Quantitative Methods in Demography: Methods and Related Applications in the Covid-19 Era (The Springer Series on Demographic Methods and Population Analysis, 52) 3030930041, 9783030930042

Citation preview

The Springer Series on Demographic Methods and Population Analysis 52

Christos H. Skiadas Charilaos Skiadas   Editors

Quantitative Methods in Demography Methods and Related Applications in the Covid-19 Era

The Springer Series on Demographic Methods and Population Analysis Volume 52

Series Editor Scott M. Lynch, Department of Sociology, Duke University, Durham, NC, USA

This series is now indexed in Scopus. In recent decades, there has been a rapid development of demographic models and methods and an explosive growth in the range of applications of population analysis. This series seeks to provide a publication outlet both for high-quality textual and expository books on modern techniques of demographic analysis and for works that present exemplary applications of such techniques to various aspects of population analysis. Topics appropriate for the series include: • • • • • • • • • • •

General demographic methods Techniques of standardization Life table models and methods Multistate and multiregional life tables, analyses, and projections Demographic aspects of biostatistics and epidemiology Stable population theory and its extensions Methods of indirect estimation Stochastic population models Event history analysis, duration analysis, and hazard regression models Demographic projection methods and population forecasts Techniques of applied demographic analysis, regional and local population estimates and projections • Methods of estimation and projection for business and health care applications • Methods and estimates for unique populations such as schools and students. Volumes in the series are of interest to researchers, professionals, and students in demography, sociology, economics, statistics, geography and regional science, public health and health care management, epidemiology, biostatistics, actuarial science, business, and related fields.

Christos H. Skiadas • Charilaos Skiadas Editors

Quantitative Methods in Demography Methods and Related Applications in the Covid-19 Era

Editors Christos H. Skiadas ManLab Technical University of Crete Chania, Greece

Charilaos Skiadas Dept of Mathematics/Comp Science Hanover College Hanover, Indiana, USA

ISSN 1389-6784 ISSN 2215-1990 (electronic) The Springer Series on Demographic Methods and Population Analysis ISBN 978-3-030-93004-2 ISBN 978-3-030-93005-9 (eBook) https://doi.org/10.1007/978-3-030-93005-9 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2022, corrected publication 2022 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

We thank Anthi for continuous encouragement and support in developing this volume

Preface

Demography: Demographic Data. Data collection was a primary issue for countries and societies since their first steps toward social organization. Families, births and deaths, injuries and diseases, property occupation, villages and cities, small territories and huge landscapes, and insurance and risk covering came as aftereffects of the old and modern societal needs. As data collection became more accurate and important for an organized society, data analysis methods and techniques became important tools. Prior to that, the invention or introduction of writing, reading, and numbering took societies to the next step of social organization. As the writing, reading, and numbering tools evolved, data storage and analysis required organization; this necessitated the formation of registry offices, notaries, and other organizational structures for the benefit of society, especially when money in any type and form was invented and introduced. The main part of these forms, quantifying Demography, introduced and accepted hundreds and even thousands of years ago and are so strong that form the basis of our societies. It looks like that writing, reading, and numbering dominate our societies and our everyday life. Some interesting features of numbers and numbering were discovered and developed during the past centuries. The discovery of algebra, number theory, geometry, and other mathematical forms and tools opened new horizons to that we call as data analysis in demography and population studies. Probability, an important element of demographic and population analysis, was developed late after the Middle Ages as a scientific tool (Girolamo Gardano, 1564). It remained to explore a good data series for births and deaths to form the Life Table ((J. Graunt, 1665) and (E. Halley, 1693)). Pierre de Fermat and Blaise Pascal (1654) and Christiaan Huygens (1657) had done important work on the scientific basis of probability supporting quantitative demographic analysis. It took a few centuries for the advent of actuarial science, until 1825, when Gompertz proposed the famous model for insurance estimated by the actuaries. From then onward, actuarial science developed. Important for the use of the Gompertz model was the introduction of the Tables of Logarithms. The famous vii

viii

Preface

Verhulst model (1838), also known as the Logistic, was introduced to estimate and forecasts population size. Around 1900, the Runge–Kutta method for estimating differential equations provided new approaches of using data analysis tools in demography and population analysis. During the same period, Luis Bachelier proposed the theory of speculation in his PhD dissertation, the theory of stochastic processes that came to be an important part of many developments in the twentieth century. The stochastic theory is important to estimate the health state of a population as well as the healthy life years lost and the healthy life expectancy estimation. Jacques Janssen and Christos Skiadas (1995) introduced stochastic modeling to life table data sets with applications to France and Belgium. Later on, stochastic modeling was further developed and applied to find the healthy life years lost and the healthy life expectancy (see Skiadas C.H., Skiadas C. (2020) Preliminary Notes. In: Skiadas C.H., Skiadas C. (eds) Demography of Population Health, Aging and Health Expenditures. The Springer Series on Demographic Methods and Population Analysis, vol 50. Springer, Cham. https://doi.org/10.1007/978-3030-44695-6_1). The stochastic theory, even though certain difficulties arise in theory and applications, is a very important tool, thus providing demography and population analysis a methodology to do stochastic simulations to make estimates for the healthy life years lost and the healthy life expectancy (HLE). The latter confirmed with successful comparisons with the HALE estimates from the World Health Organization (WHO). According to WHO: HALE is an estimate of the average number of years that a person can expect to live in “full health” by taking into account years lived in less than full health due to disease and/or injury. Following this definition, we have explored the average measures of mortality from a life table. The averaging form xmx /sum(m0 :mx ) was proven to be very successful in providing the healthy life years lost (HLYL) and then the HLE = LEHLYL, where (LE) is the life expectancy. The importance of this measure coming directly from life tables is that we estimate the HLE in the past as far as life tables are provided. No information on the health status of a population is required. The influence of diseases and injuries and their aftereffects are included in the averaging formula. The HLE years resulting from this method are similar to HALE. For further information, see the above references at https://doi.org/10.1007/978-3-030-446956_1. However, the quantitative methods in demography and population analysis and health state estimates were mainly developed with the aid of informatics and computing and received special attention in the last decades of the twentieth century. Further developments and applications continue in the first part of the twenty-first century. These methods already have ushered considerable changes in various scientific fields and of course in estimating vital aspects of demography, health, insurance, social development, life expectancy, and population changes. Mortality, population aging, and data analysis are further developed, and the tools used are friendly to the end user. People are “ready” to understand and apply the new

Preface

ix

tools. A fast-growing literature, both theoretical and applied, and many “computer packages” and visual observation are ready to support the very many applications. Interdisciplinary works are an important task of the new era, and new fields like data science and big data analysis, important to handle large data sets familiar in international studies in demography, health sciences, and population, are developed. The twenty-first century is already characterized by an optimistic way of data analysis approaches. Large number of people are educated and trained to collect and store data sets. Large and expanded networks disseminate information. Demography, health, and population studies have benefited the most, and more developments are in underway. This book provides quantitative material in the Covid-19 era exploring important issues in demography, population studies, and health state estimates along with the healthy life expectancy calculation in the past for the World Health Organization members. Even more, mortality and survival data analysis was done, followed by various methodologies performed and applications in several issues, including society, economy, insurance, and classification, important in demographic studies. Quantitative and applied methodologies in demography are included as the use of full life tables for developing countries to estimate the healthy life years lost. This book is a valuable guide for researchers, theoreticians, and practitioners in various scientific fields.

