Controlling The Transmission Dynamics of COVID-19
Controlling The Transmission Dynamics of COVID-19
Controlling The Transmission Dynamics of COVID-19
Abstract. The outbreak of COVID-19 caused by SARS-CoV-2 in Wuhan and other cities
in China in 2019 has become a global pandemic as declared by World Health Organization
(WHO) in the first quarter of 2020 [19]. The delay in diagnosis, limited hospital resources
and other treatment resources leads to rapid spread of COVID-19. In this article, we
consider an optimal control COVID-19 transmission model and assess the impact of some
control measures that can lead to the reduction of exposed and infectious individuals in
the population. We investigate three control strategies for this deadly infectious disease
using personal protection, treatment with early diagnosis, treatment with delay diagnosis
and spraying of virus in the environment as time-dependent control functions in our
dynamical model to curb the disease spread.
Key words: COVID-19, delay in diagnosis, dynamic model, compartmental models, op-
timal control, Hamiltonian
1 Introduction
The recent outbreak of the deadly and highly infectious COVID-19 disease caused by
SARS-CoV-2 in Wuhan and other cities in China in 2019 has become a global pandemic
as declared by World Health Organization (WHO) in the first quarter of 2020 [19]. The
most vulnerable people to develop serious complications from this dangerous disease
are the elderly with underlying medical problems. As at 29th March, 2020, the 69th
situation report by WHO indicated that the deadly COVID-19 disease had globally
infected 634, 835 people with 29, 891 deaths, see [20].
The understanding of the transmission dynamics of infectious diseases has been
well-studied and researched in mathematics and usually referred to as mathematical
epidemiology. These mathematical models have played a major role in increasing un-
derstanding of the underlying mechanisms which influence the spread of diseases and
provide guidelines as to how the spread can be controlled [2, 7, 25]. The recent outbreak
of the deadly and highly infectious COVID-19 disease has attracted the attention of
many authors who have discussed and studied the nature of the virus, its transmission
dynamics and the basic reproduction number of the disease, see eg. [3,5,9,26,27,29]. Re-
cently, Elsevier and Springer have made open access to several literature for interested
researchers [4, 14].
?
Corresponding author : stephen.moore@ucc.edu.gh
2 S. E. Moore and E. Okyere
In this section, we formulate an optimal control model for COVID-19 to derive four con-
trol measures with minimal implementation cost to eradicate the disease after a defined
period of time. Our new epidemiological time-dependent control model is an extended
and modified version of the COVID-19 transmission dynamical model introduced in [28].
Here, we note that the population is divided into susceptible (S), self-quarantine sus-
ceptible (Sq ), exposed (E), infectious with timely diagnosis (I1 ), infectious with delayed
diagnosis (I2 ), hospitalized (H), recovered (R) and the viral spread in the environment
(V ). Following the compartmental transition diagram as shown in Figure 1, the eight-
state dynamical model describing the transmission dynamics of COVID-19 is given by
dS
= −(βe E + βi1 I1 + βi2 I2 + βv V )S − qS + q1 Sq
dt
dSq
= qS − q1 Sq
dt
dE
= (βe E + βi1 I1 + βi2 I2 + βv V )S − ωE
dt
Controlling the Transmission Dynamics of COVID-19 3
Fig. 1. Compartmental diagram for the transmission dynamics of COVID-19, see [28].
dI1
= φωE − γ1 I1 − µI1 (2.1)
dt
dI2
= (1 − φ)ωE − γ2 I2 − µI2
dt
dH
= γ1 I1 + γ2 I2 − mH − µH
dt
dR
= mH
dt
dV
= f1 E + f2 I1 + f3 I2 − dv V,
dt
where βe , βi1 , βi2 and βv denote the transmission rates from the exposed, infectious with
timely diagnosis, infectious with delay diagnosis and virus in the environment to the
susceptible, respectively. Also fi , i = 1, 2, 3 is the rate of virus in the environment from
both the exposed and the infectious and removed at rate dv . For the rest of the model
parameters, we refer the interested reader to reference [28] for detailed descriptions.
