds.

There is limited study on the effects of isolation with treatment

as a control strategy against the spread of avian influenza.

Isolation is also used when adding new flocks of birds to the

poultry farm in order to prevent possible transmission of disease

to the current flock. We investigated the effects of isolation-

treatment strategy as a promising policy in controlling an

outbreak. We modified the isolation model of Lee and Lao

(2017) and emphasize the role of treatment in utilizing this

strategy. We focused on the impact of isolation control and

treatment control in applying this strategy. Isolating infected

birds is an effective measure to reduce the spread of H5N6 in the

population, as claimed by Lee and Lao (2017). We followed up

confinement by applying treatment during isolation, which turns

out to have a significant role in applying confinement. Through

our simulation in Figures 7–9, we showed that transmission of

H5N6 virus in the poultry population can be reduced by isolating

at least 78% of the infected birds. In addition, at least 62% of the

isolated birds must successfully recover from the infection

within the first week.

Using optimal-control theory, we showed that the success of

vaccination is highly dependent on the effectiveness of the

chosen vaccine. A less-effective vaccine will make vaccinated

birds vulnerable to acquiring the virus. Vectormune AI is a

rHVT-H5 vaccine which provides 73% protection against AIV

H5 type (Kilany et al. 2014). In the study of Cornelissen and

colleagues (2012), the NDV-H5 vaccine induced 80% immunity

to chicken against H5N1. A fowlpox vector vaccine TROVAC-

H5 protected chickens against avian influenza for at least 20

weeks (Bublot et al. 2006). Despite effective vaccines, there is a

possibility for the effectiveness of the vaccine to decline over

time, so we suggest that vaccination should be implemented

together with other intervention strategies in preventing the

spread of H5N6 in the population.

Mass culling of birds is the current policy used when detecting

an outbreak of avian influenza, which is applied to the infected

farm and a short radius around the infected premises (OIE, 2020).

We considered a modified culling strategy, as suggested in the

study of Gulbudak and Martcheva (2013), which focused on

culling infected birds as well as high-risk susceptible birds that

are in contact with infected birds. We showed that culling only

the infected birds is not enough to contain the spread of H5N6.

Instead, culling 78% of susceptible birds and at least 63% of

infected birds within 30 days can prevent an outbreak and avoid

further transmission of the virus in the poultry population.

The modified culling strategy has the cheapest implementation

cost with the least number of infected birds after 100 days. It

should be implemented if rapid eradication of the outbreak is

necessary, with the understanding that the consequence is losing

a large number of birds in the process. On the other hand, if we

aim to conserve the poultry population, then the isolation with

treatment strategy will potentially prevent the outbreak with

most of the birds recovered from the infection. This strategy can

be achieved through a rapid isolation of infected birds and a

reliable treatment policy. Conversely, vaccination should be

implemented only with other intervention strategies.

Note that we used three different models for each strategy, which

limits our comparison of the three control strategies. Future

work will consider combinations of strategies and conduct

numerical continuation studies to track both stable and unstable

steady states and bifurcation points in the systems in order to

gain better understanding and new discoveries of the overall

dynamics of the epidemiological systems.

Using optimal-control theory gives us a better understanding of

H5N6 outbreak prevention. By applying optimal control to

different strategies against H5N6, we have illustrated the effects

of each policy, together with its respective implementation cost.

Every intervention strategy against H5N6 has advantages and

disadvantages, but proper execution and appropriate application

is a significant factor in achieving a desirable outcome.

Vol. 13 | No. 02 | 2020 Philippine Science Letters

169

ACKNOWLEDGMENTS

Lucido acknowledges the support of the Department of Science

and Technology-Science Education Institute (DOST-SEI),

Philippines for the ASTHRDP Scholarship grant together with

the Career Incentive Program (CIP). Lao holds research

fellowship from De La Salle University. Smith? is supported by

an NSRC Discovery Grant. For citation purposes, please note

that the question mark in “Smith?” is part of his name. We thank

the anonymous reviewers whose comments helped improve and

clarify this manuscript.

CONFLICTS OF INTEREST

Lucido, Smith? and Lao declare that they have no conflict of

interest.

REFERENCES

Agusto FB. Optimal isolation control strategies and cost-

effectiveness analysis of a two-strain avian influenza model.

Biosys2013; 113 (3):155–164.

