.
Navarro Romero, Gelves Alarcón & Carcía Corrales / INGE CUC, vol. 17 no. 1, pp. 285–302. Enero - Junio, 2021

Correlational analysis between the economics, socio-demographic indices and statistics of contagion due to Covid-19, applying the Clustering methodology in countries of America

Análisis correlacional entre la economía, los índices sociodemográficos y las estadísticas de contagio por Covid-19, aplicando la metodología de Clustering en países de América

DOI: http://doi.org/10.17981/ingecuc.17.1.2021.21

Artículo de Investigación Científica. Fecha de Recepción: 01/12/2020. Fecha de Aceptación: 25/01/2021.

Elisa del Carmen Navarro-Romero

Universidad Santo Tomás. Bogotá, D.C. (Colombia)

elisanavarro@usantotomas.edu.co

Óscar Mauricio Gelves-Alarcón

Universidad Militar Nueva Granada. Bogotá, D.C. (Colombia)

oscar.gelves@unimilitar.edu.co

Natalia García-Corrales

Environment & Technology Foundation. Cereté (Colombia)

etf@environmenttechnologyfoundation.org

To cite this article:

E. Navarro Romero, O. Gelves Alarcón & N. Carcía Corrales, “ Correlational analysis between the economics, socio-demographic indices and statistics of contagion due to Covid-19, applying the Clustering methodology in countries of America”, INGECUC, vol. 17. no. 1, pp. 285–302. DOI: http://doi.org/10.17981/ingecuc.17.1.2021.21

Abstract

Introduction— This research is motivated, by the current world situation, caused by the pandemic declared by the WHO before the spread and severity of the coronavirus disease (COVID-19), notified for the first time in Wuhan (China) on December 31 of 2019. Through mathematical and statistical analysis, it seeks to show and explain in an expeditious manner, the causes for which there is a higher rate of contagion and lethality due to the virus, in different countries, taking into consideration patterns associated with social political behavior and economic, as a first approach to knowing statistics that allow generating forecasts for future periods, given the conditions.

Objective— The main objective of this work is to define the correlation of the economic, social and demographic variables of the countries of America, with respect to the contagion of the virus, proposing a forecast model on the level of contagion in each cluster proposed by the different regions of the American continent.

Methodology— The study performs clustering (grouping) of the countries of America with respect to their geographical position North America, Central America and the Caribbean islands and South America, followed by a search for statistical data on social, economic and demographic indicators of the countries of America in recent years and statistics of levels of contagion of COVID 19 in sources such as international organizations regulating health issues. Next, a characterization and correlation of the collected data was carried out, to finally, based on the results of the correlation, make a forecast of the level of contagion that would be reached by each of the regions.

Results— The purpose of this document is to provide information on the countries of North America, Latin America and the Caribbean with respect to the analysis of mortality from COVID-19, through methods of analysis of mortality from all causes as one of the approaches proposed to contribute to the assessment of the true magnitude of the burden of the COVID-19 epidemic in these countries.

Conclusions— The results show interesting information, since the Latin American curve turned out to be much less pronounced than that of the United States, in terms of contagion and deaths, despite the socio-demographic conditions, economic, technological and political opportunities. This analysis invites us to find out which are those correlations that directly impact the behavior of infections, taking into account variables such as age, gender, stratum, level of education, and other sociodemographic characteristics that may influence the spread or containment of the virus.

Keywords— Grouping; data; ICR; mathematical model; pandemic; linear regression; multiple regression

Resumen

Introducción— Esta investigación está motivada, por la actual situación mundial, provocada por la pandemia declarada por la OMS ante la propagación y gravedad de la enfermedad por coronavirus (COVID-19), notificada por primera vez en Wuhan (China) el 31 de diciembre de 2019. A través del análisis matemático y estadístico, se busca mostrar y explicar de manera expedita, las causas por las cuales existe una mayor tasa de contagio y letalidad por el virus, en diferentes países, tomando en consideración patrones asociados al comportamiento político social y económico, como una primera aproximación para conocer estadísticas que permitan generar pronósticos para periodos futuros, dadas las condiciones.

Objetivo— El objetivo principal de este trabajo es definir la correlación de las variables económicas, sociales y demográficas de los países de América, con respecto al contagio del virus, proponiendo un modelo de pronóstico sobre el nivel de contagio en cada cluster propuesto por las diferentes regiones del continente americano.

Metodología— El estudio realiza una clusterización (agrupación) de los países de América con respecto a su posición geográfica América del Norte, América Central e islas del Caribe y América del Sur, seguido de una búsqueda de datos estadísticos sobre indicadores sociales, económicos y demográficos de los países de América en los últimos años y estadísticas de niveles de contagio del COVID 19 en fuentes como los organismos internacionales que regulan los temas de salud. Luego, se realizó una caracterización y correlación de los datos recolectados, para finalmente, en base a los resultados de la correlación, realizar un pronóstico del nivel de contagio que alcanzaría cada una de las regiones.