This Book Includes Five Parts – The first part with six chapters is related to the Covid-19 pandemic. The death/cases development, chaotic forms present, socioeconomic and health issues, and life expectancy loss are targets of the chapters. – The second part on global health and longevity includes four chapters on the estimation of healthy life years lost and the healthy life expectancy in the far past and the global health and longevity: an analysis of post-World War II data in the WHO countries. – The third part on mortality–survival includes ten chapters related to mortality and survival in the Czech Republic, Greece, Nigeria, Brazil, and Russia. – The fourth part, with four chapters on special methods, includes America’s Zika virus and its similarities with African and Asian lineages, and a relative entropy measure in labor market outcomes by educational attainment. Also, the intergenerational educational mobility in European countries and a different approach in grouping European countries in terms of mortality. – The fifth part includes nine chapters on insurance, risk and health risk, social change indicators, the performance of socio-economic classification in Europe, life expectancy and the Chinese retirement plan, and life expectancy and population loss due to mental disorder in Russia.

x

Preface

We thank all the contributors of this book, the chapter authors, and of course the Springer team and Evelien Bakker for help, guidance, and support. Athens, Greece Hanover, IN, USA

Christos H. Skiadas Charilaos Skiadas

Contents

Part I 1

2

3

4

5

6

COVID-19 Studies

Reaction to COVID-19 Pandemic: An Evaluation of Pandemic Management Around the World . . . . . . .. . . . . . . . . . . . . . . . . . . . Yiannis Dimotikalis and Christos H. Skiadas Effects of the Covid-19 Pandemic in the Area of Tension Between the Economy and Climate Change: A Case Study at Rural and City District Level in Southern Germany .. . . . . . . . . . . . . . . Moneim Issa and Rolf Bergs Predicting the Second Wave of COVID-19 Pandemic Through the Dynamic Evolving Neuro Fuzzy Inference System . . . . . . Susanna Levantesi, Andrea Nigri, and Gabriella Piscopo Losses in Life Expectancy at Birth from 2020: The Impact of COVID-19 on the Structure of Mortality by Sex and Age in Brazil .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . Denise Helena França Marques and Geovane Máximo Stochastic Comparison Between the Original SARS-CoV 2 Genetic Structure and SARS-CoV 2 - P.1 Variant . .. . . . . . . . . . . . . . . . . . . . Jesús E. García, V. A. González-López, and G. H. Tasca Epidemic Management in the Emergency. Protection Measures, Cost and Compliance with Safety Protocols of the Employees of the Health Units. The Case of the General University Hospital of Heraklion “Venizeleio”, the Management and Pandemic of SARS-CoV-2 . . . . . . . Anna Kefalaki and George Matalliotakis

3

15

37

47

63

77

xi

xii

Contents

Part II 7

8

9

Global Health – Longevity

How to Estimate of the Healthy Life Expectancy (HLE) in the Far Past: Switzerland (1876 –2016) and Forecasts to 2060 with Comparisons with HALE . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . Christos H. Skiadas, Charilaos Skiadas, and Konstantinos N. Zafeiris Global Health and Longevity: An Analysis of Post-World War II Data.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . Konstantinos N. Zafeiris and Christos Skiadas

91

97

Health Care Need Adjusted Prospective Old-age Dependency Ratio in Selected European Countries . . . . . . . . . . . . . . . . . . . . 143 Tomáš Fiala, Jitka Langhamrová, and Jana Vrabcová

10 Spreading Disease Modeling Using Markov Random Fields . . . . . . . . . . 155 Stelios Zimeras Part III

Mortality – Survival

11 Mortality in Greece Before and During the Recent Economic Recession: Short-Terms Effects of the Economic Austerity .. . . . . . . . . . . 167 Byron Kotzamanis, Konstantinos Zafeiris, and Anastasia Kostaki 12 Age Exaggeration Ruses: Infrequent Age Overstatement Distorts the Mortality Curve at Old Age . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 189 Dalkhat M. Ediev 13 Completeness Assessment of Neonatal Deaths in a Region of Brazil: Linkage and Imputing Missing Data . . . . .. . . . . . . . . . . . . . . . . . . . 207 Neir Antunes Paes, Carlos Sérgio Araújo dos Santos, and Tiê Dias de Farias Coutinho 14 A Decomposition Analysis of Differences in Length of Life in the Czech Republic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 219 David Morávek and Jitka Langhamrová 15 Modelling Nigerian Female Mortality: An Application of Four Stochastic Mortality Models . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 229 Oluwaseun Eniola Adegbilero-Iwari and Angela Unna Chukwu 16 Gender, Health and Socio-demographic Influences on Updating Subjective Survival Probabilities . . . . . .. . . . . . . . . . . . . . . . . . . . 245 Apostolos Papachristos and Georgia Verropoulou 17 Estimating Alcohol-Atributable Mortality in Czechia . . . . . . . . . . . . . . . . . 261 Jana Vrabcová, Pechholdová Markéta, and Svaˇcinová Kornélia 18 Alcohol Consumption and Marital Status in the Czech Republic. . . . . 277 Kornélia Svaˇcinová, Markéta Pechholdová, and Jana Vrabcová

Contents

xiii

19 Drug Addiction Mortality Among Young Muscovites: Official Rates and Actual Scale . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 291 G. Semyonova Victoria, E. Ivanova Alla, P. Sabgayda Tamara, V. Zubko Aleksandr, S. Gavrilova Natalia, N. Evdokushkina Galina, and G. Zaporozhchenko Vyacheslav 20 Factors Reducing Child Mortality from Congenital Heart Defects in Russia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 317 A. V. Zubko, T. P. Sabgayda, and V. G. Semyonova Part IV

Special Methods

21 America’s Zika Virus and Its Similarities with African and Asian Lineages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 337 Jesús E. García and V. A. González-López 22 A Relative Entropy Measure of Divergences in Labour Market Outcomes by Educational Attainment .. . . . .. . . . . . . . . . . . . . . . . . . . 351 Maria Symeonaki 23 Assessing the Intergenerational Educational Mobility in European Countries Based on ESS Data: 2002 –2016 .. . . . . . . . . . . . . . 359 Maria Symeonaki and Paraskevi Tsinaslanidou 24 A Different Approach to Current Developments in the Twenty-First Century – Grouping European Countries in Terms of Mortality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 373 Panagiotis Andreopoulos, Fragkiskos G. Bersimis, and Alexandra Tragaki Part V

Various Applications

25 Examining Items’ Suitability as the Marker Indicator in Testing Measurement Invariance . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 389 Anastasia Charalampi, Catherine Michalopoulou, and Clive Richardson 26 Real Estate Pension Schemes: Modeling and Perspectives.. . . . . . . . . . . . 403 Valeria D’Amato, Emilia Di Lorenzo, Gabriella Piscopo, Marilena Sibillo, and Roberto Tizzano 27 Insurance Incentives to Pursue Social Well-Being . .. . . . . . . . . . . . . . . . . . . . 415 Valeria d’Amato, Emilia di Lorenzo, Gabriella Piscopo, and Marilena Sibillo 28 Improved Insurer’s Capital Adequacy of Reserve Risk Using Copula Approach and Hypothesis Tests . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 423 Ilze Zarin¸a, Irina Voronova, and Gaida Pettere

xiv

Contents

29 Assessing the Performance of the European Socio-economic Classification (ESeC) in Eight European Countries for 2018 .. . . . . . . . . 433 Aggeliki Yfanti, Anastasia Charalampi, and Catherine Michalopoulou 30 A Multisite-Multivariate AQI and the Determination of New Threshold Values for Health Risk Categories . . . . . . .. . . . . . . . . . . . . . . . . . . . 449 Giuliana Passamani and Paola Masotti 31 Two Indicators for the Social Sciences . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 461 Alan Araújo Freitas 32 Life Expectancy and Different Parameter Identification in Chinese Retirement Plan . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 473 Yichao Li, Nan Li, and Hong Mi 33 Population Loss Due to Mental Disorders Caused by Deviant Behavior in the 2000s in Russia . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 487 G. Semyonova Victoria, E. Ivanova Alla, P. Sabgayda Tamara, V. Zubko Aleksandr, and S. Gavrilova Natalia Correction to: Two Indicators for the Social Sciences . . .. . . . . . . . . . . . . . . . . . . . Alan Araüjo Freitas

C1

Correction to: Quantitative Methods in Demography . . .. . . . . . . . . . . . . . . . . . . . Christos H. Skiadas and Charilaos Skiadas