Bi Y, Chen Q, Wang Q, Shi W, Liu D, Gao GF. Genesis,

Evolution and Prevalence of H5N6 Avian Influenza Viruses

in China. Cell Host Microbe 2016; 20 (6):810–821.

Bublot M, Pritchard N, Swayne DE, Selleck P, Karaca K,

Suarez D, Audonnet J-C, Mickle T. Development and use of

fowlpox vectored vaccines for avian influenza.

Ann N Y

Acad Sci. 2006;1081:193-201.

Butler D. Vaccination will work better than culling, say bird

flu experts. Nature 2005; 434:810.

Capasso V, Serio G. A generalization of the Kermack-

McK

endrick deterministic epidemic model. Math

Biosci1978; 42 (1):43–61.

Chen H. Avian influenza vaccination: the experience in China.

Revue Scientifique et Technique de l’OIE2009; 28 (1):267–

274.

Chong NS, Tchuenche JM, Smith? RJ. A mathematical model

of avian influenza with half-

saturated incidence. Theory

Biosci2014; 133 (1):23–38.

Cornelissen LA, Leeuw OS, Tacken MG, Klos HC, Vries RR,

Boer-Luijtze EA, Zoelen-Bos EA, Rigter A, Rottier PJ,

Moormann RJ, Haan CA. Protective efficacy of Newcastle

disease

virus expressing soluble trimeric hemagglutinin

against highly pathogenic H5N1 influenza in chicken and

mice. PLoS ONE 2012; 7(8):1–11.

FAO. The Global Strategy for Prevention and Control of H5N1

Highly Pathogenic Avian Influenza, 2007.

Fleming W, Rishel R. Optimal deterministic and stochastic

control. Springer, 1975.

Gulbudak H, Martcheva M. Forward hysteresis and backward

bifurcation caused by culling in an avian influenza model.

MathBiosci2013; 246 (1):202–212.

Gulbudak H, Ponce J, Martcheva M. Coexistence caused by

culling in a two-strain avian influenza model. Preprint, J Biol

Dynamics 2014; 367 (1):1–22.

Gumel AB. Global dynamics of a two-strain avian influenza

model. Int J of Comp Math 2009; 86 (1):85–108.

Joob B, Viroj W. H5N6 influenza virus infection, the newest

influenza. Asian Pac J Trop Biomed 2015; 5 (6): 434–437.

Jung E, Iwami S, Takeuchi Y, Jo T-C. Optimal control strategy

for prevention of avian influenza pandemic. J Theor Biol.

2009; 260(2):220–229.

Kilany W, Dauphin G, Selim A, Tripodi A, Samy M, Sobhy H,

VonDobschuetz S, Safwat M, Saad M, Erfan A, Hassan M,

Lubroth J, Jobre Y. Protection conferred by recombinant

turkey herpesvirus avian influenza (rHVT-H5) vaccine in the

rearing period in two commercial layer chicken breeds in

Egypt. Avian Pathol. 2014; 43:6:514-523.

Kim S, de los Reyes AA, Jung E. Mathematical model and

intervention strategies for mitigating tuberculosis in the

Philippines. J Theor Biol. 2018; 443:100–112.

Lee H, Lao A. Transmission dynamics and control strategies

assessment of avian influenza A (H5N6) in the Philippines.

Infec Dis Model2018; 3:35–59.

Liu S, Ruan S, Zhang X. Nonlinear dynamics of avian

influenza epidemic models. Math Biosci 2017; 283:118–

135.

Liu Z, Fang C-T. A modeling study of human infections with

avian influenza A H7N9 virus in mainland China. Int J Infect

Dis 2015; 41:73–78.

Mills CE, Robins JM, Lipsitch M. Transmissibility of 1918

pandemic influenza. Nature 2004; 432 (7019):904–906.

OIE. World Animal Health Information System, 2020.

Okosun KO, Smith? R. Optimal control analysis of malaria–

schistosomiasis co-

infection dynamics. Math BiosciEng

2017; 14:377–405.

Pontryagin LS, Boltyanskii V, Gamkrelidze R, Mishchenko E.

Mathematical Theo

ry of Optimal Processes. CRC Press

Book, 1986.

Roche B, Lebarbenchon C, Gauthier-Clerc M, Chang C-M,

Thomas F, Renaud F, van der Werf S, Guegan J-F. Water-

borne transmission drives avian influenza dynamics in wild

birds: The case of 2005-2006 epidemics in the Camargue

Area. Infect GenetEvol2009; 9:800-805.