Resultados— El propósito de este documento es proporcionar información sobre los países de América del Norte, América Latina y el Caribe con respecto al análisis de la mortalidad por COVID-19, a través de métodos de análisis de la mortalidad por todas las causas como uno de los enfoques propuestos para contribuir a la evaluación de la verdadera magnitud de la carga de la epidemia de COVID-19 en estos países.

Conclusiones— Los resultados muestran información interesante, ya que la curva latinoamericana resultó ser mucho menos pronunciada que la de Estados Unidos, en términos de contagio y muertes, a pesar de las condiciones sociodemográficas, económicas, tecnológicas y políticas. Este análisis invita a averiguar cuáles son las correlaciones que impactan directamente en el comportamiento de los contagios, teniendo en cuenta variables como la edad, el género, el estrato, el nivel de educación y otras características sociodemográficas que pueden influir en la propagación o contención del virus.

Palabras clave— Agrupación; datos; ICR; modelo matemático; pandemia; regresión lineal; regresión múltiple

I. Introduction

Humanity has been affected by different pandemics throughout its history. This term alludes to diseases that attach themselves to individuals from diverse territories or multiple countries [1]. As stated by several researchers [2] the efficiency of a pandemic peaks with a person-to-person transmission. Within the pandemics it’s worth mentioning the ones caused by virus such as H1N1, VIH, and most recently, SARS-CoV-2 (Coivid-19). The propagation of disease due to Coronavirus (COVID-19) has presented a menace to health around the globe. By March 17th of 2020 the number of confirmed cases had risen above 179 000, with more than 7 000 confirmed dead in at least 150 countries [3].

Granted, Covid-19 is considered an airborne transmitted disease, considering that the virus can transfer itself through sneezing, coughing, talking or any activity that result in the generation of aerosolized particles [4]. This requires establishing means of distancing people and generating protection mechanisms such as the use of face masks in environments where the virus can be transmitted with ease.

The impact to the populational density, regarding infectious diseases has been subject to various studies. Density referring to contact and interaction between inhabitants, which makes this variable crucial for the propagation of emerging infectious diseases. In the case of global pandemics such as the recent COVID-19 virus [5], the larger and denser urban centers, particularly those involved with tourist activities, could become in epicenters of a health crisis resulting in the death of thousands. In a similar manner, health & educational systems could help minimize the disease’s total impact for those who have been infected, which leads to a higher recovery rate and a lower mortality rate. Densities tend to vary higher than the confirmed infection rates. The simple correlation between the two figures approximates to 0.48, which means that, for a country, the density only represents around 23% variation of the virus’ propagation. Density and exposure are not a straight line. Dense areas can have a higher chance of putting into practice policies that ensure distancing, reducing real infection rates or simply leading to a large-scale social distancing due to a greater quantity of people conscious about the threat.

During this period the pandemic is taking place in, humanity has understood the importance of possessing data that allows processing, extracting information, and thus, predicting its behavior. As a number of studies [6] epidemiology and biostatistics oversee analyzing the patterns described within the population’s diseases, and in mathematical terms, projecting their future. The Pan-American organization of health [7] highlights the importance of possessing high quality, accessible, trustworthy, opportune, open and reliable data analyzed recently that allow us to generate information geared towards decision making during a pandemic; while at the same time warning us about the difficulties caused by “infopedia”, referring to the health problems rooted in the excess of unreliable information regarding the pandemic’s current behavior, which resulted in trouble for Covid-19 related knowledge by February 2020 [8]. The concern the excess or lack of trustworthy information geared towards decision making that determines the population’s behavior, and thus, influences their health, is approached as an analysis that provides a tool for measuring the concern of catching Covid-19, In a sample of the Peruvian population; which concludes, information is one of the factors that influences the behavior of a pandemic throughout a territory.

In accordance with different research [10], the metropolitan population is one of the most important infection predictors, populations with larger metropolitan areas, suffer higher infection indices and mortality rates. However, the opposite is true in regard to the internal level of the states or departments of the different countries, and the mortality rate, possibly due to bigger adherence to the politics and practices of social distancing and better health services. These findings suggest that connectivity is a greater factor towards Covid-19 propagation than density.

Furthermore, other research [11] a study was conducted regarding the correlation between the demographic variables and the lethality of Covid-19 in OCDE countries, considering them homogeneous. Result show that the mortality and lethality of Covid-19 are not associated with demographic variables, the country’s budget directed towards sanitary services. On the other hand, they are directly related with the number of medical professionals and tests. The results are outstanding since these sorts of analysis haven’t been taken into account to adjust the safety measures and politics adopted by various countries in order to counter the Virus’ spread.

For Colombia specifically, the Ministry of health, in conjunction with Jhon Hopkins University, carried out a validation of the best indicators to trace Covid-19’s behavior, which feature [12]:

In the previous indicators, a factor tied to population and populational density wasn’t taken into consideration, as a purpose for this study these variables, and their association with the effects of the pandemic, are considered.

On the other hand, it establishes the necessity to group together information that allows the development of corresponding analysis. Precedents of Clustering can be found in Villavicencio [13], where it’s applied, not to find an algorithm, instead to split logically the data in groups that could be processed.