C3

Part I

COVID-19 Studies

Chapter 1

Reaction to COVID-19 Pandemic: An Evaluation of Pandemic Management Around the World Yiannis Dimotikalis and Christos H. Skiadas

1.1 Theory and Applications Covid-19 pandemic dominated the health socioeconomic and political issues the last few years in all countries of the World. It was clear from the beginning that the thread was very serious and radical actions should apply. The spread of the disease followed an Exponential growth. Without taking immediate action the health systems would collapse. Radical actions applied at least to reduce the speed of epidemics spread waiting for the appropriate vaccine invention and perhaps new drugs or treatment methodologies. Data are very important. For the first time in human history our national and international systems to collect, store and analyze datasets are so-advanced. However, the analysis of so-many datasets came to be a puzzle difficult to solve. The task was to reduce the growth speed of epidemics; but how fast and what measures where the most appropriate. At least to save the socioeconomic and political system while improving the health systems as well. Now, after almost 20 months after covid-19 invasion, we have enough experience from fighting the pandemic in several countries. 16 countries are selected and the covid-19 spread is presented in the following. In all cases, the 1st Part was a rapidly (exponentially) growth that forced countries to take appropriate measures followed by the 2nd Part with slower epidemic spread. The 3rd almost slightly growing linear part came

Y. Dimotikalis Department of Management Science & Technology, Hellenic Mediterranean University, Crete, Greece e-mail: [email protected] C. H. Skiadas () ManLab, Technical University of Crete, Crete, Greece e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 C. H. Skiadas, C. Skiadas (eds.), Quantitative Methods in Demography, The Springer Series on Demographic Methods and Population Analysis 52, https://doi.org/10.1007/978-3-030-93005-9_1

3

4

Y. Dimotikalis and C. H. Skiadas

after the introduction of vaccines along with measures that came from previous experience from Parts 1 and 2. However, some lessons of the analysis from non-linear systems theory and the related chaotic behavior are missing. Nonlinear systems may result in chaos or chaotic attractors especially when “Time Delays” are present. It means that as long as a delay between treatment and cure appear chaos is present. Even more longer delays appear from local Authorities measures and interventions. Governments tend to propose measures and correct again and again after collecting appropriate data. This could act like to try to correct the Stock Exchange fluctuations by many repeated actions. The best, in this case, is to carefully study the selected actions to be effective and designed for a large time horizon. An example is presented here for Greece while the readers of the paper can apply the same methodology for the other Countries included by introducing data from https:// ourworldindata.org/coronavirus. The chaotic like behavior is presented in the next figure for GREECE. X-Axis is for new deaths and Y-Axis represents the first difference of deaths. The 1st Part (yellow) is followed by the 2nd Part (red) and the 3rd Part (green). The final Part (cyan) is like a continuation of the 1st Part curve. The graph includes chaotic like forms with 28-days delay. Note that a lager limit circle is followed by a smaller one (see Fig. 1.1). A 28days moving average form is selected. It looks like the main delays are coming from a one-month response of the system. This is the time period that should be taken

Fig. 1.1 Greece Covid-19 data, New Deaths to First Difference of Deaths

1 Reaction to COVID-19 Pandemic: An Evaluation of Pandemic Management. . .

5

Fig. 1.2 Greece Covid-19 data, Total Deaths to Total Cases

under consideration when important actions are needed. No immediate response is possible. The reaction time of a large system as a country is critical. Following the analysis above it is clear that the chaotic circles formation need much time to appear. It could that explained the very few papers appearing in the literature. A chaotic attractor is presented for the World covid-19 data without China in the paper by e Fernandes (2020) while Debbouche et al. (2021) study “Chaotic dynamics in a novel COVID-19 pandemic model described by commensurate and incommensurate fractional-order derivatives.” In the following the cases of 16 countries are presented for total deaths/million population vs total cases (pop%) with comments1 (Fig. 1.2). The slope of deaths vs cases curve declines rapidly in most countries due to: protection measures in vulnerable groups and improvement of medical treatment of cases (practically visible in the data curve of mortality from the disease) (Figs. 1.3 and 1.4). Comments from a comparison of 3 countries Greece-Sweden-UK: • More deaths/million in the UK (~2000), Greece (1400), Sweden (1460). • Cases (pop%) in UK (~12%), Sweden (>12%) while in Greece (~6%). • Sweden therefore better treated than UK, Greece with half cases (~6%) had the same deaths result as Sweden (~12%).

1

The authors invite readers feedback in YouTube® channel: COVID-19 Data Analysis: https:// www.youtube.com/channel/UCa553hVoILqn4CJsIhiWW3w

6

Fig. 1.3 Sweden Covid-19 data, Total Deaths to Total Cases

Fig. 1.4 UK Covid-19 data, Total Deaths to Total Cases

Y. Dimotikalis and C. H. Skiadas

1 Reaction to COVID-19 Pandemic: An Evaluation of Pandemic Management. . .

7

Fig. 1.5 France Covid-19 data, Total Deaths to Total Cases

• the different policy in Greece (Lockdown, etc) led to a long 1st Part (until 26 Jan 2021, 12 months) with exponential (polynomial) growth of deaths/cases while in UK and Sweden 1st Part duration is 8 months, (until Aug 2020). 3rd Part of Greece seems to be moving away from the linear model with an increasing trend (new pandemic wave?) (Fig. 1.5). France: Large deviations in 2nd Part, change in the way of measurement. In mid-May 2021 in France there was a revision in the collection of cases resulting in discontinuity in the curve. Because we used the data from OurWorldInData.org website “as it is” we did not make any changes or corrections (Fig. 1.6). Spain: Very close to French (and Italian) data (Fig. 1.7). Belgium: Very similar to UK and France (Fig. 1.8). Italy: Same form as Spain, France, Belgium and UK but the number of total deaths/million is higher close to 2150 (Fig. 1.9). Germany: Similar graph shape for Deaths to Total Cases to Sweden and Greece (Fig. 1.10). Czechia: The European Country from selected cases with the highest number for deaths/million (Fig. 1.11). Canada: The graph form is similar to USA and UK but with only 30% of the death cases from USA or UK (Fig. 1.12). Israel: Similar graph to Czechia but with only 50% deaths/million than Czechia (Fig. 1.13). USA: The graphs are similar to UK. However, the 3rd part tends to Exponential, an alarm for a new wave (Fig. 1.14).

8

Y. Dimotikalis and C. H. Skiadas

Fig. 1.6 Spain Covid-19 data, Total Deaths to Total Cases

Fig. 1.7 Belgium Covid-19 data, Total Deaths to Total Cases

Brazil: the related graph has 3 similar parts with small deviations to differentiate between parts. A country with very high total deaths per million of the population (Fig. 1.15).

1 Reaction to COVID-19 Pandemic: An Evaluation of Pandemic Management. . .

9

Fig. 1.8 Italy Covid-19 data, Total Deaths to Total Cases

Fig. 1.9 Germany Covid-19 data, Total Deaths to Total Cases

Japan shows the smallest 1st Part from the countries selected though it lasted until July 2020. For the 2nd Part a form like a double sigmoid appears. In total Japan has a very low deaths/million number (Fig. 1.16).

10

Y. Dimotikalis and C. H. Skiadas

Fig. 1.10 Czechia Covid-19 data, Total Deaths to Total Cases

Fig. 1.11 Canada Covid-19 data, Total Deaths to Total Cases

South Korea shows an exponential 1st Part and a double sigmoid 2nd Part followed by an almost linear slightly growing 3rd Part. The total deaths per million of the population are less than 50 (Fig. 1.17).

1 Reaction to COVID-19 Pandemic: An Evaluation of Pandemic Management. . .

11

Fig. 1.12 Israel Covid-19 data, Total Deaths to Total Cases

Fig. 1.13 USA Covid-19 data, Total Deaths to Total Cases

Australia: A small number of Deaths/million similar to South Korea with exponential Parts 1 and 2 and an almost linear 3rd Part.

