COVID 19 Pandemic Dynamics Mathematical Simulations

Visit the link below to download the full version of this book:

https://medipdf.com/product/covid-19-pandemic-dynamics-mathematical-simulations/

Click Download Now

To my parents, Kateryna and Georgii, who
are bravely fighting the pandemic in isolation
Introduction

The COVID-19 pandemic poses a great threat due to millions of infected people,
high mortality, and a very negative impact on the economy. Its detailed investi-
gations are still ahead, but the public is already interested in the duration of the
pandemic, the expected number of patients, estimations of quarantine measures,
the scale of recurrences, etc. The threats of the COVID-19 pandemic require the
mobilization of scientists, including mathematicians familiar with methods of
infectious disease simulation.
The more complex the mathematical model, the more unknown parameters it
contains, the values of which must be determined using a limited number of
observations of the disease over time. Even long-term monitoring of the epidemic
may not provide reliable estimates of its parameters due to the constant change in
quarantine and testing conditions, in algorithms of isolation of infected persons, in
pathogen activity, etc.
Any mathematical modeling of the epidemic dynamics will be of particular value
if we make an accurate long-term forecast of its duration and number of diseases
using statistics data sets obtained immediately after the outbreak. That is why many
authors were trying to predict the COVID-19 pandemic dynamics in many countries
and regions. We will not dwell on a detailed analysis of these studies and only note
that the correct mathematical simulation of the COVID-19 pandemic is very dif-
ficult for at least two reasons.
First, data on the number of cases are clearly incomplete immediately after onset,
there are quite long hidden periods. The reason is the large number of asymptomatic
patients and the lack of skills to detect a new disease. It must be noted that a large
discrepancy between the registered and actual number of cases occurred even for
the later periods of the COVID-19 pandemic. Adequate modeling is further com-
plicated by the fact that we do not know when the number of reported cases is
approaching the actual number. The second reason for the limited accuracy of
long-term forecasts is the constant changes in the conditions of the pandemic
(changing quarantine measures, social behavior, virulence of the pathogen, etc.).
Therefore, a prediction made using statistics for a certain time period is not suitable
for other periods of time.

vii
viii Introduction

In this book, we will focus on the simplest models that describe the development
of infectious diseases over time. In particular, the initial stages of epidemics in
different regions, which are characterized by an exponential increase in the number
of cases, will be studied. Simple comparisons of the pandemic dynamics in different
countries and its trends will be presented. In these parts of the book, I shall use
some results of articles and preprints, written together with Gerhard Demelmair,
Ihor Kudybyn, Anatolii Nikitin, and Bohdan Shepetyuk, to whom I am also very
grateful for collecting and systematizing statistical information on the number of
registered cases of the disease.
In this book, we will also use the classical SIR model, with three differential
equations for the evolution of the number of susceptible persons—S; infected,
spreading the infection—I and removed persons—R, which is the sum of isolated,
immunized, and deceased persons. This model contains only four parameters, the
values of which can be estimated using a statistical approach developed and suc-
cessfully applied for investigations of the mysterious children's disease that
occurred in the Ukrainian city of Chernivtsi in 1988.
The SIR model is unable to determine the duration of the incubation period and to
predict separately the number of deaths caused by coronavirus, but allows us to make
adequate predictions of the duration of epidemics in different countries, estimates
of their actual beginning (they may precede the time of registration of the first patient),
to calculate the time dependences of the total number of patients V = I + R and I, to
estimate the probability of meeting an infected person and the effective reproduction
numbers. Corresponding results for the first COVID-19 epidemic waves in mainland
China, USA, Germany, the UK, the Republic of Korea, Austria, Italy, Spain, France,
the Republic of Moldova, Ukraine, the city of Kyiv and for the whole world are
already published in my articles and preprints. In the book, these results are sys-
tematized and conclusions are drawn based on current information about the course
of the pandemic. In Chap. 7 we will compare and discuss the characteristics of the first
waves of the COVID-19 pandemic in different countries and WHO regions.
Constant changes in the pandemic conditions (i.e., in the peculiarities of quar-
antine and its violation, in situations with testing and isolation of patients, in
coronavirus activity due to its mutations, etc.) cause changes in the values of
parameters of the mathematical models and lead to new pandemic waves. In par-
ticular, in October 2020 we observed a sharp increase in the daily number of new
cases in many European countries. We will develop simple methods of detecting
these changes and propose a simple method of identifying new pandemic waves.
The numerical differentiation of smoothed dependences of the accumulated number
of cases allows selecting periods with different values of SIR parameters.
To simulate different pandemic waves (periods with more or less constant values
of its dynamics parameters), a general SIR model and its exact solution will be
proposed. The identification procedures for the parameters of the general SIR model
will be described. The characteristics of several pandemic waves in Ukraine and the
world will be calculated and corresponding predictions will be presented.
To have good accuracy of predictions, the pandemic dynamics must be updated
with the use of new data sets. Because of this, a simple method to assess the final
Introduction ix