II. Metodology

The study type is explorative since it searches to answer a hypothesis regarding the behavior of data and phenomena. In accordance with the information type, the investigation identifies itself as quantitively, although certain qualitative variables were used to develop further analysis of the results.

As for the data collection and organization, it was approached through four phases: data recompilation, selecting variables for analysis and constructing regression models to determine the correlation between data.

.

A. Data collection

For the first phase the Systematic Literature Review (SLR) was used, making use of Google® search engines to obtain data for secondary sources regarding the pandemic’s behavior. To guarantee homogeneous data of various countries, data was compiled during the period between 10-03-2020 and 15-08-2020, making use of the search equation CODIV-19 AND America AND deaths AND cases AND CIFRAS OR stadistics. The recompiled information was double checked in order to delete any unreliable or repeated sources, guaranteeing the validity of the sources for the analyzed territories.

B. Clustering

In the second phase the data went through a clustering process by region classifying them under North American cluster, Central American cluster, and South American cluster. To accomplish this a relationship was established between Infected and three independent variables, those being Number of inhabitants, PIB per Capita and the flights per every country. With that information the coefficients of correlation and equations were determined, with this construct it was determined the levels of infection that could be reached by the different regions, with an average percentage to determine the number of deaths.

C. Variable selection and correlation

For the third phase, variable selection and correlation, the Excel® tool was used in order to organize the data and the correlation of the different variables, such as the levels of infection, PIB per capita, number of inhabitants per region, thousands of flights using multiple regression. Based on the results an equation for the infection forecast was found, taking into account correlation indicators between different variables. The methods used to find the correlations between the selected variables were lineal regression, multiple regression, specifying dependent and independent variables in both cases.

D. Construction of Regression models

In the fourth and final phase, the regression models were developed based on the correlation of the data, and through that an infection forecast was determined in every region.

III. Results

For the study of the correlation of scalar demographic variables the following were taken into account:

In terms of Covid indicators, the following were considered:

Throughout development it was established to only use sovereign countries since various states associated or colonies of other countries were present, such as Puerto Rico being an associate with the United States or Curazao being dependent of Holland. Data for the development of the study was taken on the 15th of August 2020.

For development we started with the correlation between infected per country and the population (Table 1), the population was identified as the independent variable while the number of infected per country was identified as the dependent variable, linear regression was used for the correlation.

Table 1. Number of inhabitant’s vs Infected per country.

Country

Infected

Population

% Infected

Chile

383 902

18 650 114

2.1%

United States

5 529 750

329 995 528

1.7%

Panamá

79 402

5 005 246

1.6%

Brazil

3 317 832

211 823 665

1.6%

Peru

516 296

33 050 325

1.6%

Colombia

456 689

50 220 856

0.9%

Bolivia

97 950

11 969 649

0.8%

Dominican Republic

85 545

10 606 865

0.8%

Argentina

289 100

45 030 748

0.6%

Ecuador

100 688

17 080 778

0.6%

Honduras

49 467

8 893 259

0.6%

Costa Rica

27 737

6 172 543

0.4%

Guatemala

62 313

15 289 958

0.4%

Mexico

511 369

128 166 749

0.4%

El Salvador

22 314

6 356 670

0.4%

Canada

121 889

37 411 590

0.3%

Paraguay

22 314

7 612 812

0.3%

Cuba

3 229

1 179 995

0.3%

Haiti

7 810

11 485 800

0.1%

Uruguay

1 421

3 286 314

0.0%
Source: [15].

From the previous table we can observe that the countries with higher percentage of infection to date are Chile, United States, Panama, Brazil, Peru, which, especially in the case for US and Brazil, belonging to countries with a high number of inhabitants. Taking into account that the variables being used are scalar, a simple regression is a tool tat can offer information about the data’s behavior.

Previously an analysis regarding the data was made, where we can observe the following behavior for the every one of the following variables (Fig. 1):

Fig. 1. Data behavior involving infected.
Source: Authors.

Considering the lack of collinearity in regards to the variables, the results of the regression are as follows (Table 2):

Fig. 2. Lineal Regression Infected vs Population.
Source: Authors.
Table 2. Infected Lineal Regression results vs Lineal population.

Correlation Coefficient

0.963633972

Determination Coefficient R^2

0.928590432

R^2 adjusted

0.927162241

Common error

238896.8553

Observations

52
Source: Authors.

Column F, from Table 3, showcases the statistic value of testing, and the probability column showcases the lowest level of significance to reject the null hypothesis of equal means. If probability is lower than 0.05 it rejects the null hypothesis, in this case it’s higher, thus we accept that the infected are associated to the volume of the population (number of in­habitants).

Table 3. Infected variance analysis vs Population.