12

Fig. 1.14 Brazil Covid-19 data, Total Deaths to Total Cases

Fig. 1.15 Japan Covid-19 data, Total Deaths to Total Cases

Y. Dimotikalis and C. H. Skiadas

1 Reaction to COVID-19 Pandemic: An Evaluation of Pandemic Management. . .

Fig. 1.16 South Korea Covid-19 data, Total Deaths to Total Cases

Fig. 1.17 Australia Covid-19 data, Total Deaths to Total Cases

13

14

Y. Dimotikalis and C. H. Skiadas

References Data Retrieved from: https://ourworldindata.org/coronavirus [Online Resource]. Data accessed 23 Sept 2021. Debbouche, N., Ouannas, A., Batiha, I. M., et al. (2021). Chaotic dynamics in a novel COVID-19 pandemic model described by commensurate and incommensurate fractional-order derivatives. Nonlinear Dynamics. https://doi.org/10.1007/s11071-021-06867-5 e Fernandes, T.d.S. (2020). Chaotic model for COVID-19 growth factor. Research on Biomedical Engineering. https://doi.org/10.1007/s42600-020-00077-5

Chapter 2

Effects of the Covid-19 Pandemic in the Area of Tension Between the Economy and Climate Change: A Case Study at Rural and City District Level in Southern Germany Moneim Issa and Rolf Bergs

2.1 Introduction In recent months, since the beginning of the Covid-19 pandemic outbreak, several surveys and research studies have been conducted to explore the effects and the consequences of the global crisis on the labour market and the economy during and after the on-going pandemic. Simultaneously, thousands of businesses and institutions are trying to figure out how to remain operational despite the crisis. Perhaps one of the immediately observable consequences of the crisis is the relocation of some of the private and public sector activities to the private home. Working from Home (WFH) proved to be an outstanding alternative for partially maintaining the functionality of production and administration. Interestingly, the results of many relevant analytical efforts indicate that most employees as well as employers foresee unprecedented positive opportunities for WFH after the pandemic, particularly with regard to productivity, competitiveness and life quality.1 Further to that, WFH has a demonstrable impact on the environment (Issa & Bergs, 2020). There is a slowdown of economic activity, less commuting and less international transport; all this leads to a rapid reduction of greenhouse gas (GHG)

1

Currently, during the crisis around 8 million people in Germany work from home in a home office, almost 18% the total labour force (Federal Minister of Labour and Social Affairs 26.04.2020). According to provisional calculations by the Federal Statistical Office (Destatis), around 45.0 million persons resident in Germany were in employment in March 2020. Destatis: Pressemitteilung Nr. 150 vom 30. April 2020 https://www.destatis.de/DE/Presse/Pressemitteilungen/2020/ 04/PD20_150_132.html

M. Issa · R. Bergs () PRAC Bergs & Issa Partnership Co., Bad Soden, Germany e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 C. H. Skiadas, C. Skiadas (eds.), Quantitative Methods in Demography, The Springer Series on Demographic Methods and Population Analysis 52, https://doi.org/10.1007/978-3-030-93005-9_2

15

16

M. Issa and R. Bergs

emissions. All this is well visible in recent satellite imagery from CopernicusSentinel. It shows a major recovery impact of the slowdown on air quality (ESA, 2020a, b). Certainly, WFH has a predominantly economic function; it has evidently alleviated the already predicted devastating mechanisms of the crisis, even though the pandemic staggered national and global economies in a rigid state with a deep uncertainty about future developments. In the forthcoming months, many countries are likely to suffer significant economic imbalances. Based on estimations during the crisis, the global economic growth is expected to shrink by 2.5%, after growth of 3.5% had been forecast in the pre-Corona phase (Michelsen et al., 2020, p. 201). The coronavirus pandemic is currently causing massive changes in wide areas of the economy, labour force and society in almost all parts of the world (CNBC, 2020): Among others, fewer vehicles are driven because commuters now work from home, public transport is largely paralyzed, airplanes stay on the ground, key industries close factories, tourism and business travel comes to a standstill. The German economy and the economy of all individual federal states are also strongly affected by the fast spread of the Covid-19 (e.g. Wollmershäuser et al., 2020; Wollmershäuser & Wohlrabe, 2020). It is to be stressed that without the technological opportunities of WFH the economic crisis would be even considerably stronger. Since the pandemic has forced employers and employees to make more use of home-office-based work, specific potential efficiency gains have become visible in terms of the cost-output ratio in the private (and public) sector and for the consumption of natural resources (such as clean air). This deserves further attention. Prompted by current events, the study at hand is a rapid spatial data exploration restricted by some constraints in the database. It aims to address the peculiar initial effects in the area of tension between the economy and climate change from two points of view, (i) a general sociological one by viewing the acceptance of a changing working life (more WFH) linked to the relationship between working life and the environment (observed in survey data available) and (ii) a specific spatial one by exploring higher resolved primary data. As regards the latter viewpoint, the authors look at the German Bundeslaender Hessen, Bavaria and Baden-Württemberg, which are among the wealthiest and economically most active regions in the EU. Also here, the primary implications of the pandemic have been a massive decline of economic activity in the majority of sectors. Short-term work, redundancy and growing bankruptcy of firms have become rampant as anywhere worldwide. Hence, any positive outcome of this massive reduction of economic activity on the environment and the climate has to be weighed against the welfare loss, i.e. the economic costs of Covid-19. Cleaning of the air during March and April 2020 thus stems from less production and less traffic. Less traffic consists of less commuting, less carriage and less passenger travelling by aircraft, train and ferry. With the easing of the lockdown, emissions rose again immediately, so that the time horizon of this study is just mid-March to mid-April 2020. The study is structured into a methodological introduction including the description of data, a descriptive analysis of the situational perception by secondary data (surveys) and an econometric analysis of the perceived relationships by exploring

2 Effects of the Covid-19 Pandemic in the Area of Tension Between. . .

17

data for different predictors of the slump in greenhouse gas emissions during March to April 2020. Further to that there is a brief conclusion (showing the “what” of the study) and an outlook (discussing the “so what” of the results).

2.2 Methodology and Data The emergence of the pandemic crisis has affected various sectors in the German economy and might have already subtly changed specific perceptions of the working life. In the global descriptive analysis these effects will be reflected. Results of three surveys are reviewed to demonstrate the effects of the crisis on the labour conditions from the perception viewpoint. Based on data from international health and environmental institutions, the global current as well as the pre-pandemic health and environmental situation will be regarded to direct our attention to the relationship between lockdown, the pandemic and the peculiar environmental reaction (GHG emissions) from a spatial viewpoint. With a more technical focus we then model the situation on the ground by merging several spatial datasets available with a view to inspect different possible predictors (representing the current slow-down of economic activity) that could have an immediate effect on GHG reduction at the level of rural and city districts in Hessen, Bavaria and Baden-Württemberg. We selected those three geographically contiguous Bundeslaender because together they represent a large growth pole of the German economy with several major urban agglomerations. The model introduced above is first cast into a simple OLS log-linear regression and then subsequently augmented within more advanced spatial econometric procedures in order to capture neighbourhood effects that are quite likely is such a context. The simple OLS model is: ln(E) = a + b1 ln(I ) + b2 ln(T ) + b3 ln(S) + ε where E means the reduction of NO2 reductions in percent during March and April detected by image analysis of Copernicus-Sentinel-5P satellite imagery (https:// s5phub.copernicus.eu/dhus/#/home). The two netCDF files from 13 March and 13 April 2020 were transformed into 16-bit tiff-images for direct image analysis. NUTS-2 regions were cropped from those files; NO2 emission differences in percentage (March to April 2020) were calculated for every region by ImageJ.2 The variable I means the total local incidence of Covid-19 infection from 5th to

