Since January 2020 Elsevier has created a COVID-19 resource centre with

free information in English and Mandarin on the novel coronavirus COVID-

19. The COVID-19 resource centre is hosted on Elsevier Connect, the
company's public news and information website.

Elsevier hereby grants permission to make all its COVID-19-related

research that is available on the COVID-19 resource centre - including this
research content - immediately available in PubMed Central and other
publicly funded repositories, such as the WHO COVID database with rights
for unrestricted research re-use and analyses in any form or by any means
with acknowledgement of the original source. These permissions are
granted for free by Elsevier for as long as the COVID-19 resource centre
remains active.
 Physica D 411 (2020) 132599

Contents lists available at ScienceDirect

Physica D
journal homepage: www.elsevier.com/locate/physd

COVID-19: Development of a robust mathematical model and

simulation package with consideration for ageing population and time
delay for control action and resusceptibility
∗
Kok Yew Ng a,c , , Meei Mei Gui b
a
Nanotechnology and Integrated BioEngineering Centre (NIBEC), Ulster University, Jordanstown Campus, Shore Road, Newtownabbey BT37 0QB, UK
b
School of Chemistry and Chemical Engineering, David Keir Building, Queen’s University Belfast, Stranmillis Road, Belfast BT9 5AG, UK
c
Electrical and Computer Systems Engineering, School of Engineering, Monash University, Malaysia

article info a b s t r a c t

Article history: The current global health emergency triggered by the pandemic COVID-19 is one of the greatest
Received 21 April 2020 challenges we face in this generation. Computational simulations have played an important role to
Received in revised form 27 May 2020 predict the development of the current pandemic. Such simulations enable early indications on the
Accepted 27 May 2020
future projections of the pandemic and is useful to estimate the efficiency of control action in the
Available online 9 June 2020
battle against the SARS-CoV-2 virus. The SEIR model is a well-known method used in computational
Communicated by V.M. Perez-Garcia
simulations of infectious viral diseases and it has been widely used to model other epidemics such as
Keywords: Ebola, SARS, MERS, and influenza A. This paper presents a modified SEIRS model with additional exit
COVID-19 conditions in the form of death rates and resusceptibility, where we can tune the exit conditions in the
Coronavirus model to extend prediction on the current projections of the pandemic into three possible outcomes;
SEIRS death, recovery, and recovery with a possibility of resusceptibility. The model also considers specific
Mathematical modelling and simulation information such as ageing factor of the population, time delay on the development of the pandemic
Time delay
due to control action measures, as well as resusceptibility with temporal immune response. Owing
Resusceptibility
to huge variations in clinical symptoms exhibited by COVID-19, the proposed model aims to reflect
better on the current scenario and case data reported, such that the spread of the disease and the
efficiency of the control action taken can be better understood. The model is verified using two case
studies based on the real-world data in South Korea and Northern Ireland.
© 2020 Elsevier B.V. All rights reserved.

1. Introduction distress syndrome. Based on clinical data reported from Wuhan

where the outbreak began, elderly patients have been identified
The coronavirus disease COVID-19 is a respiratory infection to have higher odds to experience severe symptoms with higher
disease caused by the novel coronavirus, SARS-CoV-2. The first mortality rate compared to people of younger age [2]. Study
COVID-19 outbreak was reported in Wuhan of Hubei Province, also shows that up to approximately 80% of the people infected
China at the end of December 2019. Within just two months, with SARS-CoV-2 are asymptomatic carriers, i.e. they experience
the disease has rapidly spread across the world and it has been no or mild symptoms but are still able to transport the virus
declared a global pandemic in early March 2020. As of 20th to others [3]. This has caused the detection and containment
April 2020, the virus has affected close to 2.5 million people of the SARS-CoV-2 virus to be much more challenging. As a
with approximately 170,000 confirmed deaths across at least 184 result, social distancing has been widely implemented in many
countries [1]. countries worldwide to slow down the transportation of the
The symptoms caused by the SARS-CoV-2 virus have large virus through minimised human-to-human contact. Individuals
variations with most people only experiencing mild to moderate
who have recovered from COVID-19 after experiencing mild or
respiratory illnesses and only a smaller group of people would
moderate symptoms are more likely to develop temporary re-
develop complications of respiratory failure or acute respiratory
sistance towards the virus and are unlikely to experience severe
respiratory illnesses [4]. However, in rare occasions, there have
∗ Corresponding author at: Nanotechnology and Integrated BioEngineer-
been clinical findings showing that patients who have recovered
ing Centre (NIBEC), Ulster University, Jordanstown Campus, Shore Road,
Newtownabbey BT37 0QB, UK.
from the disease have been tested positive again. For instance, in
E-mail addresses: [email protected] (K.Y. Ng), [email protected] February 2020, a patient in Osaka, Japan, has been tested positive
(M.M. Gui). towards the SARS-CoV-2 a few days after being discharged from

https://doi.org/10.1016/j.physd.2020.132599
0167-2789/© 2020 Elsevier B.V. All rights reserved.
2 K.Y. Ng and M.M. Gui / Physica D 411 (2020) 132599