size and duration of epidemic waves was proposed in Chap. 13 and applied for
USA, Germany, the UK, the Republic of Korea, Italy, Austria, Spain, France, the
Republic of Moldova, Sweden, Ukraine, the city of Kyiv and the whole world.
I would like to express my sincere thanks to Profs. Dirk Langemann (Technische
Universitaet Braunschweig) and Juergen Prestin (Universitaet zu Luebeck) for their
support in developing the used optimization approach. I would also like to thank
Profs. Alberto Redaelli, Giuseppe Passoni, and Gianfranco Fiore (Politecnico di
Milano), Sergei Pereverzyev (RICAM, Linz, Austria) for involving me in very
interesting biomedical investigations under the EU-funded Horizon 2020 projects:
EUMLS (grant PIRSES-GA-2011-295164-EUMLS) and AMMODIT (grant
MSCA-RISE 645672).
Finally, my thanks to Academician Viktor Grinchenko, Profs. Volodymyr
Tymofeyev, Alexander Galkin, Pavlo Maslianko, my friends Liudmyla Trotsenko,
Maia Troian, Nina Basiuk, Anna Voronka, Daria Cusitcaia, Damien Berezenko,
Anatolii Podkur, Volodymyr Borysenko, and Oleksii Rodionov for their support
and help in collecting and processing data.
Contents

1 Data Used for Calculations, Comparisons, and Verifications . . . . . 1

2 Early Stages of Epidemics and Exponential Growth . . . . . . . . . . . . 7
3 Comparisons of the Early Stages of the COVID-19 Pandemic
in Different Regions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4 Classical SIR Model and the Exact Solution of Differential
Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
5 Statistics-Based Procedure of Parameter Identification
for the Classical SIR Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
6 SIR Simulations for the First Waves of the COVID-19 Pandemic
in Different Countries and Regions . . . . . . . . . . . . . . . . . . . . . . . . . 37
6.1 Mainland China . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
6.2 The Republic of Korea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
6.3 Italy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
6.4 Austria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
6.5 Spain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
6.6 France . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
6.7 Germany . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
6.8 The UK . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
6.9 USA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
6.10 The Republic of Moldova . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
6.11 Ukraine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
6.12 Kyiv . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
6.13 The World . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
7 Comparison of the First Waves of the COVID-19 Pandemic
in Different Countries and Regions . . . . . . . . . . . . . . . . . . . . . . . . . 89
8 Identification of the New Waves of the COVID-19 Pandemic . . . . . 109

xi
xii Contents

9 General SIR Model and Its Exact Solution . . . . . . . . . . . . . . . . . . . 127

10 Procedures of Parameter Identification for the Waves
of Epidemics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
11 Applications of the General SIR Model for Calculations
of the COVID-19 Epidemic Waves in Ukraine . . . . . . . . . . . . . . . . 141
12 Global Waves of the COVID-19 Pandemic . . . . . . . . . . . . . . . . . . . 147
13 Long-Time Predictions for the Pandemic Dynamics . . . . . . . . . . . . 153
14 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
Chapter 1
Data Used for Calculations,
Comparisons, and Verifications