2

The variable has some important limitations: Within NUTS-2 regions data are invariant, hence some error may be induced by comparing a coarse NUTS-2 resolved dataset with more precise NUTS-3 resolved predictors. Due to typical seasonal fluctuations, weather and underlying secular trends of air quality, the simple comparison of two consecutive months is imprecise. However, in contrast to the sharp decrease in 2020 the average NO2 pollution (μg/m3 ) of the 2 months in 2019 had been at the same level for rural and urban areas and traffic routes (German Environment

18

M. Issa and R. Bergs

14th week 2020; data for that are ready-made at district level and taken from the Robert-Koch-Institute (SurvStat@RKI 2.0 application). T means “teleworkability” of a regional economy, i.e. the specific potential of working from home. During the pandemic it is assumed that the levels of teleworkability and true WFH tend to equalise. This concept recurs to Fadinger and Schymik (2020). The data background for that is the production statistics (NACE) at the NUTS-3 level of the three Bundeslaender viewed. The teleworkability index is based on the calculation of Dingel and Neiman (2020)3 and then merged with the regional production data published by Eurostat (table “namq_10_a10”). The variable S represents the change of short-term work at district level in percent. These data are ready-made available at district level from the Federal Agency of Labour. We have not deemed the change of unemployment a meaningful predictor because during March to April 2020 most companies have used short-term work to bridge the economic downturn and to resume full production within weeks or few months. The variables I and S on the one hand and T on the other are to be distinguished in their characters. While T is a lasting independent variable, I and S are predictors that just take effect during the pandemic. They just measure whether infection incidence or short-term work leads to more or less GHG emissions. The variable I is not expected to directly influence GHG emissions but it could shed light on how timely and appropriately people have reacted with lockdown and sufficient protection against contagion. It is to be stressed that infection incidence only represents cases confirmed by PCR4 tests and not those many undetected cases. Since the proportion of confirmed to undetected cases may vary across the spatial units, inconsistency in estimation cannot be fully ruled out. Apart from I, T and S there are however important spatial effects to be expected; such rural districts are close or distant to each other, and particularly a change of GHG emission is also influenced by spatial spillover effects. We therefore address those effects within an augmented spatial autoregressive and spatial error model (Anselin & Florax, 1995; LeSage & Pace, 2009; Dubé & Legros, 2014; Golgher & Voss, 2016). ln(E) = a + ρW ln(E) + b1 ln(I ) + b2 ln(T ) + b3 ln(S) + ε

(SAR)

or 

ln(E) = a + b1 ln(I ) + b2 ln(T ) + b3 ln(S) + ν ν = λW ν + ε

(SEM)

Agency, 2020, p. 14). Therefore, the simple direct comparison of March and April 2020 is deemed feasible for that purpose. 3 Cf. Supplementary replication package: column ,teleworkable_emp“ in file ,NAICS_workfromhome.csv“ stored in the sub-folder ,national_measures“ (Internet link: see references). 4 Polymerase chain reaction.

2 Effects of the Covid-19 Pandemic in the Area of Tension Between. . .

19

where W represents a row-standardised spatial weight matrix, ρ and λ are the spatial coefficients. The error term ε is i.i.d. with mean zero and a constant variance: ε  N(0, σ 2 ). In the SEM case ν is the sum of spatial (λWν) and non-spatial error (ε). The right choice between both models is determined by different tests, such as Moran’s I of the residuals and (Robust) Lagrange multiplier (LM) tests. In addition to the above standard spatial procedures we also explored the context in a Spatial Durbin model in that spatial lags of both, the dependent variable and the predictors are included, a SLX model where instead of the spatial lag of the dependent variable the lags of predictors are considered and a Spatial Durbin Error model that extends the SLX model. In case of the Spatial Durbin model the loglikelihood function did not converge to a maximum, therefore we had to drop this approach. The SLX model (Halleck Vega & Elhorst, 2015) to be efficiently, consistently and unbiasedly estimated by OLS is formulated as follows: ln(E) = a + b1 ln(I ) + b2 ln(T ) + b3 ln(S) + θ1 V ln(I ) + θ2 V ln(T ) + θ3 V ln(S) + ε

V, in contrast to W, is a binary spatial weight matrix where the off-diagonal elements are set 1 for inter-district distance dmn < 30 kilometers5 and set zero otherwise. The coefficient θ represents spatial spillover effects of the predictors which are indirectly implied by respective average influences from neighbour regions, hence bi + θ i shows direct plus indirect (=total) effects of an independent variable. The log-linear form allows to directly interpret coefficients as elasticities, thus if e.g. the sum of coefficients of predictors X and VX is 0.5, a 1% increase of X in region i, and on average in the neighbour regions, would lead to a total 0.5 increase of the dependent variable Y when keeping the other influences constant. Since one can assume that there could be a possible significant spatial error in the residuals that might affect the estimates, it is suggested to extend the SLX model by applying a Spatial Durbin Error procedure (SDEM): 

ln(E) = a + b1 ln(I ) + b2 ln(T ) + b3 ln(S) + θ1 V ln(I ) + θ2 V ln(T ) + θ3 V ln(S) + ν ν = λW ν + ε

W is the weight matrix (inverse distance) as explained earlier. The SDEM procedure should contribute to a more precise estimate. Our study addresses the actual determinants of the changes in air pollution. We hypothesise that teleworkability is a general predictor, while the other are specific ones that accrue from the current constraints in public life. It should be noted, however, that there could be also reverse relationships, namely (i) air pollution (particulate matter) as a suspected carrier of outdoor contagion, so that GHG emissions are also a potential predictor of the geographical distribution of the pandemic (Setti et al., 2020) or (ii) that the infection rate I depends on teleworkability T (Fadinger & Schymik, 2020). Since there is only the proof of a

5

Distance is calculated from coordinates by the Vincenty formula.

20

M. Issa and R. Bergs

virus concentration on particulate matter but no respective evidence of contagion, a further investigation in this direction would be speculative by now. Fadinger and Schymik (2020) estimated the infection rate in an epidemiological model after first estimating the impact of WFH on the contact rate at a NUTS 2 resolution (nonspatial). They found strongly significant negative effects of WFH on the infection rate. We also tested this model with a spatial regression procedure for our data at NUTS-3 level but had to conclude with an insignificant estimate.6

2.3 Results 2.3.1 Acceptance of Teleworking and Its Possible Impact on the Environment and Public Health A sustainable reduction of GHG emissions needs teleworkability (the potential of WFH) to materialise. Apart from legal and technical preconditions this largely depends on readiness on the part of employers and employees. The Corona crisis has forced people to WFH and may have changed minds simply due to their newly acquired experience with WFH. In the following section three consecutive recent surveys among potentially teleworking employees are illustrated. An early survey conducted in March 2020 by the BVDW (Bundesverband Digitale Wirtschaft) predicted up to 75% of the employees to be prepared for WFH during the pandemic (Fig. 2.1). Regarding the expectations of the employees toward their employers, over 66% expect their employer would undertake adequate technical measures for WFH, 45.3% assume that their employer is technically capable to provide the necessary WFH facilities, and 45.7% are still sceptical about the willingness of the employer (Statista, 2020). Another survey executed by Bitkom mid-March 2020 (Bitkom, 2020) found that already one out of two gainfully employed is working entirely or at least partially at home as a result of the Corona pandemic. Advanced communication technology has vastly contributed to digitize business activities and has strongly facilitated the rise of WFH, as responders argued. Meanwhile, many activities could be carried out from a digitalized home office such as web conferences, transfer of legal documents, organizing customer and supplier meetings and a lot more. However, still 62% of employees with a home office permit preferred the company’s office as their place of work. Now it would be interesting to see how those perceptions and standpoints have changed in these times of classical uncertainty, namely the Corona crisis.

6

Data and estimates are available from the authors on demand. We suspect a mismatch between teleworkability and WFH or different spatial resolution levels to be reasons of such strikingly different results.