Philippine Science Letters Vol. 13 | No. 02 | 2020

170

Shi Z, Zhang X, Jiang D. Dynamics of an avian influenza

model with half-

saturated incidence. ApplMath Comput

2019; 355:399–416.

Simeon, L. (2017, September 08). Avian flu poultry ban

partially

lifted. Retrieved August 23, 2020, from

https://www.philstar.com/headlines/2017/08/22/1731805/av

ian-flu-poultry-ban-partially-lifted.

Teng Y, Bi D, Guo X, Hu D, Feng D, Tong Y. Contact

reductions from live poultry market closures limit the

epidemic of human infections with H7N9 influenza. J Infect

2018; 76 (3):295–304.

van den Driessche P, Watmough J, Reproduction numbers and

sub-threshold endemic equilibria for compartmental models

of disease transmission. Math Biosci2002; 180:29–48.

Ward MP, Maftei D, Apostu C, Suru A. Estimation of the basic

reproductive number (R0) for epidemic, highly pathogenic

avian influenza subtype H5N1 spread. Epi Infect 2009; 137

(2):219–226.

WHO. Influenza (Avian and other zoonotic), 2020.

Vol. 13 | No. 02 | 2020 Philippine Science Letters

171

Appendix A. Variables and parameters

Here, we describe each variable and parameter that we used in

each model.

Notation Description or Label

() Susceptible birds

() Infected birds

() Isolated birds

() Recovered birds

() Vaccinated birds

() Total bird population

Constant birth rate of birds

Natural death rate of birds

Rate at which birds contract avian

influenza

Half-saturation constant for birds

Additional disease death rate due to

avian influenza

Proportion of vaccinated poultry

Efficacy of the vaccine

Isolation rate of identified infected birds

Releasing rate of birds from isolation

Proportion of recovered birds from

isolation

Culling frequency for susceptible birds

Culling frequency for infected birds

() Culling rate of susceptible birds

() Culling rate of infected birds

The initial conditions are based on Philippine Influenza A

(H5N6) outbreak report given by the OIE (2020): (0) =

407 837 and (0) = 73 360. We calculated transmissibility of

the disease (= 0.025) using the basic reproduction number

in (5) and equating it to 3, the value of the basic reproduction

number of AIV without intervention (Mills et al. 2004, Ward et

al. 2009).We calculated parameter values that reduce the basic

reproduction number below one and control the spread of AIV

in the poultry population.

Definition Symbol Value Source

Constant birth

rate of birds

per day

(Chong et

al. 2013)

Natural

mortality rate

3.4246 ×

10

per day

(Liu et al.

2017)

Transmissibility

of the disease

0.025per

day

Calculated1

Half-saturation

constant for

birds

180 000birds (Lee and

Lao 2018)

Disease-

induced death

rate of poultry

4×10per

day

(Liu et al.

2017)

Proportion of

vaccinated

poultry

0.50 Calculated1,2

Vaccine

efficacy

0.90 Calculated1,2

Waning rate of

the vaccine

0.00001per

day

Calculated1

Isolation rate of

identified

infected birds

0.01per day Calculated1,2

Release rate of

birds from

isolation

0.09per day Calculated1

Proportion of

recovered

birds from

isolation

0.5 Calculated1,2

Culling

frequency for

susceptible

birds

per day Estimated2

Culling

frequency for

infected birds

per day Estimated2

1Calculated means we compute this value using t he basic reproduction

number

2These values will become the controls when optimal-control theory is

applied.

Philippine Science Letters Vol. 13 | No. 02 | 2020

172

Appendix B. Non-existence of backward bifurcation

Appendix B.1. Vaccination

In showing that a backward bifurcation does not exist for the

vaccination model, we solve for

=±

such that

=(+)[(1)+(+)(+)+(1)],

=(1)+(+)(+)(1)

(+)[(1)

+(+)(+)],

=(+)(+)(1).

Theorem B.1.The vaccination model

(3) has no endemic

equilibrium when 1 and has a unique endemic

equilibrium when > 1.

Proof. We obtain two possible endemic equilibria

and

for the vaccination model. From (. 1), we establish the

relationship between and such that

> 1 > 0 = 1 = 0 < 1 < 0

From (. 1), it is clear that < 0. Consider the cases when >

0, when > 0 and = 0 or 4= 0, and when

< 0, > 0, and 4> 0.

Case 1: > 0

When > 0, we have > 1. Since < 0, it follows

that

=

< 0

=4

2> 0.