Sources of data sets for COVID-19 pandemic dynamics, some examples and
principles of their using will be presented in this chapter.
For calculations, we will use the official data about the accumulated numbers of
confirmed COVID-19 cases Vj; j ¼ 1; 2; 3; . . .; n in different countries and world-
wide. The number of observations n may vary for different investigations. The main
source of this information is WHO daily situation reports, [1]. The corresponding
time moments tj are measured in days. We will use different zero points for the time,
specifying them for every calculation (see examples in Tables 1.1 and 1.2). It must
be noted that Vj shows the accumulated number of laboratory confirmed cases at the
end of corresponding day. It means that the number of infected people could be
much higher at corresponding day, since there were and still are rather long delays
between the day of testing and day when the results of tests are ready. For example,
in Ukraine this delay was between 1 and 10 days (according to the type laboratory
and the length of the queue).
We will use and analyze also the daily data for the number of confirmed cases in
mainland China, which origins from National Health Commission (NHC) of the
People’s Republic of China [2] (see Table 1.2). On February 12, 2020, NHC has
added 12,289 new cases (not previously included in official counts) as “clinically
diagnosed cases”. The cases, reported by this official organization before, have the
name of “tested confirmed cases” [2]. To avoid confusing, we will denote “tested
confirmed cases” as Wj and the “clinically diagnosed cases” as Qj; j corresponds to
the different time moments tj (see Table 1.2). A part of the official diagram (its
version, presented on February 15, 2020) is shown in Fig. 1.1. In some urgent

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 1
I. Nesteruk, COVID-19 Pandemic Dynamics,
https://doi.org/10.1007/978-981-33-6416-5_1
2 1 Data Used for Calculations, Comparisons, and Verifications

Table 1.1 Official cumulative numbers of confirmed cases in the Republic of Korea
Day in Time Accumulated Day in Time Accumulated
February moment number of cases in March moment number of cases in
2020 tj the Republic of 2020 tj the Republic of
Korea Vj, [1] Korea Vj, [1]
17 −1 31 1 12 4212
18 0 51 2 13 4812
19 1 104 3 14 5328
20 2 204 4 15 5766
21 3 346 5 16 6284
22 4 602 6 17 6767
23 5 763 7 18 7134
24 6 977 8 19 7382
25 7 1261 9 20 7513
26 8 1766 10 21 7755
27 9 2337 11 22 7869
28 10 3150 12 23 7979
29 11 3736 13 24 8086
– – – 14 25 8165
The information from [1]. The corresponding time moments tj and the accumulated confirmed
numbers of cases Vj in South Korea

situations (e.g., after the epidemic outbreak in Italy in February 2020), the infor-
mation from media [3, 4] was used. The number of cases in Europe was taken from
[5]. The accumulated number of cases in Wuhan (China) in December 2019 and
January 2020 was calculated with the use of daily data reported in [6] (see
Table 1.2).
Some data sets or their parts will be used for calculations, some of them only for
verifications of calculations and comparisons. For example, most of the values Wj
and Wj + Qj shown in Table 1.2 were used in two SIR simulations of epidemic
dynamics in mainland China, some of them to verify the reliability of these cal-
culations, and the values Vj (also listed in Table 1.2) only for comparisons. All data
used for calculations will be given in the relevant tables.
Table 1.2 Data sets used for calculations, comparisons and verifications
Day in Time Accumulated Day in Time “Tested Day in Time The sum of “tested confirmed
December moment number of January and moment confirmed February moment cases” and “clinically
2019 and tj cases Vj, [6] February tj cases” Wj, and March tj diagnosed cases” Wj + Qj, [2]
January 2020 2020 [2] 2020
8 −39 1 16 0 45 12 27 58,761
10 −37 2 17 1 62 13 28 63,851
13 −34 3 18 2 121 14 29 66,492
15 −32 5 19 3 198 15 30 68,500
16 −31 6 20 4 291 16 31 70,548
17 −30 7 21 5 440 17 32 72,436
19 −28 9 22 6 571 18 33 74,185
20 −27 13 23 7 830 19 34 74,576
21 −26 17 24 8 1287 20 35 75,465
22 −25 20 25 9 1975 21 36 76,288
23 −24 24 26 10 2744 22 37 76,936
25 −22 28 27 11 4515 23 38 77,150
26 −21 29 28 12 5974 24 39 77,658
1 Data Used for Calculations, Comparisons, and Verifications

27 −20 32 29 13 7711 25 40 78,064

28 −19 34 30 14 9692 26 41 78,497
29 −18 40 31 15 11,791 27 42 78,824
30 −17 45 1 16 14,380 28 43 79,251
31 −16 47 2 17 17,205 29 44 79,824
1 −15 57 3 18 20,440 1 45 80,026
2 −14 64 4 19 24,324 2 46 80,151
3 −13 77 5 20 28,018 – – –
(continued)
3
 4