the hospital for treatment with the disease [5]. Due to very dE(t)
= β (1 − σ )S(t)I(t) − (µ + α )E(t), (2)
limited knowledge on the immune response of humans to this dt
novel virus, the possibility of reinfection cannot be ruled out at dI(t)
= α E(t) − (µ + γ )I(t) − ∆I(t) (3)
the moment. dt
Mathematical modelling and computational simulations have dR(t)
played important roles in describing the dynamics of infectious
= γ I(t) − µR(t) − Rs (t), (4)
dt
diseases using nonlinear systems so that their risks could be dD(t)
better understood [6–11]. Most notably, the SEIR (Susceptible– = ∆I(t), (5)
dt
Exposed–Infected–Removed/Recovered) model has been widely
where S(t), E(t), I(t), R(t), and D(t) represent the susceptible, ex-
reported during the past decades in quantitative modelling stud-
posed, infected, recovered/removed, and deaths compartments,
ies of infectious epidemics/pandemics, such as the severe acute
respectively. It is established that S(t) + E(t) + I(t) + R(t) + D(t) =
respiratory syndrome (SARS) in 2002 [12], the influenza A (pH1N1)
N(t), where N(t) is the total stock population. The constant Λ is
pandemic in 2009 [13], the Middle East respiratory syndrome
the birth rate in the overall population and µ is the death rate due
(MERS) pandemic in 2013 [14], as well as the latest Ebola out-
to conditions other than COVID-19. The parameter β is the rate
break in 2018 [15]. The SEIR model represents a typical infectious
of transmission per S-I contact, α is the rate of which an exposed
epidemic disease using four distinctive phases; susceptible (S)
person becomes infected, and γ is the recovery rate. Therefore,
represents the population that has yet to be infected by the virus;
the incubation and recovery times are τinc = 1/α and τrec = 1/γ ,
exposed (E) represents the number of individuals exposed to the
respectively. The constant σ is the efficiency of the control action
virus; infected (I) models the number of people infected, have
to reduce the infection rate and to flatten the curve. It has a
demonstrated symptoms, and are able to spread the virus to
direct effect on the basic reproduction number R0 , which will be
the people in the S compartment; and lastly, recovered/removed
explained further later in this paper.
(R) models the number of people who have recovered and are
The parameter ∆ = δ [(1 − κold )Nold + (1 − κ )(1 − Nold )] comes
assumed to have immune response towards the virus [16–19].
into effect in the worst case scenario where the patient does not
Thus, based on the model, the S compartment will slowly deplete
recover from the virus. We model the fatality rate with influence
as the outbreak prolongs further, and the virus will eventually
of the fraction of elderly population (above 65 years of age)
die out due to insufficient population within the S compartment.
within the community Nold , where the percentages of nonelderly
This compartmental modelling method allows transport of pop-
and elderly who recovered are κ and κold , respectively. The time
ulation from one compartment to another, where the disease
spent hospitalised or infected in fatal cases is τhosp = 1/δ . In this
transmission rates with respect to time can be simulated.
paper, we establish that τhosp = τrec , assuming that patients spend
In this work, we propose a modified SEIRS model based on the
the same amount of time hospitalised or infected, whether they
Kermack–McKendrick model [20] with consideration for time de-
recover from the virus or not.
lay and resusceptibility to the virus after recovery. In this model,
The function Rs (t) represents the resusceptible stock, which
the probability of a recovered patient to be reinfected with SARS-
can be computed from the recovered population using
CoV-2 is taken into consideration to predict the future projection
of COVID-19 cases. Resusceptibility is one of the crucial keys that Rs (t) = ξ R(t), (6)
could possibility lead to a prolonged COVID-19 pandemic. Time
delay in the control action representing the time taken for the where ξ is the percentage of the recovered population who
authorities to act on the virus and also the duration of short- are resusceptible to the virus. The number of actual recovered/
term immunity after recovery, which may lead to resusceptibility, removed cases with permanent immunity can then be written
are also considered in the model. The time delay factor is ap- using
plied to enhance the robustness and accuracy of the model and Rc (t) = R(t) − Rs (t). (7)
simulation, hence to better reflect on the timely situation with
specific measures to control the transmission of the disease. The In an ideal situation, population who recovered develop perma-
consideration for resusceptibility with time delay is an important nent immunity against the virus, i.e. ξ = 0. As a result, (6)
highlight in this paper as it has rarely been considered in the becomes Rs (t) = 0 and subsequently, (7) becomes Rc (t) = R(t).
SEIR models reported thus far [18,21,22]. Other than that, we also
included information such as demographic details for the ageing 3. Positivity, boundedness, and equilibrium analysis of the
population, who seem to experience a higher fatality rate due to model
COVID-19 [23].
This paper is organised as follows: Section 2 introduces the 3.1. Positivity of the solutions
mathematical model; Section 3 presents the theoretical proofs
for the positivity, boundedness, and stability of the model, as Lemma 1. The solutions to all subpopulations (S(t), E(t), I(t), R(t),
well as describes the model with time delay factors for the D(t)) in the system (1)–(5) are nonnegative for all time t ≥ 0
control action and potential resusceptibility; Section 4 verifies the given any finite nonnegative initial conditions of (S(0) ≥ 0, E(0) ≥
proposed model and also provides some extensive results and 0, I(0) ≥ 0, R(0) ≥ 0, D(0) ≥ 0).
discussions on predictions using the model through case studies
based on real-world data; and Section 5 concludes the paper. Proof. Firstly, it is established that all subpopulations (S(t), E(t),
Appendix A presents the design of the simulation package using I(t), R(t), D(t)) defined by the system (1)–(5) are continuously
the MATLAB/Simulink environment. differentiable. As such, if all subpopulations have nonnegative
initial conditions, and that if any of the subpopulations is zero
2. Modelling COVID-19 using modified SEIRS at time t = ti ≥ 0, its derivative is nonnegative by inspection.
Assume that S(0) ≥ 0, S(t1 ) = 0, and Rs (t1 ) ≥ 0 at time instant
First, let us consider the modified SEIRS model system below, t = t1 . Then, we can rewrite (1) using
dS(t) dS(t1 )
= Λ − µS(t) − β (1 − σ )S(t)I(t) + Rs (t), (1) = Λ + Rs (t1 ) ≥ 0,
dt dt
 K.Y. Ng and M.M. Gui / Physica D 411 (2020) 132599 3

where we can establish that S(t1+ ) ≥ 0 and hence, S(t) is non- 3.3. Disease-free equilibrium
negative for all time t ≥ 0. Next, assume that E(0) ≥ 0, E(t2 ) =
0, S(t2 ) ≥ 0, and I(t2 ) ≥ 0 at time instant t = t2 . We can rewrite Lemma 3. The disease-free equilibrium EDFE is locally asymptoti-
(2) using cally stable if the basic reproduction number R0 < 1.