Variance Analysis

 

Degree of freedom

F

Critical value of F

Regression

1

650,1862

2,56828E-30

Remains

50

Total

51

 

 

 

Coefficients

Probability

Below 95%

Interception

–189350.2

0.053446622

–139902.757

Variable X 1

0.01522

2.56828E-30

0.014025368
Source: Authors.

The equation for the regression (1) is represented in the following manner:

Within the results obtained its possible to observe a strong correlation of the data due to the correlation of 0.964, however the critical value of F is rather small which validates the use of the equation as a forecast model. Even though the relation is high there can exist other variables that can influence this result, thus the relation between populational density and the level of infection is presented, with the independent variable being the population density and the infection.

Then, the connection between infections originating from Covid 19 and the populational density, the data is organized upward based on the cases of infection (Table 4).

Table 4. Populational density vs Infections.

Countries

Covid

Density

United States

5 529 750

34.5

Brazil

3 317 832

24.1

Peru

516 296

23.9

Mexico

511 369

62.7

Colombia

456 689

41.3

Chile

383 902

23.3

Argentina

289 100

15.7

Canada

121 889

3.5

Ecuador

100 688

56.7

Bolivia

97 950

9.9

Dominican Republic

85 545

217.9

Panama

79 402

49.1

Guatemala

62 313

139.4

Honduras

49 467

79

Venezuela

31 381

33.7

Costa Rica

27 737

95.3

El Salvador

22 314

292.6

Puerto Rico

10 730

389.76

Paraguay

9 022

16.8

Haiti

7 810

377.8

Nicaragua

4 115

45.7

Cuba

3 229

100.8

Surinam

2 838

3.57

Uruguay

1 421

19

Trinidad and Tobago

426

238
Source: [15].

In Table 5, the values obtained from the variance are showcased, note how the correlation coefficient Is below 0.5, which indicates that there is no correlation between populational density and infections caused by Covid 19, a possible explanation might involve mass prevention of the virus would result in avoidance of its infection.

Table 5. Density regression results vs Covid 19 infections.

Regression statistics

Multiple Correlation Coefficient

0.173537668

Determination Coefficient R^2

0.030115322

R^2 adjusted

0.009909391

Common Error

897.391

Observations

50
Source: Authors.

The correlation’s calculation has a result of 0.2 which implies a low value of correlation between the populational density and the level of infection, thus it’s not viable to develop a regression equation (Table 6).

Table 6. Density variance analysis vs Covid 19 Infection.

Variance analysis

 

Degree of freedom

Critical value of F

Regression

1

0.228114531

Remains

48

Total

49

 

Coefficients

Below 95%

Interception

357883.9353

31958.53308

Variable X 1

–597.577067

–1581.75309
Source: Authors.

Following, the contents of Table 7 establish a connection between ICRS and the amount of infections per country, the ICRS defined as the compound index of health results. The Bloomberg index of OMS is present

Table 7. Infection statistics vs ICRS.

Countries

Infection

Icrs

Canada

121 889

100

United States

5 529 750

92

Uruguay

1 421

81

Costa Rica

27 737

79

Chile

383 902

76

Cuba

3 229

76

Argentina

289 100

72

México

511 369

63

Panama

79 402

62

Paraguay

9 022

57

Brazil

3 317 832

56

Colombia

456 689

56

Ecuador

100 688

56

Dominican

Republican

85 545

50

Honduras

49 467

50

El Salvador

22 314

50

Guatemala

62 313

44

Peru

516 296

42

Bolivia

97 950

24

Haiti

7 810

0
Source: [16].
Table 8. Infection regression results vs ICRS.

Regression Statistics

Multiple Correlation Coefficient

0.286191896

Determination Coefficient R^2

0.081905801

R^2 adjusted

0.030900568

Common Error

1351213.604

Observations

20
Source: Authors.
Table 9. Infection variance analysis vs ICRS.

Variance Analysis

 

Degree of freedom

F

Critical value of F

Regression

1

1.605831324

0.221229133

Remains

18

 

 

Total

19

 

 

 

Coefficients

Probability

Below 95%

Interception

–435872.4487

0.618195746

–2241464.467

Variable X 1

17193.23269

0.221229133

–11311.53562
Source: Authors.

Within the analysis we can observe that coefficient R adjusted is 0.081905 which implies that the connection between infection variables and the health quality index is null. Statistic wise countries such as the US, Canada and Brazil which possess a high index of health are some of the countries with the biggest amount of cases of infection, meanwhile countries with lower indices also present lower infection rates. It’s important to note that the number of tests provided by the country might distort the statistics.

Then, it proceeds to form a correlation of the variables infection vs people transported by aerial transports, showcased in the following Table 10 are the results of the correlation:

Table 10. Infection statistics vs Flights (thousands) per country.