Following from the system (2.1), we modified the transmission rate by reducing the
factor by (1 − u1 ), where u1 measures the effort of individuals to protect themselves
(i.e. personal protection). The control variable u2 measures the treatment rate of timely
diagnosed individuals whiles the u3 measures the treatment rate of delayed diagnosed
individuals. We assume that u2 I1 and u3 I2 individuals are removed from the timely
diagnosed class and delayed diagnosed class and added to the Hospitalized class. The
fourth control variable u4 measures the spraying of the environment to prevent viral
release. We also assume that u4 V virus are removed from the environment. We further
assume that individuals that recovers at any time t after hospitalization and treatment
are removed from the hospitalized class to the recovered class. With regards to these
4 S. E. Moore and E. Okyere
assumptions, the dynamics of system (2.1) are modified into the following system of
equations:
dS
= −(1 − u1 )(βe E + βi1 I1 + βi2 I2 + βv V )S − qS + q1 Sq
dt
dSq
= qS − q1 Sq
dt
dE
= (1 − u1 )(βe E + βi1 I1 + βi2 I2 + βv V )S − ωE
dt
dI1
= φωE − u2 I1 − µI1 (2.2)
dt
dI2
= (1 − φ)ωE − u3 I2 − µI2
dt
dH
= u2 I1 + u3 I2 − mH − µH
dt
dR
= mH
dt
dV
= f1 E + f2 I1 + f3 I2 − dv V − u4 V,
dt
where q is the rate at which susceptible (S) individuals move into self-quarantine (Sq )
and q1 is the rate at which self-quarantined individuals become susceptible again. Also,
µ is the death rate and hospitalized individuals are decreased at recovery rate m.
In this section, we will formulate an objective functional and present the existence of
optimal control by means of Pontryagin’s Maximum Principle. Given the optimal con-
trol problem (2.2), we prove the existence of control problem following [25] and then
characterizing it for optimality. The objective functional J formulates the optimization
problem of identifying the most effective strategies. The overall preselected objective
involves the minimization of the number of exposed, delayed diagnosed infectious indi-
viduals and the viral spread in the environment over a finite time interval [0, T ]. We
define the objective functional J , as follows
Z T 4
1X 2
J (u1 , u2 , u3 , u4 ) := A1 E + A2 I2 + A3 V + Ci ui (t) dt. (3.1)
0 2
i=1
We aim to minimize the cost functional (3.1) which includes the number of exposed
(E), infectious with delay diagnosis (I2 ), and the virus in the environment (V ), as well
as the social costs related to the resources needed for personal protection C1 u21 , early
detected treatment C2 u22 , delay detected treatment C3 u23 , and spraying of environment
C4 u24 . The control effort is modeled by means of a linear combination of quadratic
terms, u2i (t), i = 1, . . . , 4. The constants Aj , j = 1, . . . , 3 and Ci , i = 1, . . . , 4 represent a
measure of the relative cost of the interventions over time [0, T ].
Controlling the Transmission Dynamics of COVID-19 5
The objective of the control problem is to seek functions (u∗1 (t), u∗2 (t), u∗3 (t), u∗4 (t))
such that
J (u∗1 (t), u∗2 (t), u∗3 (t), u∗4 (t)) := min{J (u1 , u2 , u3 , u4 ), (u1 , u2 , u3 , u4 ) ∈ U}, (3.2)
subject to the COVID-19 model with controls (2.2) and appropriate initial conditions.
In the next section, we prove the existence of an optimal control for the system (2.2)
and then derive the optimality system. It is well known that Pontryagin’s maximum
principle (PMP) is required to solve this control problem and the derivation of the
necessary conditions [21, 22].
Theorem 1. Given the objective functional J (u1 , u2 , u3 , u4 ) as in (3.1), where the con-
trol set U given by (3.3) is measurable subject to (2.2) with initial conditions given at
t = 0, then there exists an optimal control u∗ = (u∗1 (t), u∗2 (t), u∗3 (t), u∗4 (t)) such that
J (u∗1 (t), u∗2 (t), u∗3 (t), u∗4 (t)) := min{J (u1 , u2 , u3 , u4 ), (u1 , u2 , u3 , u4 ) ∈ U}.