2 Effects of the Covid-19 Pandemic in the Area of Tension Between. . .

21

employees sceptical about companies willingness for WFH employees assuming that company has the capability for WFH employees expecting company to supple neccessary infrastructure for WFH employees uncertain about WFH employees prepared for WFH

0%

10%

20%

30%

40%

50%

60%

70%

80%

Data source: Statista (2020)

Fig. 2.1 BVDW Survey (March 2020). (Data source: Statista, 2020)

According to a representative survey implemented in April 2020 by the BITD (Bavarian Research Institute for Digital Transformation), already two-thirds of the employees surveyed would prefer to have more home office after the Corona crisis than before. This suggests that there may have been some habituation effect and an acceptance of WFH well increased during the early weeks of the pandemic (Stürz et al., 2020). Further to that and unexpectedly, the majority of employers seem to be technically well prepared for the home-office option, and a growing number of employees is satisfied with this new way of working arrangement. They favour to continue working from home after the crisis.7 The surveys reviewed above clearly confirm a growing readiness among employers and employees, triggered by the Covid-19 crisis, to make more use of WFH. Still, there is some uncertainty about sustainability of this mutual readiness in post-Corona times. But apart from a possible protection against contagion during lockdown, WFH has definitively a beneficial and immediately perceivable impact on the environment and public health in general. Through strong multiplier effects along the production chain GHG emissions in Germany may decrease by at least 50 million tons of GHG by the end of this year compared to 2019. Depending on further development in the Covid-19 pandemic, there may be up to 120 million tons of CO2 reduction. This would result in a reduction in emissions by 40–45% compared to 1990 (Agora, 2020). In this connection, it can be estimated that factories’ shutdown will inevitably reduce the demand for steel, cement and chemicals and if the production discontinues over a few weeks, the GHG reduction will be around 10 million tons. Given that the shutdown activities are going to continue for three further months, the emissions

7

Nonetheless, there is still a widespread uncertainty among various firms operating in specific sectors that the situation may again change after the pandemic, due to unknown changes in the economic frame conditions, particularly with regard to productivity and competitiveness. These arguments are countered by several studies, which predicted more advantages for companies than disadvantages. A 2015 study conducted by the Stanford University found that productivity among call-center employees at the Chinese travel agency Ctrip went up by 13% when they worked from home (Bloomberg, 2020).

22

M. Issa and R. Bergs

will be reduced by around 18 million tons as a result of production stoppages. If the crisis lasts longer, the GHG emission reduction of about 25 million tons is likely to occur (Agora, 2020). The environmental situation and the status of public health are closely interrelated. According to the World Health Organization (WHO), air pollution causes lung and respiratory diseases, cancer, heart disease and diabetes and kills about seven million people around the world every year, and unfortunately over “95 percent of the EU urban population remain exposed to pollutant concentrations above WHO air quality guidelines.” Traffic noise is a further related burden (The European Environment Agency, 2019). In connection with the Covid-19 pandemic, the WHO noted that people who contract SARS-Cov-2 have much higher death rates, between 700% and 1400% higher mortality rates if they have any of the above stated preconditions (WHO, 2020). Covid-19 enters the respiratory tract, attacks the lung cells and becomes severe in almost 20% of cases. Pneumonia with Covid-19 can result in more severe than “normal” pneumonia because it occurs in all areas of the pulmonary system, and there is no treatment yet available. Preliminary evidence suggests that areas with poorer air quality are more vulnerable. Lombardy, Italy, has one of the worst air quality levels in Europe, and it became Italy’s most Covid-19 affected area with deaths. A similar environmental situation could be found in Wuhan that became the first centre of Covid-19 deaths (WHO, 2020; see also: Setti et al., 2020). Health, climate change and the patterns of our working life are thus closely interlinked. The Corona crisis provides a real-world experimental laboratory to explore exactly that context on the ground. With other words: How does teleworking (as a proxy for lockdown) affects the environment and what do most recent data at rural and city district tell us about that context?

2.3.2 The Empirical Analysis at a Regional Level In the empirical section we aim to closer inspect the context of relationships and influences statistically. The pace of cleaning the air over Europe appears tremendous when comparing the Copernicus-Sentinel 5P images that suggest a major reduction of GHG emissions (here represented by NO2 ) during March to April 2020.8 With image analysis (ImageJ) it is possible to directly estimate the change of emissions along the change of the Digital Number (DN) of pixels in 16-bit transformed tiff-files (Table 2.1). The direct percentage change of mean DN for Southern Germany as shown by the two images is thus 14.8%. A similar estimate (daily change compared to averages

8

The climate change in Europe during the pandemic has been illustrated in the comparison of nitrogen dioxide emissions over Europe between March/April 2019 and 2020. Copernicus: Sentinel-5P (Precursor – Atmospheric Monitoring Mission) 2020: https://www.esa.int/ Applications/Observing_the_Earth/Copernicus/Sentinel-5P/Air_pollution_remains_low_as_ Europeans_stay_at_home

2 Effects of the Covid-19 Pandemic in the Area of Tension Between. . .

23

Table 2.1 Basic moments of NO2 emissions compared (March–April 2020) March 2020 April 2020

Mean DN 179.232 210.280

Standard dev. DN 34.915 23.192

Mode DN 193 226

Min DN 30 52

Max DN 249 250

Source: Copernicus-Sentinel; own calculations

Fig. 2.2 Reduction of GHG emissions at NUTS-2 level in Hessen, Baden-Württemberg and Bavaria: Reduction of GHG emissions March–April 2020 (percent). (Source: Copernicus Sentinel; own calculations)

in former years) is given by Le Quéré et al. (2020) at a global level. The further analysis of this dataset has been done at the level of NUTS-2. Results are displayed by the following Fig. 2.2: The distribution appears non-linear with one noticeable outlier, namely the NUTS-2 region Darmstadt. Hence, reduction of GHG emissions in the Frankfurt/Rhein-Main region are by far the largest in Southern Germany. Interestingly, the images suggest a levelling in the distribution of GHG emissions. The April 2020 image appears more homogeneous than that from March. Therefore, the top and bottom ranked districts will be subject to a closer inspection by comparing observed and estimated GHG reductions. These data are used to fill the vector of the dependent variable. Independent variables are deemed the teleworkablity (as a proxy to represent WFH during the pandemic), the sudden surge of short-term work implied by the lockdown and the local infection incidence. According to the data from the RKI, total infection incidence is significantly higher in Bavaria and Baden-Württemberg as compared with Hessen. This might stem from random incidence, but also systematic differences, at least the variation appears impressively strong which significantly less incidence in Hessen (Fig. 2.3). As regards teleworkability the variation ranges between o. Denote by S = A × Ao the state space of

5 Stochastic Comparison Between the Original SARS-CoV 2 Genetic. . .

65

the process (Xt ). Consider the notation akn meaning the concatenation of elements ak ak+1 . . . an where ai ∈ A, ∀i : k ≤ i ≤ n. The parameters that define the behavior of the process are the transition probabilities, introduced here, −1 P (a|(z, s)) = Prob(Xt = a|Xt −G = z, Xtt −o = s),

(5.1)

for each a ∈ A and (z, s) ∈ S. Then, we need to identify the set of probabilities representing the behavior of the process, {P (a|(z, s)), a ∈ A, (z, s) ∈ S}. Definition 2.1 A G-Markov chain (Xt ) is a discrete time Markov chain on a finite alphabet A, with state space S = A × Ao , where o < ∞, transition probabilities following Eq. (5.1), for a finite G, such that G > o. To estimate the transition probabilities, under Definition 2.1, consider x1n a sample  of the process Xt , (z, s) ∈ S, a ∈ A and n > G. Denote by N((z, s)) the number of occurrences of the state (z, s) in x1n , this is N((z, s)) = |{t : G < t ≤ −1 n, xt −G = z, xtt−o = s}| and the ocurrences of (z, s) ∈ A × Ao followed by a ∈ A −1 is N((z, s), a) = |{t : G < t ≤ n, xt −G = z, xtt−o = s, xt = a}|. Thus, for each N((z,s),a) a ∈ A and (z, s) ∈ S, N((z,s)) is the estimator of P (a|(z, s)) given by Eq. (5.1), N((z, s), a) , a ∈ A, (z, s) ∈ S. Pˆ (a|(z, s)) = N((z, s))