Table 1.2 (continued)

Day in Time Accumulated Day in Time “Tested Day in Time The sum of “tested confirmed
December moment number of January and moment confirmed February moment cases” and “clinically
2019 and tj cases Vj, [6] February tj cases” Wj, and March tj diagnosed cases” Wj + Qj, [2]
January 2020 2020 [2] 2020
4 −12 90 6 21 31,161 – – –
5 −11 105 7 22 34,568 – – –
6 −10 124 8 23 37,198 – – –
7 −9 155 9 24 40,171 – – –
8 −8 199 10 25 42,638 – – –
1

– – – 11 26 44,653 – – –
The information from [2] and [6]. The corresponding time moments tj and the numbers of Vj, Wj and Wj + Qj
Data Used for Calculations, Comparisons, and Verifications
1 Data Used for Calculations, Comparisons, and Verifications 5

Fig. 1.1 Part of official diagram with the numbers of Wj + Qj (last column) and Wj (previous
column), [2]
Chapter 2
Early Stages of Epidemics
and Exponential Growth

The early stages of the COVID-19 epidemic outbreaks and the pandemic dynamics
in March and April 2020 will be analyzed. The statistics-based method of parameter
estimations for the exponential growth will be presented.
For the initial stage of every epidemic, the exponential growth in the number of
cases V over time t is typical:

V ¼ ect þ b ð2:1Þ

where c and b are constant parameters. This fact was confirmed also for COVID-19
pandemic (see, e.g., [6–32]) and follows from the simple mathematical model,
stating that the increase in the number of cases dV during the time interval dt is
proportional to their number V:

dV ¼ cVdt ð2:2Þ

Formula (2.1) yields the solution to the differential Eq. (2.2) and means that
corresponding points Vj (taken in logarithmic scale) versus time must follow the
straight lines. This simple model does not take into account the restricted volume of
population (according to Eq. (2.1) the number of cases tends to infinity as time
increases) and the fact that some infected persons become isolated or dead (i.e., they
stop to spread the infection). To simulate the deviation from the exponential growth
(2.1) and stabilization of the number of cases at some saturation level, more
complicated approaches are necessary. We will use two of them (classical and
generalized SIR models) later.
Equation (2.1) yields the linear dependence

y  ln V ¼ ct þ b ð2:3Þ

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 7
I. Nesteruk, COVID-19 Pandemic Dynamics,
https://doi.org/10.1007/978-981-33-6416-5_2
8 2 Early Stages of Epidemics and Exponential Growth

and the time of cases duplication:

ln 2
sd ¼ ð2:4Þ
c

To estimate the values of parameters c and b, we can treat the values

yj ¼ logðVj Þ and corresponding time moments tj as random variables and use the
observations of the accumulated number of cases (e.g., presented in Table 1.1)
_
_
and the linear regression [33] in order to calculate the coefficients c and b of the
regression line
_
_ _
y ¼ ct þ b ð2:5Þ

using the standard formulas (see, e.g., [33]):

Pn P P 
n n
_
n j¼1 yj tj  j¼1 yj j¼1 tj
c¼ Pn P 2 ; ð2:6Þ
n
j¼1 tj 
n 2
j¼1 tj

Pn _ Pn
_
j¼1 yj c j¼1 tj
b¼ ð2:7Þ
n

Here n is the number of observations.

_
_
Values c and b can be treated as statistics-based estimations of parameters c and
b from relationships (2.1) or (2.3). The reliability of this estimation can be checked
by calculating the correlation coefficient r:
P P P 
n nj¼1 yj tj  n
j¼1 y j
n
j¼1 t j
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
r ¼ s  ffi ð2:8Þ
Pn 2  Pn  2 Pn 2 Pn 2
n j¼1 yj  j¼1 yj n j¼1 tj  j¼1 tj

If the values of jr j are close to unit, there is a strong linear relationship between
variables y and t.
We can use also the F-test for the null hypothesis that says that the proposed
linear relationship (2.3) fits the data set. The experimental value of the Fisher
function can be calculated with the use of the formula:

r 2 ðn  mÞ
F¼ ð2:9Þ
ð1  r 2 Þðm  1Þ