dE(t2 ) Proof. The disease-free equilibrium can be obtained by equating

= β (1 − σ )S(t2 )I(t2 ) ≥ 0,
dt equations (1)–(4) to zero, hence satisfying
so that E(t2+ ) ≥ 0 and hence, E(t) is nonnegative for all time t ≥ 0. Λ − µS(t) − β (1 − σ )S(t)I(t) + Rs (t) = 0, (11)
Assume that I(0) ≥ 0, I(t3 ) = 0, and E(t3 ) ≥ 0 at time instant
β (1 − σ )S(t)I(t) − (µ + α )E(t) = 0, (12)
t = t3 . Eq. (3) then becomes
α E(t) − (µ + γ )I(t) − ∆I(t) = 0, (13)
dI(t3 )
= α E(t3 ) ≥ 0, γ I(t) − µR(t) − Rs (t) = 0, (14)
dt
+ of which the disease-free equilibrium is given by EDFE = (Λ/µ,
where we can establish that I(t3 ) ≥ 0 and hence, I(t) is nonneg-
0, 0, 0). Eq. (5) can be removed from this analysis without loss of
ative for all time t ≥ 0. Assume that R(0) ≥ 0, R(t4 ) = 0, and
generality as other equations do not depend on it. It can then be
I(t4 ) ≥ 0 at time instant t = t4 . Eq. (4) can be rewritten using shown that the Jacobian for (1)–(4) at EDFE is written using
dR(t4 )
− β Λ(1µ−σ ) ξ
⎡ ⎤
= γ I(t4 ) ≥ 0, −µ 0
dt β Λ(1−σ )
JDFE = ⎢
⎢ 0 −(µ + α ) µ
0 ⎥ . (15)
⎥
+
so that R(t4 ) ≥ 0 and hence, R(t) is nonnegative for all time t ≥ 0. ⎣ 0 α − (µ + γ + ∆ ) 0 ⎦
Finally, assume that D(0) ≥ 0, D(t5 ) = 0, and I(t5 ) ≥ 0 at time 0 0 γ − (µ + ξ )
instant t = t5 . We can then rewrite (5) using
The characteristic equation can subsequently be obtained by
dD(t5 ) subtracting λ from the diagonal elements and then computing the
= ∆I(t5 ) ≥ 0, determinant, which yields
dt
where we can establish that D(t5+ ) ≥ 0 and hence, D(t) is (−µ − λ)(−(µ + ξ ) − λ)f1 (λ) = 0, (16)
nonnegative for all time t ≥ 0. where
It can be seen that since none of the subpopulations would αβ Λ(1 − σ )
have a negative derivative at any time instant of t = ti when all f1 (λ) = (−(µ + γ + ∆) − λ)(−(µ + α ) − λ) − . (17)
µ
other subpopulations are nonnegative, then it can be concluded
The first two eigenvalues in (16) can be easily computed to be
that all subpopulations are nonnegative for all time t ≥ 0. As a
λ1 = −µ, λ2 = −µ − ξ , and that they are negative. As for the
result, given that N(t) = S(t) + E(t) + I(t) + R(t) + D(t), the stock
remaining eigenvalues, they can be found by solving the quadratic
population N(t) is also nonnegative for all time t ≥ 0. Hence, the f1 (λ) in (17), which can be expended and represented using
proof is complete. □
a1 λ2 + a2 λ + a3 = 0,
where
3.2. Boundedness of the solutions
a1 = 1,
Lemma 2. The stock population N(t) is finitely upperbounded for a2 = 2µ + γ + α + ∆,
any nonnegative initial conditions. αβ Λ(1 − σ )
a3 = (µ + γ + ∆)(µ + α ) − .
µ
Proof. The dynamics of the stock population can be written using
For the disease-free equilibrium to be stable, i.e. all eigenval-
dN(t) dS(t) dE(t) dI(t) dR(t) dD(t) ues are negative, it is required that
= + + + + ,
dt dt dt dt dt dt αβ Λ(1 − σ )
= Λ − µ(S(t) + E(t) + I(t) + R(t)), − (µ + γ + ∆)(µ + α ) < 0,
µ
= Λ − µ(N(t) − D(t)). (8) αβ Λ(1 − σ )
< 1.
µ(µ + γ + ∆)(µ + α )
Assuming that N(t) ≫ D(t), and since Lemma 1 has estab-
lished that all subpopulations are nonnegative and given that all As such, the basic reproduction number with the control ac-
tion is defined using
parameters are assumed to be positive, then (8) becomes
αβ Λ(1 − σ )
dN(t) R0 = , (18)
≈ Λ − µN(t). (9) µ(µ + α )(µ + γ + ∆)
dt
where for a disease-free system that is locally asymptotically
We can then deduce that stable, we need to ensure that R0 < 1 while an unstable EDFE
dN(t) would translate to R0 > 1. Hence, the proof is complete. □
≤ Λ − µN(t), (10)
dt Should there be no control action taken, i.e. σ = 0, then the
where an integration of the inequality (10) yields basic reproduction number in (18) becomes
αβ Λ
Λ( Λ
( )
R0 = ,
N(t) ≤ N(0)e−µt + 1 − e−µt ≤ max N(0), ,
)
µ(µ + α )(µ + γ + ∆)
µ µ
which is similar to other models found in the literature [6,9,10].
for all time t ≥ 0. As a result, the stock population is finitely See [24] and the references within for a brief study on using
upperbounded and hence, the proof is complete. □ control theory to fight COVID-19.
 4 K.Y. Ng and M.M. Gui / Physica D 411 (2020) 132599

3.4. Endemic equilibrium factors are also considered for the control action taken as well
as resusceptibility. As a result, the system (1)–(5) becomes
Lemma 4. The endemic equilibrium EEE is locally asymptotically dS(t)
stable if the basic reproduction number R0 > 1. = −β (I(t)S(t) − σ I(t − τσ )S(t − τσ )) + Rs (t , τξ ), (26)
dt
dE(t)
Proof. Let us assume that the system (1)–(4), which other than = β (I(t)S(t) − σ I(t − τσ )S(t − τσ )) − α E(t), (27)
the disease-free equilibrium has a unique equilibrium at EEE = dt
(S ∗ , E ∗ , I ∗ , R∗ ) such that dI(t)
= α E(t) − γ I(t) − ∆I(t), (28)
dt
Λ
S = ∗
, (19) dR(t)
µR0 = γ I(t) − Rs (t , τξ ), (29)
dt
µΛ(µ + γ + ∆)(µ + ξ )(R0 − 1) dD(t)
E∗ = , (20) = ∆I(t). (30)
αβ Λ(1 − σ )(µ + ξ ) − αµγ ξ R0 dt
µΛ(µ + ξ )(R0 − 1)
I∗ = , (21) and hence, the basic reproduction number in (18) becomes
β Λ(1 − σ )(µ + ξ ) − µγ ξ R0
β (1 − σ )
µγ Λ(R0 − 1) R0 = . (31)
R∗ = . (22) γ +∆
β Λ(1 − σ )(µ + ξ ) − µγ ξ R0
The time delay τσ = τpre−σ + τpost −σ indicates the time taken
It can be seen that the model has no positive endemic equilibrium
for the control action to take effect in flattening the infection
if R0 < 1 since E ∗ , I ∗ , and R∗ would be negative, which indicate
curve, where τpre−σ ≥ 0 represents the time to initiate the control
an unrealistic biological system. If R0 = 1, then we would have
action after the first confirmed case at t = 0, and τpost −σ ≥ 0
the disease-free equilibrium EDFE discussed earlier in Section 3.3.
represents the time after the control action has been initiated
Hence, for a positive endemic equilibrium system, we would
but before the effects are evidenced in the outputs of the system.
require that R0 > 1.
In practical scenarios, τpost −σ can be used to model the delay for
For the stability analysis of the endemic equilibrium, we use
the population to effectively respond to the rules introduced by
the Jacobian for (1)–(4) at EEE , which is written using
the control action, such as social distancing, self-isolation, and
− (µ + η I ∗ ) −ηS ∗ ξ lockdown.
⎡ ⎤
0
ηI ∗ −(µ + α ) ηS ∗ 0 The function Rs (t , τξ ) represents the resusceptible stock with
JEE =⎣ ⎦,
⎢ ⎥
0 α −(µ + γ + ∆) 0 the consideration for temporal immunity, of which we can then
0 0 γ − (µ + ξ ) rewrite (6) using