Countries

Infected

Flights (Thousands)

United States American

5 529 750

889 022

Canada

121 889

121 889

Brazil

3 317 832

102 109

Mexico

511 369

64 529

Colombia

456 689

33 704

Chile

383 902

19 519

Argentina

289 100

18 081

Peru

516 296

17 758

Panama

79 402

12 939

Ecuador

100 688

5 365

Bolivia

97 950

4 122

El Salvador

22 314

2 545

Costa Rica

27 737

1 948

Uruguay

1 421

563

Cuba

3 229

561

Paraguay

22 314

560

Honduras

49 467

411

Guatemala

62 313

146

Dominican Republic

85 545

111

Haiti

7 810

12
Source: [15].
Table 11. Infection regression results vs Flights by Thousand.

Regression Statistics

Multiple Correlation Coefficient

0.892087715

Determination Coefficient R^2

0.795820491

R^2 adjusted

0.784477185

Common Error

637081.2013

Observations

20
Source: Authors.
Table 12. Infection variance analysis vs Flights by thousand.

Variance Analysis 

 

Degree of freedom

Critical value of F

Regression

1

1.27E-07

Remains

18

 

Total

19

 

Coefficients

Below 95%

Interception

181932.3743

–133918.4567

Variable X 1

6.210669634

4.652873464
Source: Authors.

It is noted a correlation coefficient of 0.89, which indicates a strong connection between air transport and infection, we can conclude the following; aerial transport of passenger might be a variable for high levels of infections, which justifies one of the objectives for various governments was to cancel national and international flights, mainly to avoid further spread of the virus. We can define it with (2):

A. Multiple regression analysis

for the development of the multiple regression analysis it was decided to combine the PIB per capita from 2019 (economic type variable) with the number of inhabitants and the level of infection, with the purpose to establish if there exists any connection from an economic standpoint. Below is the table with the results of the analysis.

Table 13. Infection statistics vs PIB per capita and number of inhabitants.

Countries

Infected

Pib Per Capita

Population

USA

5 529 750

65 118

329 995 528

Brazil

3 317 832

8 717

211 823 665

Peru

516 296

6 977

33 050 325

México

511 369

9 863

128 166 749

Colombia

456 689

6 432

50 220 856

Chile

383 902

14 896

18 650 114

Argentina

289 100

10 006

45 030 748

Canada

121 889

46 194

37 411 590

Ecuador

100 688

6 183

17 080 778

Bolivia

97 950

3 552

11 969 649

Republican Dominican

85 545

8 282

10 606 865

Panamá

79 402

15 731

5 005 246

Guatemala

62 313

4 620

15 289 958

Honduras

49 467

2 574

8 893 259

Costa Rica

27 737

12 238

6 172 543

El Salvador

22 314

4 187

6 356 670

Paraguay

22 314

5 415

7 612 812

Haiti

7 810

755

11 485 800

Cuba

3 229

8 821

1 179 995

Uruguay

1 421

16 190

3 286 314
Source: [15].
Table 14. Multiple regression Infected vs population and PIB per capita .

Regression Statistics

Multiple Correlation Coefficient

0.96666375

Determination Coefficient R^2

0.9344388

R^2 adjusted

0.92672572

Common Error

371470.177

Observations

20
Source: Authors.
Table 15. Infection variance analysis vs population and PIB per capita .

Variance Analysis

 

Degree of freedom

Critical value of F

Regression

2

8.74E-11

Remains

17

 

Total

19

 

Coefficients

Below 95%

Interception

–219248.241

–449109.8121

Variable X 1

6.1405665

–8.980253

Variable X2

0.01511054

0.012297051
Source: Authors.

According to the data we can observe that the correlation coefficient is 0.965 which implies a strong correlation of the studied variables. There is a higher number of cases in countries that present a high economic level, such being the case of US and Brazil. The mathematical model for the multiple regression is the following:

Y being = Dependent variable number of cases

X1 = PIB Per Capita value per country

X2 = Number of inhabitants per country.

Next, the multiple regression possesses a dependent variable (Infections) and three independent variables, those being PIB per Capita, Population and number of flights (by thousand).

Table 16. Infection Statistics vs PIB per Capita, Number of inhabitants and Flights by thousand.

Countries

Infection

Pib Per capita

Population

Flights (thousand)

USA

5 529 750

65 118

329 995 528

5 529 750

Mexico

511 369

9 863

128 166 749

64 529

Canada

121 889

46 194

37 411 590

121 889

Dominican Republic

85 545

8 282

10 606 865

111

Panama

79 402

15 731

5 005 246

12 939

Honduras

49 467

2 574

8 893 259

411

Guatemala

62 313

4 620

15 289 958

146

El Salvador

22 314

4 187

6 356 670

2 545

Haiti

7 810

755

11 485 800

12

Cuba

3 229

8 821

1 179 995

561

Costa Rica

27 737

12 238

6 172 543

1 948

Brazil

3 317 832

8 717

211 823 665

102 109

Peru

516 296

6 977

33 050 325

17 758

Colombia

456 689

6 432

50 220 856

33 704

Chile

383 902

14 896

18 650 114

19 519

Argentina

289 100

10 006

45 030 748

18 081

Ecuador

100 688

6 183

17 080 778

5 365

Bolivia

97 950

3 552

11 969 649

4 122

Paraguay

22 314

5 415

7 612 812

560

Uruguay

1 421

16 190

3 286 314

563
Source: Authors.

The results from the regression are as follows (Table 17):

Table 17. Infection multiple regression results vs PIB per Capita, Population and Flights by thousand.