(5.2)

For an efficient estimate to be achieved, we introduce a model that produces a reduction in the number of probabilities to be estimated. The idea is to use different states to estimate the same probability. Definition 2.2 Let (Xt ) be a Markov chain following Definition 2.1, of order o on a finite alphabet A, parameter G > o and, state space S = A × Ao , i. v, r ∈ S are equivalent if P (a|v) = P (a|r) ∀a ∈ A. ii. (Xt ) is a G-Markov chain with partition I = {I1 , I2 , . . . , I|I| } if this partition is the one defined by the relationship introduced by i. Let I = {I1 , I2 , . . . , I|I| } be a partition of S, now define the probabilities in terms of parts (of I), then let be P (I, a) = r∈I P (r, a), P (I ) = r∈I P (r). If P (I ) > 0 we can define P (a|I ) =

P (I, a) . P (I )

(5.3)

Then, ∀a ∈ A, P (a|I ) = P (a|r), ∀r ∈ I, meaning that we use all the states r ∈ I to estimate the same parameter. This way to represent a stochastic process (Xt ) is called Partition Markov Model (PMM), see García and González-López (2017) and

66

J. E. García et al.

García et al. (2020). Clearly, the notions of probability defined in terms of a part, require that for the states that are within that part, at least item i of Definition 2.2 be verified. Fact that requires some strategy to determine which states are grouped in the same part and which would be the optimal configuration (partition representing Definition 2.2 ii.). The estimation of the model proposed by Definition 2.2 is carried out by maximizing the Bayesian Information Criterion (BIC), see Schwarz (1978) and García and González-López (2017). Suppose that we consider the sample x1n of the process (Xt ), define N(I ) = (z,s)∈I N((z, s)) and N(I, a) = (z,s)∈I N((z, s), a), for a ∈ A, the BIC is given by, BIC(x1n , I) = ln



 a∈A,I ∈I

N(I, a) N(I,a)  N(I )

−

(|A| − 1)|I| ln(n) . α

(5.4)

We see then that the first term is the logarithm of the maximum value of the pseudolikelihood and the second term is the penalty term, with α being a positive value, (in Schwarz (1978) α = 2). In practice, the partition is obtained through a BIC-based metric (proposed in García & González-López 2017) and, by means of algorithms, as the one proposed in García and González-López (2011), which produces a consistent estimation of I. The following notation allows us to compare samples of stochastic processes, under some G-model identified through the BIC. Definition 2.3 Consider two G-Markov chains (following Definition 2.1) (X1,t ) and (X2,t ), of order o, with parameter G > o, finite alphabet A, state space n1 n2 S = A × Ao and independent samples x1,1 , x2,1 , respectively. i. For a state r ∈ S, ⎧   ⎨  α Nk (r, a) n1 n2 Nk (r, a) ln dr (x1,1, x2,1 ) = ⎩ (|A| − 1) ln(n1 + n2 ) Nk (r) a∈A

k=1,2



−N1,2 (r, a) ln

N1,2 (r, a) N1,2 (r)

 ,

ii. n1 n2 n1 n2 , x2,1 ) = max{dr (x1,1 , x2,1 )}, dmax(x1,1 r∈S

with N1,2 (r, a) = N1 (r, a) + N2 (r, a), N1,2 (r) = N1 (r) + N2 (r), where N1 and N2 n1 n2 are given as usual, computed from the samples x1,1 and x2,1 respectively. With α a real and positive value.

5 Stochastic Comparison Between the Original SARS-CoV 2 Genetic. . .

67

dr and dmax are statistically consistent, that is, by increasing the min{n1 , n2 } grows their ability to detect (1) discrepancies, when the underlying laws are different and, (2) similarities, when the underlying laws are the same. dr and dmax have useful properties, for instance dr is a local metric (restricted n1 n2 N2 (r,a) to the state r): dr (x1,1 , x2,1 ) ≥ 0 with equality ⇔ NN11(r,a) ∀a ∈ A; (r) = N2 (r) n3 n3 n1 n2 n2 n1 n1 n2 n1 n2 , x2,1 ). dr (x1,1 , x2,1 ) = dr (x2,1 , x1,1 ) and, dr (x1,1 , x2,1 ) ≤ dr (x1,1 , x3,1 ) + dr (x3,1 dmax is a global notion of proximity, since it can be computed for all the states r ∈ S, but it is not a metric. In the application we use α = 2 (see Definition 2.3-i.), which is related to the Bayesian Information Criterion, in its original version, see Schwarz (1978) and García et al. (2018). In the next section, we apply the model given by Definition 2.1 to complete genomic sequences of SARS-CoV-2 P.1 variant. We also measure the discrepancies between each sequence of SARS-CoV 2 - P.1 variant and the original sequence of SARS-CoV-2 (Wuhan).

5.3 SARS-CoV 2 and Variant P.1 In this section, we describe, in Sect. 5.3.1, the sequences of SARS-CoV 2 - P.1 variant that are investigated under the notions introduced in Sect. 5.2. In Sect. 5.3.2, we show the models that are selected for each sequence introduced in Sect. 5.3.1. In order to identify evidence of changes from the original sequence of SARS-CoV 2 and the set of sequences of SARS-CoV 2 - P.1 variant, we compare those sequences using the tools introduced in Sect. 5.2.

5.3.1 SARS-CoV 2 Data Sets The database consists of a collection of genetic sequences in FASTA format. For that reason, the alphabet that is considered is the genomic one, that is, A = {a, c, g, t}. The complete genome sequences of SARS-CoV 2 - P.1 variant used in this paper can be found in GISAID source (https://gisaid.org), the sequences are listed in Table 5.1. Table 5.1 records the Accession ID of each sequence, the collection data (January of 2021) and the sample sizes (at least 29,593). The originating lab of the sequences EPI_ISL_1034306, EPI_ISL_106828x for x = 1, 2, 3, 4 is Laboratório de Ecologia de Doenças Transmissíveis na Amazônia, Instituto Leonidas e Maria Deane - Fiocruz, Amazônia. The originating lab of EPI_ISL_906071 is LACEN Laboratório Central de Saúde Pública Dr. Costa Alvarenga, Piauí. The originating lab of the sequences EPI_ISL_90608x for x = 0, 1 is Hospital Beneficência Portuguesa, São Paulo. And, the originating lab of EPI_ISL_94062x for x = 6, 7 is Hospital Central São Caetano do Sul, São Paulo.

68 Table 5.1 Collection of sequences of SARS-CoV 2 P.1 variant, obtained from GISAID source

J. E. García et al. Accession ID EPI_ISL_1034306 EPI_ISL_1068281 EPI_ISL_1068282 EPI_ISL_1068283 EPI_ISL_1068284 EPI_ISL_906071 EPI_ISL_906080 EPI_ISL_906081 EPI_ISL_940626 EPI_ISL_940627

Collection Data 2021-01-29 2021-01-06 2021-01-11 2021-01-12 2021-01-13 2021-01-19 2021-01-22 2021-01-22 2021-01-21 2021-01-22

Sample Size 29,593 29,593 29,741 29,784 29,784 29,867 29,858 29,874 29,835 29,848

In the following subsection, we show the results of the fit of model Definition 2.1 in each sequence listed in Table 5.1 and a comparative study that seeks to identify the similarity and divergence between the sequences listed in Table 5.1 with the sequence MN908947 (Wuhan).