(23) Rs (t , τξ ) = ξ R(t − τξ ), (32)

where η = β (1 −σ ). The characteristic equation can subsequently where the time delay τξ ≥ 0 represents the duration of temporal
be obtained by subtracting λ from the diagonal elements and then immune response of the recovered population. Hence, we can
computing the determinant, which yields also update the expression for the number of recovered cases
introduced in (7) using
(−(µ + ξ ) − λ)(−λ3 + b1 λ2 − b2 λ + b3 ) − αγ ηξ I ∗ = 0, (24)
Rc (t) = R(t) − Rs (t , τξ ). (33)
where
In an ideal situation where population who recovered develop
b1 = −(µ + ηI ∗ ) − (µ + α ) − (µ + γ + ∆),
permanent immunity against the virus, ξ = 0 and τξ → ∞. As a
= −3µ − α − γ − ∆ − ηI , ∗
result, (32) becomes Rs (t , ∞) = 0 and subsequently, (33) can be
b2 = (µ + ηI ∗ )(µ + α ) + (µ + ηI ∗ )(µ + γ + ∆) rewritten as Rc (t) = R(t).
+ (µ + α )(µ + γ + ∆) − αηS ∗ , The block diagram of the proposed SEIRS model with time
delay is shown in Fig. 1. The system with time delay is assumed
= 3µ2 + 2µα + 2µγ + 2µ∆ + αγ + γ ∆ to be stable and will exhibit similar disease-free equilibrium
+ ηI ∗ (2µ + α + γ + ∆) − αηS ∗ , and endemic equilibrium properties as the system without time
delay provided that the time delay parameters are nonnegative,
b3 = −(µ + ηI ∗ )((µ + α )(µ + γ + ∆) − αηS ∗ ) − αη2 S ∗ I ∗ ,
i.e. τσ ≥ 0, τξ ≥ 0. Detailed discussions on the theoretical
= −(µ + ηI ∗ )(µ2 + µ(α + γ + ∆) + α (γ + ∆)) + µαηS ∗ . stability analysis of SEIR and similar epidemic models with time
delay can be found in studies such as [19,25–27].
Assuming that αγ ηξ I ∗ ≪ (−(µ + ξ ) − λ)(−λ3 + b1 λ2 − b2 λ +
b3 ), then the characteristic equation in (24) becomes
4. Case Studies
(−(µ + ξ ) − λ)(−λ3 + b1 λ2 − b2 λ + b3 ) ≈ 0. (25)
4.1. Case Study 1: Verification using data in South Korea
It is obvious that the first eigenvalue is λ1 = −µ−ξ . It can also be
shown that in order for the remaining eigenvalues to be negative
South Korea is used as a case study due to the amount of
such that the endemic equilibrium is locally asymptotically stable,
data available given that it is one of the first few countries to be
i.e. b1 < 0, b2 > 0, and b3 < 0, we would require that R0 > 1
directly affected by the virus outside of China. Its first confirmed
while an unstable EEE would translate to R0 < 1. Hence, the proof
case was reported on 20th January 2020 [28]. The other reason
is complete. □
is that South Korea is also one of the very few countries that
For the remaining of this paper where we verify the model managed to effectively flatten the infection curve and it has set
and perform predictions using real-world data in Section 4, and itself apart from others in leading the fight against COVID-19,
also for the model used in the simulation package presented in at least for the moment. For example, the country started vigor-
Appendix A, we assume a closed population with negligible birth ous testing among its population with contact tracing, especially
and death rates, i.e. Λ/µ ≈ 1, Λ ≈ 0, µ ≈ 0. Time delay those of confirmed and suspected cases during the early stage of
 K.Y. Ng and M.M. Gui / Physica D 411 (2020) 132599 5

Table 1
Initial parameters used to fit the model to the data in South Korea.
Parameter Value
Stock population, N 51.5×106
Fraction of elderly population, Nold 0.15
Percentage of recovery, κ 0.98
Fatality rate for elderly, 1 − κold 0.08
Incubation time, τinc 5.1 days
Recovery time, τrec 18.8 days
Basic reproduction number, R0 5.1 (95% CI: 4.97–5.23)
Initial infected cases, I(0) 4
Initial exposed cases, E(0) 80

Fig. 1. The block diagram of the proposed SEIRS system used in the simulation
package.
acceptable while modelling an actual epidemic as most countries
are still coming to terms with the virus during the first month
the epidemic. The government accomplished this by maintaining and the data do not usually represent the actual number of cases
a public database keeping track of mobile phones, credit cards, due to lack of testing for confirmed cases.
and other data of patients who tested positive [29]. Also, on Once we have the initial fitting of the model, we introduced
16th March 2020, the authorities in the country began screening control action in line with the mitigation and preventive mea-
every person, both domestic and international, who arrived at its sures taken by the government. Due to the aforementioned vig-
airports. orous testing, contact tracing, and isolation efforts taken, we
As of 20th April 2020, there have been 10,674 confirmed cases assumed that the control action has an efficiency of 88% (σ =
and 236 fatalities in South Korea [1]. As a result, we used the 0.88). As a result, the reproduction number could be reduced to
following parameters for our simulation. First, we assumed that R0 ≈ 0.61. We also assumed that there was a time delay of 30
the population of South Korea to be approximately N = 51.5×106 days since the first confirmed case before the control action was
with an elderly population of about 15% (Nold = 0.15) [30]. We introduced (τpre−σ = 30) and a further delay of approximately 13
then set the percentage of recovery to be 98% (κ = 0.98) for the days before the control action could be properly executed in the
general public [1] and a fatality rate of 8% ((1 − κold ) = 0.08) community (τpost −σ = 13). Fig. 3 shows the simulation results;
for the elderly [31]. We then assumed the incubation time and Fig. 3a shows that the trajectory of the modelled infected and
recovery time to be τinc = 5.1, and τrec = 18.8, respectively death cases matches the real-world data after the control action
in accordance with [32]. The basic reproduction number was set was introduced. Fig. 3b shows the simulation results until the
to R0 = 5.1 (95% CI: 4.97–5.23) based on the early growth- model stabilises assuming no subsequent control action being
rate of the epidemic in South Korea. The initial infected and taken to further reduce the reproduction number. Comparing
exposed cases were assumed to be I(0) = 4 and E(0) = 20I(0), Fig. 3b with Fig. 2b, the peak of the number of infected cases could
respectively. See Table 1. Fig. 2a shows the results of the initial be reduced from about 19 million cases to about 7500 cases.
fitting of the model based on the data in South Korea while Fig. 2b Subsequently, beginning 20th March 2020, stronger infectious
shows the projections of the model when no control action is disease control measures for travellers coming from overseas
taken. There are some minor discrepancies between the modelled were enforced, where all Koreans and foreigners with residence
values and the real-world data during the initial stage of the in Korea arriving from all countries would be subject to self-
simulation as seen in Fig. 2a. This is absolutely reasonable and quarantine for 14 days upon entry. All short-term travellers will