Regression Statistics

Multiple Correlation Coefficient

0.97506441

Determination Coefficient R^2

0.9507506

R^2 adjusted

0.94151633

Common Error

331868.067

Observations

20
Source: Authors.
Table 18. Multiple regression variance analysis vs population, PIB per capita and flights by thousand.

Variance Analysis

 

Degree of freedom

Critical value of F

Regression

3

1.13E-10

Remains

16

 

Total

19

 

 

Coefficients

Below 95%

Interception

–51321.308

–309176.9448

Variable X 1

–5.6637407

–23.05349225

Variable X 2

0.01286134

0.009595049

Variable X 3

0.30823405

0.024385496
Source: Authors.

The multiple correlation coefficient is 0.97506441 which indicates a direct connection between the variables and establishes the following linear equation (4):

B. Clustering process

A grouping technique was utilized to explore the data attempting to maximize the similarities between the elements of the class and minimize the similarities between groups. Throughout the clustering process the countries were classified by region: North American, Central American (plus the Caribbean islands) and South American.

1) North America Cluster

Table 19. Infection statistics Vs PIB Per capita and Cluster North America population.

Countries

Infected

Pib Percapita

Población

Flights in Thousands

USA

5 529 750

65 118

329 995 528

5 529 750

Mexico

511 369

9 863

128 166 749

64 529

Canada

121 889

46 194

37 411 590

121 889
Source: Authors.

For this cluster the Infection (y), PIB per capita (X1) and Population (X2) were used, and yielded the following results (Table 20):

Table 20. Infection regression results vs PIB Per capita and Cluster North America population.

Regression Statistics

Multiple Correlation Coefficient

1

Determination Coefficient R^2

1

R^2 adjusted

65535

Common Error

0

Observations

3
Source: Authors.

From the analysis of variance, we obtained a strong correlation coefficient, above 0.5. With R2, we can determine that the model adjusts itself to the data (Table 21).

Table 21. Infection variance analysis Vs PIB Per capita and Cluster North America population.

Variance Analysis

 

Degree of freedom

Critical value of F

Regression

2

0

Remains

0

 

Total

2

 

 

Coefficients

Below 95%

Interception

–1905568.2

–1905568.2

Variable X 1

30.5196402

30.5196402

Variable X 2

0.01650913

0.01650913
Source: Authors.

From the results we can gather a correlation of 1, which indicates a direct connection between the variables, resulting in the following equations (5):

In accordance with the multiple regression equation North America’s cluster reached an infection level of 9 721 831 inhabitants.

2) Central American (& Caribbean Islands) Cluster

The following table selects countries of Central America (& the Caribbean Islands), with the purpose to analyze the infection and establish if there is any correlation with nearby territories (Knn) (Table 22).

Table 22. Infection statistics vs PIB per Capita, Population and flights by thousand in Central America.

Countries

Infected

Pib Percapita

Population

Flights in Thousands

Dominican Republic

85 545

8282

10606865

111

Panama

79 402

15731

5005246

12939

Honduras

49 467

2574

8893259

411

Guatemala

62 313

4620

15289958

146

El Salvador

22 314

4187

6356670

2545

Haiti

7 810

755

11485800

12

Cuba

3 229

8821

1179995

561

Costa Rica

27 737

12238

6172543

1948
Source: Authors.

For the Central American cluster there was one independent variable (infection) and three dependent variables, those being X1 (PIB per Capita), X2 (population) and X3 (Flights per year by thousand unit). The results from the lineal regression are as follows:

Table 23. Infection regression results vs PIB per Capita, Population, and Flights by thousand unit in Central America.

Regression Statistics

Multiple Correlation Coefficient

0.7956783

Determination Coefficient R^2

0.6331039

R^2 adjusted

0.3579319

Common Error

25386.36

Observations

8
Source: Authors.

In accordance with the Value obtained from F, the variance of the factors doesn’t appear to present any significant effect in the overall result (Table 24).

Table 24. Infection variance analysis vs PIB per Capita, Population and flights by thousand in Central America.

Variance Analysis

 

Degree of freedom

Critical value of F

Regression

3

0.219018715

Remains

4

 

Total

7

 

 

Coefficients

Below 95%

Interception

–36702,63

–132478.9296

Variable X 1

3,8556841

–4.4163367

Variable X 2

0,0057832

–0.0014193

Variable X 3

1,8749147

–6.866583575
Source: Authors.

Within the correlation coefficient we can observe a value of 0.79 which indicates a direct connection of medium value. The multiple regression equation is defined as (6):

According to the forecast resulting from the equation, an infected level of 668140 for the Central American & Caribbean region is estimated.

3) South American Cluster

Table 25. Infection statistics vs PIB per Capita, Population and flights by thousand in South America Cluster.

Countries

Infected

Pib Percapita

Population

Flights Thousands

Brazil

3 317 832

8 717

211 823 665

102 109

Peru

516 296

6 977

33 050 325

17 758

Colombia

456 689

6 432

50 220 856

33 704

Chile

383 902

14 896

18 650 114

19 519

Argentina

289 100

10 006

45 030 748

18 081

Ecuador

100 688

6 183

17 080 778

5 365

Bolivia

97 950

3 552

11 969 649

4. 122

Paraguay

22 314

5 415

76 12 812

560

Uruguay

1 421

16 190

3 286 314

563
Source: Authors.

On the following table are the results from the lineal regression (Table 26).