Proof. The existence of an optimal control due to the convexity of the integrand of J
with respect to the control measures ui , i = 1, . . . , 4, an a priori boundedness of the
solutions of both the state and adjoint equations and the Lipchitz property of the state
system with respect to the state variables follows from [6]. t
u
To find the optimal solution, we need the Lagrangian (L) and Hamiltonian (H) for
the optimal control problem (2.2) and (3.1). The Lagrangian of the control problem is
given by
4
1X
L := A1 E + A2 I2 + +A3 V + Ci u2i (t). (3.4)
2
i=1
Since we want the minimal value of the Lagrangian, we define the Hamiltonian function
for the system as
1
C1 u21 + C2 u22 + C3 u23 + C4 u24
H = A1 E + A2 I2 + A3 V +
2
6 S. E. Moore and E. Okyere
+ λS − (1 − u1 )(βe E + βi1 I1 + βi2 I2 + βv V )S − qS + q1 Sq
+ λSq qS − q1 Sq + λE (1 − u1 )(βe E + βi1 I1 + βi2 I2 + βv V )S − wE (3.5)
+ λI1 φwE − u2 I1 − µI1 ] + λI2 (1 − φ)wE − u3 I2 − µI2
+ λH u2 I1 + u3 I2 − mH − µH
+ λR mH + λV f1 E + f2 I1 + f3 I2 − dv V − u4 V ,
where λj , j ∈ {S, Sq , E, I1 , I2 , H, R, V } are the adjoint variables. Next, we apply the
necessary conditions to the Hamiltonian H in (3.5).
Theorem 2. Given an optimal control u∗ := (u∗1 , u∗2 , u∗3 , u∗4 ) and a solution
y ∗ = (S ∗ , Sq∗ , E ∗ , I1∗ , I2∗ , H ∗ , R∗ , V ∗ ) of the corresponding state system (2.2), there exists
adjoint variable λj , j ∈ {S, Sq , E, I1 , I2 , H, R, V } satisfying
dλS
= (λS − λE )(1 − u1 ) βe E + βi1 I1 + βi2 I2 + βv V + q(λS − λE )
dt
dλSq
= q1 (λSq − λS )
dt
dλE
= −A1 + (λS − λE )(1 − u1 )βe S + wλE − φwλI1 − (1 − φ)λI2 − τ1 λV
dt
dλI1
= (λS − λE )(1 − u1 )βi1 S + (λI1 − λH )u2 + λI1 µ − τ2 λV (3.6)
dt
dλI2
= −A2 + (λS − λE )(1 − u1 )βi2 S + (λI2 − λH )u3 + λI2 µ − τ3 λV
dt
dλH
= m(λH − λR ) + µλH
dt
dλR
=0
dt
dλV
= −A3 + (λS − λE )(1 − u1 )βv S + λV (dv + u4 )
dt
with transversality conditions
λj (T ) = 0, j ∈ {S, Sq , E, I1 , I2 , H, R, V }. (3.7)
Furthermore, the control functions u∗1 , u∗2 , u∗3 and u∗4 are given by
u∗1 = min{1, max{0, Λ1 }},
u∗2 = min{1, max{0, Λ2 }},
u∗3 = min{1, max{0, Λ3 }}, and
u∗4 = min{1, max{0, Λ4 }},
where
λE − λS (βe E + βi1 I1 + βi2 I2 + βv V )S
Λ1 = ,
C1
λI1 − λH I1 λ I2 − λ H I 2 λV V
Λ2 = , Λ3 = and Λ4 = . (3.8)
C2 C3 C4
Controlling the Transmission Dynamics of COVID-19 7
In this section, we present the numerical solutions for our optimality problem using the
fourth-order runge-kutta forward-backward sweep method. This numerical scheme is
very efficient and has been widely used by several authors in simulating their optimal
control problems [1,8,15–17]. The details of this scheme can be found in the monograph
[10]. Parameter values for our numerical illustrations are adapted from [28] where the
authors used data from the city of Wuhan in the Hubei province of China. We assume
A1 = 5, A2 = 5, A3 = 10, C1 = 10, C2 = 30, C3 = 25 and C4 = 30. Figure 2 below
represent profiles of the optimal control functions (u1 , u2 , u3 , u4 ).