Fig. 2. Subfigure (a) shows the initial fitting of the model onto the data in South Korea and subfigure (b) shows the projections of the model when no control action
is taken.
6 K.Y. Ng and M.M. Gui / Physica D 411 (2020) 132599

Fig. 3. The projections of the model onto the data in South Korea when control action with an efficiency of 88% is taken. Subfigure (a) shows the projections during
the first 100 days while subfigure (b) shows the projections until the system achieves stability.

Fig. 4. The projections of the model onto the data in South Korea when a second control action with an efficiency of 50% is taken. Subfigure (a) shows the projections
during the first 100 days while subfigure (b) shows the projections up till 300 days.

also be ordered to self-quarantine with exceptions for some lim- time after being cleared of the virus [5,35,36]. On the other hand,
ited special cases [33]. Around the same time, the Korea Centres most health authorities believe that patients who recovered may
for Disease Control and Prevention (KCDC) also started advising develop an immunity towards the virus. However, it is not sure
all people in the country to observe an ‘‘Enhanced Social Distanc- if the said immunity is short-term or long-term. Hence, further
ing Campaign’’ [34]. Inducing these control actions into the model research is required to provide clinical proofs to this hypothesis.
produces the results shown in Fig. 4. The second control action As such, we repeated the simulation for the Case Study on
adds another efficiency of 50%, hence bringing the reproduction South Korea without control action, but with the inclusion of
number further down to R0 ≈ 0.31. a possibility of resusceptibility. Here, we assumed that 1% of
the patients who recovered are resusceptible towards the virus
4.1.1. Simulation with resusceptibility (ξ = 0.01), where the patients develop temporal immunity
One of the many uncertainties about COVID-19 is whether of τξ = 0, 30, 90, 360 days, respectively after recovering from
patients who have recovered from the virus will be reinfected the initial infection of the virus. Fig. 5 shows the simulation
in the future. There have been reports in the news that patients results, where the first infection spikes shown in all subfigures
who recovered from the virus were tested positive for a second are synonymous to the result presented in Fig. 2b. The subsequent
 K.Y. Ng and M.M. Gui / Physica D 411 (2020) 132599 7

Fig. 5. Trajectories for the infected and fatalities in South Korea due to resusceptibility where it is assumed that 1% of the recovered cases are reinfected after a
time span of temporal immunity of (a) 0 day, (b) 30 days, (c) 90 days, and (d) 360 days, respectively. However, these results only apply assuming that if there is
no control action being taken to flatten the curve.

infection spikes are the result of resusceptibility, depending on Table 2

the period of temporal immune response. The results show new Initial parameters used to fit the model to the data in Northern Ireland.

surges in infection cases after the specific τξ in each case, which Parameter Value

diminish over time as more people develop immunity towards Stock population, N 1.88×106
the virus. Interestingly, for the result shown in Fig. 5d where τξ = Fraction of elderly population, Nold 0.18
Percentage of recovery, κ 0.94
360 days, it could also be used to reflect on the situation where Fatality rate for elderly, 1 − κold 0.12
the virus may exhibit similar characteristics as the seasonal flu Incubation time, τinc 5.1 days
or the pandemic influenza A (pH1N1) that it is most likely active Recovery time, τrec 18.8 days
during certain seasons of the year, e.g. autumn/winter for the Basic reproduction number, R0 5.0 (95% CI: 4.85–5.15)
Initial infected cases, I(0) 3
seasonal flu and spring/summer for the pH1N1, in which case an Initial exposed cases, E(0) 60
annual vaccine administration is necessary [37,38].

4.2. Case Study 2: Prediction using data in Northern Ireland

of the initial fitting, with Fig. 6b depicting the projections of
Given the location of which this research is based, data in the infected and deaths if no control action is taken. We then
Northern Ireland is used for prediction study of the model. The simulated the model based on the control action carried out;
reports on confirmed and death cases are published daily since most schools in Northern Ireland were closed beginning 18th
24th March 2020 by the Northern Ireland Public Health Agency March 2020 followed by a national lockdown initiated by the
(PHA) via their Daily COVID-19 Surveillance Reports [39]. The first United Kingdom government on the 23rd March 2020. As such,
confirmed case was recorded on 27th February 2020, and as of we set τpre−σ = 20 (20 days) to correspond to said dates since
20th April 2020, the total number of confirmed cases stood at the first confirmed case, and assuming that it took a further
2728 with 207 fatalities. approximately 12 days for the public to respond to these control
We used the parameters in Table 2 for the initial fitting of actions, i.e. τpost −σ = 12, we obtained the simulation results as
the model based on the data from PHA on the initial growth-rate shown in Fig. 7. The results show that in order for the model to
of the epidemic in Northern Ireland. Fig. 6a shows the results follow the projected trajectories of the data from PHA in Fig. 7a,
8 K.Y. Ng and M.M. Gui / Physica D 411 (2020) 132599

Fig. 6. Subfigure (a) shows the initial fitting of the model onto the data in Northern Ireland and subfigure (b) shows the projections of the model when no control
action is taken.