Table 26. Infection regression results vs PIB per Capita, Population and flights by thousand in South America Cluster.

Regression Statistics

Multiple Correlation Coefficient

0.989070992

Determination Coefficient R^2

0.978261427

R^2 adjusted

0.965218284

Common Error

194998.2521

Observations

9
Source: Authors.
Table 27. Infection variance analysis vs PIB per Capita, Population and flights by thousand in South America Cluster.

Variance Analysis

 

Degree of freedom

Critical value of F

Regression

3

0.000140834

Remains

5

 

Total

8

 

 

Coefficients

Below 95%

Interception

–217637.9657

–643803.084

Variable X 1

9.599893018

–33.57114109

Variable X 2

0.014013166

–0.003010874

Variable X 3

3.990495374

–30.68113114

The correlation coefficient is 0.9807 which implies a strong and direct connection of the correlation between variables, and establishes the following regression equation (7):

According to the established forecast based on the multiple regression equation South America will reach a level of infection estimated to be around 6 926 041 people (Fig. 3).

Fig. 3. Infection level per cluster
Source: Authors.

When adding up the cluster it results in an estimated infection level of 17 316 012 inhabitants for the American continent and deaths averaging 3% resulting in a value of 519 480 deaths.

According to Fig. 4 we can observe that by the Month of August (2020) the American countries represent the largest figures of COVID confirmed cases, and countries present similar figures in regards to the infection rates in concordance with the clusters previously established (North America, Central America and South America).

Fig. 4. World Health Organization’s situation reports. Johns Hopkins University CSSE, The Centers for Disease Control and Prevention, New York Times, CNBC.
Source:

IV. Conclusions

The findings suggest that connectivity plays a larger role in the propagation of the COVID-19 pandemic than populational density. Large metropolitan areas with a high number of spaces deeply connected via the economy, social and transport are the most vulnerable.

As stated by the study there appears to be a strong correlation of variables such as the amount of inhabitants, the amount of flights and the economic activity reflected by the PIB per capita of the continent’s countries, clustering the data by regions, which implies that the measures taken in regards to the border closure resulted in an alternative for controlling the propagation at a global level.

Indicators of health service quality and population density of the countries were taken, but no correlation of the data was defined, which implies that several countries with low levels of ICRS did not perform enough tests to define the number of infected people.

The economic level and the amount of flights had a correlation regarding the level of infection. This is reflected in the first measures used by the governments; this being quarantining the area (such as prohibiting travel and certain economic activities) to minimize the infection rate.

Studies found in literature use other statistical methods to develop the correlation with the variables [14], [9] and [11], some of which stand out such as the use of R-square values, Spearman correlation coefficient an the statistic analysis ANOVA, which establishes a need to develop other correlational analysis with techniques that allow the passing of time, and thus show estimates about the variable’s behavior in the pandemic’s future.

Variables such as population, PIB per Capita and flights generate a high relation with the infection level, but it’s necessary to point out that this doesn’t imply the causality of the infection since correlational analysis don’t have the reach necessary to determine that information.

Financing

This investigation belongs to the research project known as "Diseño de una metodología para la gestión de flujo de pacientes en servicios de salud basado en estrategias lean y patrones de proceso de negocios", Answers to the process betterment line belonging to the Organizational processes investigation group (GIPO) in business engineering approved by the investigation committee from the Industrial Engineering faculty of the San Tomas University.

References

[1] Real Académica Española, Diccionario de la lengua española. MD, ES: RAE, Oct. 2019. Disponible en https://dle.rae.es/

[2] T. N. Jilani, R. T. Jamil & A. H. Siddiqui, “H1N1 Influenza,” in, StatPearls [Internet]. Treasure Island, FL: StatPearls Publishing, 2020.

[3] S. Hamidi, S. Sabouri & R. Ewing, “Does Density Aggravate the COVID-19 Pandemic?,” J Am Plan Assoc, vol. 86, no. 4, pp. 495–509, 2020. https://doi.org/10.1080/01944363.2020.1777891

[4] B. Ather, T. M. Mirza & P. F. Edemekong, “Airborne Precautions,” in, StatPearls [Internet]. Treasure Island, FL: StatPearls Publishing, 2020.