Fig. 3. Solution trajectories for Exposed individuals with varying parameter φ = 0.4, φ = 0.6 and
φ = 0.8. The red line represents the controlled Exposed population whiles the blue line represents the
uncontrolled exposed population.
Fig. 4. Solution trajectories for Infectious individuals with delayed diagnosis with varying parameter
φ = 0.4, φ = 0.6 and φ = 0.8. The red line represents the controlled delayed diagnosed infectious
population whiles the blue line represents the uncontrolled infectious population.
Fig. 5. Controlling the viral spread in the environment with varying proportion of sympomatic indi-
viduals φ = 0.4, φ = 0.6 and φ = 0.8 where the red line represents the controlled envrionment and the
blue line represents the uncontrolled environment.
Controlling the Transmission Dynamics of COVID-19 9
Fig. 6. Solution trajectories for Exposed individuals with varying parameter φ = 0.4, φ = 0.6 and
φ = 0.8. The red line represents the controlled exposed population whiles the blue line represents the
uncontrolled exposed population.
Fig. 7. Solution trajectories for Infectious individuals with delayed diagnosis with varying parameter
φ = 0.4, φ = 0.6 and φ = 0.8. The red line represents the controlled delayed diagnosed infectious
population whiles the blue line represents the uncontrolled infectious population.
10 S. E. Moore and E. Okyere
Fig. 8. Controlling the viral spread in the environment with varying proportion of symptomatic indi-
viduals φ = 0.4, φ = 0.6 and φ = 0.8 where the red line represents the controlled environment and the
blue line represents the uncontrolled environment.
Fig. 9. Solution trajectories for Exposed individuals with varying parameter φ = 0.4, φ = 0.6 and
φ = 0.8. The red line represents the controlled Exposed population whiles the blue line represents the
uncontrolled exposed population.
In this subsection, solution trajectories for the number of exposed, infectious with delay
diagnosis and virus in the environment for all the three control strategies are numerically
compared with that of the non-optimal control model. Our numerical results suggest
that, if people can adhere to effective personal protection practices such as the use
of hand sanitizers, washing of hands regularly and social distancing, there will less
infections in the population. From our results, we can further argue that, effective
spraying of the environment and early diagnosis of infected or infectious individuals
and treatment can help reduce of the number of COVID-19 infections significantly.
Controlling the Transmission Dynamics of COVID-19 11
Fig. 10. Solution trajectories for Infectious individuals with delayed diagnosis with varying parameter
φ = 0.4, φ = 0.6 and φ = 0.8. The red line represents the controlled delayed diagnosed infectious
population whiles the blue line represents the uncontrolled infectious population.
Fig. 11. Controlling the viral spread in the environment with varying proportion of symptomatic indi-
viduals φ = 0.4, φ = 0.6 and φ = 0.8 where the red line represents the controlled environment and the
blue line represents the uncontrolled environment.
Fig. 12. Solutions trajectories for Exposed individuals with φ = 0.4, φ = 0.6 and φ = 0.8 .
Fig. 13. Solutions trajectories for Infectious individuals delayed diagnosis with φ = 0.4, φ = 0.6 and
φ = 0.8 .
12 S. E. Moore and E. Okyere
Fig. 14. Solutions trajectories for virus in the environment with φ = 0.4, φ = 0.6 and φ = 0.8 .
5 Conclusion
References
9. W. Jia, K. Han, Y. Song, W. Cao, S. Wang, S. Yang, J. Wang, F. Kou, P. Tai, J. Li, et al. Extended
sir prediction of the epidemics trend of covid-19 in italy and compared with hunan, china. medRxiv,
2020.