Fig. 7. The projections of the model onto the data in Northern Ireland when control action with an efficiency of 45% is taken. Subfigure (a) shows the projections
during the first 70 days while subfigure (b) shows the projections until the system achieves stability.

the control action has to achieve an efficiency of about 45% (σ = SARS-CoV-2 required critical care [41]. Meanwhile in Italy, as of
0.45), which indicates that the reproduction number could be 29th March 2020, up to 12% of all positive cases were admitted to
reduced to R0 ≈ 2.75. Comparing Fig. 6b with Fig. 7b, the peak the intensive care unit (ICU) [42]. As such, should we assume that
of the number of infected cases could be reduced from 650,000 approximately 10% of those tested positive in Northern Ireland
cases to 350,000 cases. would require ICU admission, then the peak of the infection curve
should not exceed 3300 cases, i.e. more control actions have to be
4.2.1. Further control action to meet critical Care Capacity taken to reduce the peak of 350,000 cases as seen in Fig. 7b.
However, based on the results shown in Fig. 7, it is essential Therefore, on day 38, which is about one week after the
to further flatten this curve due to the limit of about 330 critical infection curve started to flatten due to the first control action, a
care beds available in Northern Ireland (100 initial setup + 230 in- second control action was introduced into the model. This second
troduced by the newly built Nightingale hospital) [40]. According control action also reflects on the announcement made by the
to the Intensive Care National Audit & Research Centre (ICNARC) United Kingdom government in early April 2020 to allow police
with its ‘‘Report on 2249 patients critically ill with COVID-19’’ officers to enforce social distancing measures. Assuming that this
dated 4th April 2020, about 6% of those tested positive for the second control action results in a further efficiency of 66%, the
 K.Y. Ng and M.M. Gui / Physica D 411 (2020) 132599 9

Fig. 8. Subfigure (a) shows the initial projections of the infected and cumulative deaths curves in Northern Ireland after the second control action with an efficiency
of 66% was introduced on day 38 of the model. Subfigure (b) shows the remaining projections of the model.

reproduction number could then be reduced to R0 ≈ 0.93, and Interesting future research and expansion of the model include
that it would take another 7 days for the public to fully respond but not limited to the predictions for the occupancy of ICU beds,
to the control action, we could observe the results as shown in the effects of easing control action on R0 and hence, the time of
Fig. 8. With the initiation of the second control action, it can be which control action has to be reinstigated, as well as specific
seen in Fig. 8a that the peak in the infection curve is now reduced subregions analysis such as demographic information to model
to 3500 cases. As such, the critical care capacity should be able to the transmission of the virus among subregions in the country to
meet the demand for treatment based on the same assumption cater for population movements and travels.
made earlier in this section, where it is estimated about 10% of the
infected cases are admitted to the ICU. Another observation worth CRediT authorship contribution statement
noting is that the number of deaths in the worst case scenario
has also been reduced to about 3000 cases. See Fig. 8b. This Kok Yew Ng: Methodology, Software, Formal analysis, Visual-
value echos the projection made by the government in Northern ization, Writing - original draft. Meei Mei Gui: Conceptualization,
Ireland that COVID-19 could lead to 3000 deaths during the first Data curation, Formal analysis, Writing - review & editing.
wave [43].
Appendix A. Description of the simulation package
5. Conclusion
Fig. A.9 shows the graphical user interface (GUI) of the sim-
This paper has presented a robust model for COVID-19 based ulation package developed using the MATLAB/Simulink environ-
on a modified SEIRS method to include considerations for the ment. Users can use this interface to set preferred settings for the
ageing population, and time delay for control action as well simulation and also to view simulation results. The simulation kit
as resusceptibility of the recovered population due to temporal can be downloaded from https://github.com/nkymark/COVIDSim.
immunity. Two case studies using real-world data were presented
in this research; the first case for verification of the model based A.1. Establishing simulation settings
on the data in South Korea including a study on the possibility of
resusceptibility of recovered population; and the second case for At the top section of the GUI are some interactive interfaces
prediction study of the model using data and up-to-date control available for the user to set key simulation settings, which include
action and related events in Northern Ireland. The simulation the following:
results from the case studies have clearly shown that the time
• General Settings:
of which the control action is taken and also the time for the
public to properly respond to such intervention measures are – Stock Data: Use this to load real-world data of select
critical in helping to flatten the curve. Also, until a time where countries. The data are obtained from [1].
a vaccine is developed and made available to the general public, – Stock Population: The stock population N is entered
the possibility of resusceptibility, no matter how small, will lead here.
to subsequent waves of infections in the future depending on the – Recovered Cases: Use this to set the percentage of re-
time of temporal immunity. A simulation package was developed covered cases κ .
using the MATLAB/Simulink environment to ease understanding – Elderly Population: Use this to set the fraction of elderly
on the spread of the virus as well as the efficiency that needs to population (above 65 years of age) Nold .
be achieved by the control action in order to successfully flatten – Elderly Fatality Rate: Use this to set the fatality rate
the infection curve to not overload the healthcare capacity. (1 − κold ) for the elderly population.
10 K.Y. Ng and M.M. Gui / Physica D 411 (2020) 132599

Fig. A.9. The main graphical user interface of the simulation package in MATLAB. ⃝ 1 Load real-world data for the selected country. ⃝ 2 Set the stock population N
3 Set the percentage of recovered cases κ . ⃝
for simulation. ⃝ 5 Set the fatality rate for the elderly population (1 − κold ).
4 Set the fraction of elderly population Nold . ⃝
⃝6 Computed values for β, α , and γ using values entered for R0 , τinc , and τrec . ⃝ 7 Set the simulation time in days. ⃝ 8 Set the value for the basic reproduction
number R0 . ⃝9 Set the initial number of infected cases I(0). ⃝ 10 Set the incubation time τinc . ⃝
11 Set the recovery time τrec . ⃝
12 Settings for resusceptibility, including
the percentage of resusceptible cases ξ and duration of temporal immunity τξ . ⃝ 13 Settings for control action, including the efficiency rate σ as well as the time
delay during pre- and post-control action, τpre−σ and τpost −σ , respectively. ⃝
14 Reset the GUI and clear all plots. ⃝15 Run the simulation. ⃝ 16 Recreate the graphs on
external MATLAB figure windows. ⃝ 17 Graphical plots from the simulation (left figure for overall simulation while right figure compares initial projections of the
model with real-world data).