When > 1 the infected population

of the

endemic equilibrium

does not exist, and we have a

unique endemic equilibrium

.

Case 2: > 0 and either = 0 or 4= 0

Given that > 0, we consider the case when = 0 and

when4= 0.

Case 2A: = 0

Since = 0, it follows that

= 0 and

> 0. Note that

=

0 leads to the DFE. Hence, if > 0 and = 0, then

> 0, and

we have a unique endemic equilibrium

.

Case 2B: 4= 0

Considering that 4= 0, it follows that

=

and

,

> 0. Thus, if > 0 and 4= 0, then we have a

unique endemic equilibrium

=

.

Case 3: < 0, > 0, and 4> 0.

From the assumption that < 0 and < 0, it follows that

=

> 0 V2

=24

2> 0

Thus, we have two endemic equilibria

and

, which

implies that a backward bifurcation may possibly occur

whenever < 0, > 0, and 4> 0.

However, given the values of and , we can show that when

< 0 , we cannot obtain > 0 , which we prove by

contradiction. Suppose that < 0. By definition of and , the

value of both parameters ranges from 0 to 1. From

(. 1), it

follows that <

, where we define =(+

)and =(+)(+).

Using (. 1) with > 0 , we get (1)+>

2++(+)(1). By simplifying, we

obtain

(+)(1)

>(+)+(+)+(1).

In both extreme values of , it follows that

0 > (+)+(+).

Since ,,0, it implies that (+)+(+)0,

and we have a contradiction. Results above suggest that two

endemic equilibria do not exist when < 1 , since the

condition < 0, > 0, and 4 > 0, cannot be satisfied.

From Cases 1 to 3, it is evident that the vaccination model has

no endemic equilibrium when < 1 and a unique endemic

equilibrium when 1.

Appendix B.2. Culling

To show that a backward bifurcation does not exist for the

culling model, we solve for

=±

such that

=(++)(++),

=(+)(1)(+)(++),

=(+)(1).

Theorem B.2. The culling model (4) has no endemic

equilibrium when < 1 and has a unique endemic

equilibrium when > 1.

Proof. We obtain two possible endemic equilibria,

and

,

for the culling model. From

(. 3), < 0, and we consider

cases where < 1, = 1, and > 1.

Case 1: < 1

When is below unity, it follows that < 0 and < 0. Given

that < 0 and < 0, we have

=+4

2< 0

=4

2< 0.

Thus, in our case when < 1 , we have no endemic

equilibrium.

Case 2: = 1

When = 1 , we have = 0 and < 0 . It follows that

4=. Since < 0, we have

=+

2= 0

=

2< 0.

Hence, when = 1, we have no endemic equilibrium.

Case 3: > 1

When is above the unity, it follows that > 0. Given that

< 0 and > 0, we have

(. 1)

(. 2)

(. 3)

Vol. 13 | No. 02 | 2020 Philippine Science Letters

173

=+4

2< 0

=4

2> 0.

Hence, when > 1, we have

> 0 and a unique endemic

equilibrium

.

Appendix C. Proof of Theorem 4.2

Proof. The existence of optimal control (

,

) is given by the

result of Fleming and Rishel (1975). Boundedness of the

solution of (3) shows the existence of a solution for the system.

We have nonnegative values for the controls and state variables.

In our minimizing problem, we have a convex integrand for

with respect to (,). By definition, the control set is closed,

convex and compact, which shows the existence of optimality

solutions in our optimal system. We use Pontryagin's Maximum

Principle to obtain the adjoint equations and transversality

conditions. We differentiate the Hamiltonian (16) with respect

to the corresponding state variables as follows:

=

,

=

,

=

,

with = 0 where = 1 ,2, 3 . Using the optimality

condition

=()+= 0 and

=()+

+

+= 0,

we derive the optimal controls (17). We consider the bounds for

the control and conclude the characterization for

and

=min 1, max 0,

and

=min 1, max 0,

(+).

Appendix D. Proof of Theorem 4.3

Proof. Analagous to the previous proof, we differentiate the

Hamiltonian (19) with respect to the corresponding state

variables as follows:

=

, and

=

,

with = 0 where = 1, 2. We consider the optimality

condition

=()

= 0 and

=()

= 0,

to derive the optimal controls (20). We consider the bounds of

the controls and get the characterization for

and

=min 1, max 0,

() and

=min 1, max 0,

(+).