10. S. Lenhart and J. T. Workman. Optimal control applied to biological models. CRC press, 2007.
11. Q. Lin, S. Zhao, D. Gao, Y. Lou, S. Yang, S. S. Musa, M. H. Wang, Y. Cai, W. Wang, L. Yang,
et al. A conceptual model for the coronavirus disease 2019 (covid-19) outbreak in wuhan, china
with individual reaction and governmental action. International journal of infectious diseases, 2020.
12. K. Mizumoto and G. Chowell. Estimating risk for death from 2019 novel coronavirus disease, china,
january-february 2020. Emerging Infectious Diseases, 2020.
13. A. A. Momoh and A. Fügenschuh. Optimal control of intervention strategies and cost effectiveness
analysis for a zika virus model. Operations Research for Health Care, 18:99–111, 2018.
14. Springer Nature. Sars-cov-2 and covid-19: A new virus and associated respiratory disease, 2020.
15. R. L. M. Neilan, E. Schaefer, H. Gaff, K. R. Fister, and S. Lenhart. Modeling optimal intervention
strategies for cholera. Bulletin of mathematical biology, 72(8):2004–2018, 2010.
16. K. O. Okosun and O. D. Makinde. A co-infection model of malaria and cholera diseases with
optimal control. Mathematical Biosciences, 258:19–32, 2014.
17. E. Okyere, J. D. Ankamah, A. K. Hunkpe, and D. Mensah. Deterministic epidemic models for ebola
infection with time-dependent controls. arXiv preprint arXiv:1908.07974, 2019.
18. S. Olaniyi, K. O. Okosun, S. O. Adesanya, and E. A. Areo. Global stability and optimal control
analysis of malaria dynamics in the presence of human travelers. The Open Infectious Diseases
Journal, 10(1), 2018.
19. World Health Organization. WHO characterizes covid-19 as a pandemic, 2020.
https://www.who.int/dg/speeches/detail/who-director-general-s-opening-remarks-at-the-media-
briefing-on-covid-19—11-march-2020.
20. World Health Organization. WHO coronavirus disease (covid-2019) situation reports, 2020.
https://www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports.
21. L. S. Pontryagin. The mathematical theory of optimal processes and differential games. 169:119–
158, 1985.
22. L. S. Pontryagin, V. G. Boltyanskii, R. V. Gamkrelidze, and E. F. Mishchenko. The mathematical
theory of optimal processes. Wiley, New York, NY, 1962.
23. K. Roosa, Y. Lee, R. Luo, A. Kirpich, R. Rothenberg, J. M. Hyman, P. Yan, and G. Chowell.
Real-time forecasts of the covid-19 epidemic in china from february 5th to february 24th, 2020.
Infectious Disease Modelling, 5:256–263, 2020.
24. O. Sharomi and T. Malik. Optimal control in epidemiology. Annals of Operations Research, 251(1-
2):55–71, 2017.
25. C. I. Siettos and L. Russo. Mathematical modeling of infectious disease dynamics. Virulence,
4(4):295–306, 2013.
26. B. Tang, N. L. Bragazzi, Q. Li, S. Tang, Y. Xiao, and J. Wu. An updated estimation of the risk of
transmission of the novel coronavirus (2019-ncov). Infectious Disease Modelling, 5:248–255, 2020.
27. H. Wang, Z. Wang, Y. Dong, R. Chang, C. Xu, X. Yu, S. Zhang, L. Tsamlag, M. Shang, J. Huang,
et al. Phase-adjusted estimation of the number of coronavirus disease 2019 cases in wuhan, china.
Cell Discovery, 6(1):1–8, 2020.
28. R. Xinmiao, Liu Y., Huidi C., and Meng F. Effect of delay in diagnosis on transmission of covid-19.
Mathematical Biosciences and Engineering, 17(mbe-17-03-149):2725, 2020.
29. S. Zhang, M. Y. Diao, W. Yu, L. Pei, Z. Lin, and D. Chen. Estimation of the reproductive number
of novel coronavirus (covid-19) and the probable outbreak size on the diamond princess cruise ship:
A data-driven analysis. International Journal of Infectious Diseases, 93:201–204, 2020.