– SEIR Parameters: Use this to set the values for R0 , τinc , References
τrec , the initial infected cases I(0), and the simulation
time. [1] E. Dong, H. Du, L. Gardner, An interactive web-based dashboard to track
COVID-19 in real time, The Lancet Infectious Diseases 20 (5) (2020)
• Resusceptibility Settings: 533–534, http://dx.doi.org/10.1016/S1473-3099(20)30120-1.
[2] F. Zhou, T. Yu, R. Du, G. Fan, Y. Liu, Z. Liu, J. Xiang, Y. Wang, B. Song, X. Gu,
– Resusceptible Cases: Use this to set the percentage of L. Guan, Y. Wei, H. Li, X. Wu, J. Xu, S. Tu, Y. Zhang, H. Chen, B. Cao, Clinical
recovered cases who are resusceptible. course and risk factors for mortality of adult inpatients with COVID-19 in
wuhan, China: A retrospective cohort study, Lancet 395 (10229) (2020)
– Duration of temporal immunity: Use this to set the time 1054–1062, http://dx.doi.org/10.1016/S0140-6736(20)30566-3.
of short-term immune response τξ , assuming there is [3] E. Prompetchara, C. Ketloy, T. Palaga, Immune responses in COVID-19 and
no permanent immunity after recovery. potential vaccines: Lessons learned from SARS and MERS epidemic, Asian
Pacific J. Allergy Immunol. 38 (2020) 1–9, http://dx.doi.org/10.12932/AP-
• Control Action Settings: 200220-0772.
[4] I. Thevarajan, T.H.O. Nguyen, M. Koutsakos, J. Druce, L. Caly, C.E. van de
– Control Action Efficiency: Use this to set the percentage Sandt, X. Jia, S. Nicholson, M. Catton, B. Cowie, S.Y.C. Tong, S.R. Lewin,
of control action efficiency σ . K. Kedzierska, Breadth of concomitant immune responses prior to patient
recovery: A case report of non-severe COVID-19, Nat. Med. 26 (4) (2020)
– Pre-action Time Delay: Use this to set the time delay
453–455, http://dx.doi.org/10.1038/s41591-020-0819-2.
τpre−σ for the control action to be introduced after the [5] Questions raised over COVID-19 reinfection after Japanese woman devel-
first confirmed case. ops illness again, Japan Times (2020-02-28) https://www.japantimes.co.jp/
– Post-action Time Delay: Use this to set the time delay news/2020/02/28/national/coronavirus-reinfection/.
τpost −σ to mimic the time it takes for the population to [6] P. van den Driessche, Reproduction numbers of infectious disease models,
Infect. Dis. Modell. 2 (3) (2017) 288–303, http://dx.doi.org/10.1016/j.idm.
respond to the control action. 2017.06.002.
[7] R. Peng, F. Yi, Asymptotic profile of the positive steady state for an SIS
epidemic reaction–diffusion model: Effects of epidemic risk and population
A.2. Simulation results movement, Physica D 259 (2013) 8–25, http://dx.doi.org/10.1016/j.physd.
2013.05.006.
[8] H. Zhang, M. Small, X. Fu, G. Sun, B. Wang, Modeling the influence of
The simulation results are displayed at the bottom section of information on the coevolution of contact networks and the dynamics of
the GUI. The plot on the right shows the initial fit of the model infectious diseases, Physica D 241 (18) (2012) 1512–1517, http://dx.doi.
org/10.1016/j.physd.2012.05.011.
using the settings established in Appendix A.1 onto the real-world
[9] M.J. Keeling, P. Rohani, Modeling Infectious Diseases in Humans and
data of the select country, while the plot on the left shows the Animals, Princeton University Press, 2008, http://dx.doi.org/10.2307/j.
simulation results until the simulation stop time. ctvcm4gk0, http://www.jstor.org/stable/j.ctvcm4gk0.
 K.Y. Ng and M.M. Gui / Physica D 411 (2020) 132599 11