[5] P. S. Peixoto, D. Marcondes, C. Peixoto & S. M. Oliva, “Modeling future spread of infections via mobile geolocation data and population dynamics. An application to COVID-19 in Brazil,” PLoS One, vol. 15, no. 7, Jul. 2020. https://doi.org/10.1371/journal.pone.0235732

[6] D. Rosselli, “Epidemiología de las pandemias,” Rev Medicina, vol. 42, no. 2, pp. 168–174, Jul. 2020. Disponible en https://revistamedicina.net/ojsanm/index.php/Medicina/article/view/1511

[7] OPS, “Por qué es importante el desglose de datos durante una pandemia,” paho.org, 2020. Recuperado de https://www.paho.org/ish/images/docs/Data-Disaggregation-Factsheet-Spanish.pdf

[8] OPS, “Pandemia de COVID-19: estadísticas sobre el acceso a la BVS y el alcance de la cooperación técnica de BIREME, Boletín Bireme, no. 42, 2020. Disponible en https://boletin.bireme.org/2020/04/01/pandemia-de-covid-19-estadisticas-sobre-el-acceso-a-la-bvs-y-el-alcance-de-la-cooperacion-tecnica-de-bireme/

[9] P. G. Ruiz Mamani, W. C. Morales-García, M. White & M. S. Marquez-Ruiz, “Properties of a scale of concern for COVID-19: Exploratory analysis in a Peruvian sample,” Med Clin, vol. 255, no. 12, pp. 535–537, Dec. 2020. https://doi.org/10.1016/j.medcli.2020.06.022

[10] S. Hamidi, R. Ewing & S. Sabouri, “Longitudinal analyses of the relationship between development density and the COVID-19 morbidity and mortality rates: Early evidence from 1,165 metropolitan counties in the United States,” Heal Place, vol. 64, no. 2, pp. 102378–102378, Jul. 2020. https://doi.org/10.1016/j.healthplace.2020.102378

[11] A. Medeiros de Figueiredo, A. Daponte, D. C. Moreira Marculino de Figueiredo, E. Gil-García & A. Kalache, “Case fatality rate of COVID-19: absence of epidemiological pattern,” Gac. Sanit, vol. 35, no. 4, pp. 10–12, 2020. https://doi.org/10.1016/j.gaceta.2020.04.001

[12] República de Colombia. MinSalud, “Análisis de la epidemia de covid-19 en el país,” Boletín de Prensa No 223 de 2020, 2020. Disponible en https://www.minsalud.gov.co/Paginas/Analisis-de-la-epidemia-de-covid-19-en-el-pais.aspx

[13] F. Velásquez & G. D. Sosa, “Aplicación de Técnicas de Clustering en Sonidos Adventicios para Mejorar la Interpretabilidad y Detección de Estertores,” INE CUC, vol. 11, no. 1, pp. 53–62, 2015. Disponible en http://revistascientificas.cuc.edu.co/index.php/ingecuc/article/download/366/2015105

[14] I. F. Meza, A. E. Herrera & L. G. Obregón, “Determinación experimental de nuevas correlaciones estadísticas para el cálculo del coeficiente de transferencia de calor por convección para placa plana, cilindros y bancos de tubos,” INGECUC, vol. 13, no. 2, pp. 9–17, 2017. https://doi.org/10.17981/ingecuc.13.2.2017.01

[15] GBM, Banco Mundial de la Salud, datos.bancomundial.org, 2019. Disponible en https://datos.bancomundial.org/indicator/NY.GDP.PCAP.CD

[16] ACCH, Asociación Colombiana de Clínicas y hospitales, achc.org, 2019. Disponible en http://achc.org.co

Elisa del Carmen Navarro-Romero. Industrial engineering from the Universidad del Norte (Colombia). Magister in industrial engineering with emphasis on Organizational management, from the Universidad Distrital Francisco de Paula Santander (Colombia); teacher in investigational formation. 12 years of experience with business from the real sector, with positions associated with betterment processes, budget, guaranteeing quality and logistics, in Financial companies such as the Grupo Aval Holding; Service companies such as CUMANDES- Cummins de los Andes, and production companies such as Oleoflores SA, and other of the sort.
https://orcid.org/0000-0002-1825-0097

Óscar Mauricio Gelves-Alarcón. Full time teacher at the faculty of Industrial Engineering at the Universidad Militar Nueva Granada (Colombia), dedicated to educational spaces involving modern manufacture and economical engineering, with schooling in Industrial Engineering at the District University, specialized in production engineering. Master’s degree in industrial directed engineering at the Universidad de Buenos Aires (Argentina) and with experience in education in various private universities in the fields of Production, investigation and logistics, with experience in the real sector as coordinator of planning at Brightstar and coordinator of logistics at Eco de los Andes. https://orcid.org/0000-0003-0557-775X

Natalia García-Corrales. Industrial engineering at the Universidad del Norte (Colombia), Magister in administration and specialist in project management at the Universidad Pontificia Bolivariana (Colombia), and specialist in health & security management at the ECCI university. Currently, professor at the faculty of industrial engineering at the Pontificia Universidad Bolivariana (Colombia), and leader of strategic management of the Environment & Technology Foundation. 12 years of experience with management system articulation with management and strategic planning, particularly with service companies. Organizational advice in organizational transformation processes. https://orcid.org/0000-0002-6866-2401

.
© The author; licensee Universidad de la Costa - CUC.
INGE CUC vol. 17 no. 1, pp. 285-302. Enero - Junio, 2021
Barranquilla. ISSN 0122-6517 Impreso, ISSN 2382-4700 Online
.