[10] P. van den Driessche, J. Watmough, Further notes on the basic reproduction [24] G. Stewart, K. van Heusden, G.A. Dumont, How control theory can
number, in: F. Brauer, P. van den Driessche, J. Wu (Eds.), Mathemat- help us control COVID-19, IEEE Spectr. (2020) https://spectrum.ieee.org/
ical Epidemiology, Springer Berlin Heidelberg, Berlin, Heidelberg, 2008, biomedical/diagnostics/how-control-theory-can-help-control-covid19.
pp. 159–178, http://dx.doi.org/10.1007/978-3-540-78911-6_6. [25] S. Tipsri, W. Chinviriyasit, The effect of time delay on the dynamics of an
[11] P.E. Lekone, B.F. Finkenstädt, Statistical inference in a stochastic epidemic SEIR model with nonlinear incidence, Chaos Solitons Fractals 75 (2015)
SEIR model with control intervention: Ebola as a case study, Biometrics 62 153–172, http://dx.doi.org/10.1016/j.chaos.2015.02.017.
(4) (2006) 1170–1177, http://dx.doi.org/10.1111/j.1541-0420.2006.00609. [26] G. Huang, Y. Takeuchi, W. Ma, D. Wei, Global stability for delay SIR and
xCitations, http://www.jstor.org/stable/4124539. SEIR epidemic models with nonlinear incidence rate, Bull. Math. Biol. 72
[12] C. Dye, N. Gay, Modeling the SARS epidemic, Science 300 (5627) (2003) (5) (2010) 1192–1207, http://dx.doi.org/10.1007/s11538-009-9487-6.
1884–1885, http://dx.doi.org/10.1126/science.1086925. [27] P. Yan, S. Liu, SEIR Epidemic model with delay, ANZIAM J. 48 (1) (2006)
[13] M.M. Saito, S. Imoto, R. Yamaguchi, H. Sato, H. Nakada, M. Kami, S. Miyano, 119–134, http://dx.doi.org/10.1017/S144618110000345X.
T. Higuchi, Extension and verification of the SEIR model on the 2009 [28] E. Shim, A. Tariq, W. Choi, Y. Lee, G. Chowell, Transmission potential and
influenza A (H1N1) pandemic in Japan., Math Biosci 246 (1) (2013) 47–54, severity of COVID-19 in South Korea, Int. J. Infect. Dis. 93 (2020) 339–344,
http://dx.doi.org/10.1016/j.mbs.2013.08.009. http://dx.doi.org/10.1016/j.ijid.2020.03.031.
[14] G. Chowell, S. Blumberg, L. Simonsen, M.A. Miller, C. Viboud, Synthesizing [29] D. Normile, Coronavirus cases have dropped sharply in South Korea. What’s
data and models for the spread of MERS-CoV, 2013: Key role of index the secret to its success? Science (2020) http://dx.doi.org/10.1126/science.
cases and hospital transmission, Epidemics 9 (2014) 40–51, http://dx.doi. abb7566.
org/10.1016/j.epidem.2014.09.011. [30] OECD, Elderly Population (Indicator), OECD Publishing Paris, 2018, http:
[15] P. Diaz, P. Constantine, K. Kalmbach, E. Jones, S. Pankavich, A modified SEIR //dx.doi.org/10.1787/8d805ea1-en.
model for the spread of Ebola in Western Africa and metrics for resource [31] The updates on COVID-19 in Korea as of 1 April, in: Division of Risk as-
allocation, Appl. Math. Comput. 324 (2018) 141–155, http://dx.doi.org/10. sessment and International cooperation, Korea Centers for Disease Control
1016/j.amc.2017.11.039. and Prevention, 2020, https://www.cdc.go.kr/menu.es?mid=a30402000000.
[16] Q. Lin, S. Zhao, D. Gao, Y. Lou, S. Yang, S.S. Musa, M.H. Wang, Y. Cai, [32] S. Flaxman, S. Mishra, A. Gandy, et al., Report 13: Estimating the num-
W. Wang, L. Yang, D. He, A conceptual model for the coronavirus disease ber of infections and the impact of non-pharmaceutical interventions
2019 (COVID-19) outbreak in Wuhan, China with individual reaction and on COVID-19 in 11 European countries, 2020, http://dx.doi.org/10.25561/
governmental action, Int. J. Infect. Dis. 93 (2020) 211–216, http://dx.doi. 77731.
org/10.1016/j.ijid.2020.02.058. [33] The updates on COVID-19 in Korea as of 23 March, Division of Risk assess-
[17] L. López, X. Rodó, A modified SEIR model to predict the COVID-19 ment and International cooperation, Korea Centers for Disease Control and
outbreak in Spain: Simulating control scenarios and multi-scale epi- Prevention, 2020, https://www.cdc.go.kr/menu.es?mid=a30402000000.
demics (4/10/2020), Available at SSRN (2020) http://dx.doi.org/10.2139/ [34] A timeline of South Korea’s response to COVID-19, in: The Center for
ssrn.3576802. Strategic and International Studies (CSIS), 2020, https://www.csis.org/
[18] A.J. Kucharski, T.W. Russell, C. Diamond, Y. Liu, J. Edmunds, S. Funk, R.M. analysis/timeline-south-koreas-response-covid-19.
Eggo, F. Sun, M. Jit, J.D. Munday, N. Davies, A. Gimma, K. van Zandvoort, H. [35] S. Korean woman reinfected with coronavirus after recovery, Yonhap News
Gibbs, J. Hellewell, C.I. Jarvis, S. Clifford, B.J. Quilty, N.I. Bosse, S. Abbott, P. (2020-02-29) http://yna.kr/AEN20200228009700320.
Klepac, S. Flasche, Early dynamics of transmission and control of COVID- [36] E. Dahl, Coronavirus (Covid-19) outbreak on the cruise ship Diamond
19: A mathematical modelling study, The Lancet Infectious Diseases 20 (5) Princess, Int. Marit. Health 71 (1) (2020) 5–8, http://dx.doi.org/10.5603/
MH.2020.0003.
(2020) 553–558, http://dx.doi.org/10.1016/S1473-3099(20)30144-4.
[37] H. Kelly, K. Grant, Interim analysis of pandemic influenza (H1N1) 2009 in
[19] M. De la Sen, R.P. Agarwal, A. Ibeas, S. Alonso-Quesada, On a generalized
Australia: Surveillance trends, age of infection and effectiveness of seasonal
time-varying SEIR epidemic model with mixed point and distributed time-
vaccination, Eurosurveillance 14 (31) (2009) http://dx.doi.org/10.2807/ese.
varying delays and combined regular and impulsive vaccination controls,
14.31.19288-en.
Adv. Difference Equ. 2010 (1) (2010) 281612, http://dx.doi.org/10.1155/
[38] S. Cook, C. Conrad, A.L. Fowlkes, M.H. Mohebbi, Assessing google flu trends
2010/281612.
performance in the United States during the 2009 influenza virus A (H1N1)
[20] W.O. Kermack, A.G. McKendrick, A contribution to the mathematical theory
pandemic, PLoS One 6 (8) (2011) e23610, http://dx.doi.org/10.1371/journal.
of epidemics, Proc. R. Soc. Lond. A 115 (772) (1927) 700–721.
pone.0023610.
[21] K. Prem, Y. Liu, T.W. Russell, A.J. Kucharski, R.M. Eggo, N. Davies, S. Flasche,
[39] Daily COVID-19 Surveillance Reports, Northern Ireland Public Health
S. Clifford, C.A.B. Pearson, J.D. Munday, S. Abbott, H. Gibbs, A. Rosello, B.J.
Agency (PHA), 2020, https://www.publichealth.hscni.net/publications/
Quilty, T. Jombart, F. Sun, C. Diamond, A. Gimma, K. van Zandvoort, S. Funk,
covid-19-surveillance-reports.
C.I. Jarvis, W.J. Edmunds, N.I. Bosse, J. Hellewell, M. Jit, P. Klepac, The effect
[40] Coronavirus: City hospital to host NI’s first nightingale, BBC News (2020)
of control strategies to reduce social mixing on outcomes of the COVID-19
https://www.bbc.co.uk/news/uk-northern-ireland-52129644.
epidemic in Wuhan, China: A modelling study, Lancet Public Health 5 (5) [41] ICNARC report on COVID-19 in critical care, Intensive Care National Audit
(2020) e261–e270, http://dx.doi.org/10.1016/S2468-2667(20)30073-6. & Research Centre (ICNARC). https://www.icnarc.org/About/Latest-News/
[22] Y. Fang, Y. Nie, M. Penny, Transmission dynamics of the COVID-19 outbreak 2020/04/04/Report-On-2249-Patients-Critically-Ill-With-Covid-19.
and effectiveness of government interventions: A data-driven analysis, J. [42] J. Phua, L. Weng, L. Ling, M. Egi, C.-M. Lim, J.V. Divatia, B.R. Shrestha,
Med. Virol. 92 (6) (2020) 645–659, http://dx.doi.org/10.1002/jmv.25750. Y.M. Arabi, J. Ng, C.D. Gomersall, M. Nishimura, Y. Koh, B. Du, Intensive
[23] Severe outcomes among patients with coronavirus disease 2019 (COVID- care management of coronavirus disease 2019 (COVID-19): Challenges
19) — United States, February 12–March 16, 2020, MMWR Morb Mortal and recommendations, Lancet Respir. Med. 8 (5) (2020) 506–517, http:
Wkly Rep 2020 69 (2020) 343–346, http://dx.doi.org/10.15585/mmwr. //dx.doi.org/10.1016/S2213-2600(20)30161-2.
mm6912e2. [43] Coronavirus: NI could see 3,000 deaths in ‘first wave’, BBC News (2020)
https://www.bbc.co.uk/news/uk-northern-ireland-52121143.