5.3.2 Results According to the results of García et al. (2020), the best model to represent the sequence MN908947 (Wuhan) is given by Eq. (5.1) under the alphabet A = {a, c, g, t}, with o = 3 and G = 9. Then the question is, is this model also capable of representing the P.1 variant?. It is expected some maintenance since the virus is the same. In Table 5.2, we report the BIC values (see Eq. (5.4)) for nine possible models following Eq. (5.1), the higher the BIC value, the more indicated is the model. We use o = 3, since the genomic structure is organized in triples of elements of A, then essentially we need to point the adequate value for G. We verify that except for one of the sequences (sequence EPI_ISL_1068282), the indicated model is given by the Eq. (5.1) with o = 3 and G = 9. In the case of sequence EPI_ISL_1068282, the model given by Eq. (5.1) with o = 3 and G = 9 is the second-best placed. The preference indicated by the variant P.1 is clear, the winning model is the one given by the Eq. (5.1) with o = 3 and G = 9. Which makes sense since the virus is the same, and the alterations in relation to the model indicated in García et al. (2020) must somehow maintain some structure. The question is, despite the fact that the model remains, how is this new phenomenon organized, that is, how is the performance of P.1 variant organized in relation to the original version? For each pair of sequences listed in Tables 5.1, 5.3 and 5.4 record the values of dmax (see Definition 2.3) and the state where such maximum occurs. The values of dmax are used in Fig. 5.1. Also, they are all small, less than 0.014. For the definition of dmax, it is necessary to determine the transition probabilities given by the Eq. (5.2). For this, we must set the values of o and G, taking into account the

EPI_ISL_1068281 −39,224.14 −39,233.30 −39,237.48 −39,248.40 −39,264.45 −39,223.57 −39,236.87 −39,242.84 −39,230.49

EPI_ISL_906080 −39,591.33 −39,608.28 −39,603.60 −39,614.18 −39,613.44 −39,582.11 −39,612.22 −39,623.80 −39,592.86

EPI_ISL_1034306 −39,220.30 −39,234.95 −39,236.14 −39,254.11 −39,263.19 −39,210.87 −39,232.85 −39,246.43 −39,225.40

EPI_ISL_906071 −39,599.57 −39,600.95 −39,604.76 −39,618.61 −39,622.50 −39,578.74 −39,599.44 −39,614.30 −39,596.50

G 4 5 6 7 8 9 10 11 12

G 4 5 6 7 8 9 10 11 12

EPI_ISL_906081 −39,606.19 −39,610.92 −39,612.93 −39,623.53 −39,623.52 −39,586.39 −39,613.91 −39,623.22 −39,600.25

EPI_ISL_1068282 −39,425.68 −39,445.34 −39,449.49 −39,460.05 −39,466.76 −39,429.09 −39,451.08 −39,465.29 −39,437.66 EPI_ISL_940626 −39,556.51 −39,568.07 −39,573.38 −39,589.66 −39,590.44 −39,548.22 −39,577.71 −39,594.21 −39,561.07

EPI_ISL_1068283 −39,485.59 −39,496.43 −39,505.94 −39,519.41 −39,525.24 −39,475.57 −39,500.98 −39,525.31 −39,485.43

EPI_ISL_940627 −39,561.29 −39,569.12 −39,572.15 −39,585.40 −39,586.47 −39,540.60 −39,567.87 −39,585.55 −39,561.09

EPI_ISL_1068284 −39,487.87 −39,502.76 −39,502.85 −39,516.08 −39,524.72 −39,487.86 −39,496.44 −39,519.77 −39,490.65

Table 5.2 Model selection for sequences of SARS-CoV 2, P.1 variant (see list in Table 5.1), BIC values from the model given by Eq. (5.1) with o = 3 and G = 4, 5, 6, 7, 8, 9, 10, 11, 12. In bold, the highest BIC value (see Eq. (5.4)) pointing the best fit

5 Stochastic Comparison Between the Original SARS-CoV 2 Genetic. . . 69

70

J. E. García et al.

Table 5.3 From left to right, each table reports (1) the sequence with which the sequence on top is compared, (2) the value of dmax, and (3) the state where the dmax happens EPI_ISL_Z ↓Z 1034306 1068281 1068282 1068283 1068284 906071 906080 906081 940626 940627

EPI_ISL_1034306

EPI_ISL_Z ↓Z 1034306 1068281 1068282 1068283 1068284 906071 906080 906081 940626 940627

EPI_ISL_1068281

0 0.00194 0.01337 0.01337 0.01337 0.01337 0.01337 0.01277 0.01337 0.01337

– 0 0.01337 0.01337 0.01337 0.01337 0.01337 0.01277 0.01337 0.01337

(c, tga) (t, cga) (t, cga) (t, cga) (t, cga) (t, cga) (t, cga) (t, cga) (t, cga)

(t, cga) (t, cga) (t, cga) (t, cga) (t, cga) (t, cga) (t, cga) (t, cga)

EPI_ISL_Z ↓Z 1034306 1068281 1068282 1068283 1068284 906071 906080 906081 940626 940627

EPI_ISL_1068282

EPI_ISL_Z ↓Z 1034306 1068281 1068282 1068283 1068284 906071 906080 906081 940626 940627

EPI_ISL_1068283

– – 0 0.00998 0.00998 0.00998 0.00998 0.00998 0.00998 0.00998

– – – 0 0.00237 0.00481 0.00481 0.00481 0.00571 0.00395

(g, gcc) (g, gcc) (g, gcc) (g, gcc) (g, gcc) (g, gcc) (g, gcc)

(c, agg) (t, ccc) (t, ccc) (t, ccc) (a, cgc) (t, ccc)

results reported in Table 5.2 and the reference García et al. (2020), we set o = 3 and G = 9. Figure 5.1 shows the dendrogram build from the values of dmax (see Definition 2.3-ii) between the pairs of sequences listed in Table 5.1 (Tables 5.3 and 5.4). Note that the dmax values are all close to zero (see Fig. 5.1), so they do not represent a relevant divergence. Despite this, what is extracted from Tables 5.3 and 5.4 is that there are some states that predominate in the sense of producing the highest value of dmax, those are (t, cga), (g, gcc), (a, cgc), (t, ccc), (g, ctc), (a, cca). Next, we proceed to compare the sequences listed in Table 5.1 plus the sequence MN908947, we do that by means of Fig. 5.2. Figure 5.2 shows the dendrogram build from the values of dmax (see Definition 2.3-ii) between the pairs of sequences listed in Table 5.1 plus the sequence MN908947 (from Wuhan 2019) (on the left). The dendrogram shows a clear separation between the original sequence of SARS-CoV 2 and the groups formed by the P.1 variant of SARS-CoV 2. That is, the G-model (Definition 2.1) is maintained because it is the same virus, but the transition probabilities estimated empirically by Eq. (5.2) reveals, through the

5 Stochastic Comparison Between the Original SARS-CoV 2 Genetic. . .

71

Table 5.4 Continuation of Table 5.3. From left to right, each table reports (1) the sequence with which the sequence on top is compared, (2) the value of dmax, and (3) the state where the dmax happens EPI_ISL_Z ↓Z 1034306 1068281 1068282 1068283 1068284 906071 906080 906081 940626 940627

EPI_ISL_1068284

EPI_ISL_Z ↓Z 1034306 1068281 1068282 1068283 1068284 906071 906080 906081 940626 940627

EPI_ISL_906080

EPI_ISL_Z ↓Z 1034306 1068281 1068282 1068283 1068284 906071 906080 906081 940626 940627

– – – – 0 0.00481 0.00481 0.00481 0.00571 0.00395

– – – – – – 0 0.00263 0.00571 0.00242

(t, ccc) (t, ccc) (t, ccc) (a, cgc) (t, ccc)

(g, ctc) (a, cgc) (a, cca)

EPI_ISL_Z ↓Z 1034306 1068281 1068282 1068283 1068284 906071 906080 906081 940626 940627

EPI_ISL_906071

EPI_ISL_Z ↓Z 1034306 1068281 1068282 1068283 1068284 906071 906080 906081 940626 940627

EPI_ISL_906081

– – – – – 0 0.00122 0.00263 0.00571 0.00241

– – – – – – – 0 0.00571 0.00298

(a, aaa) (g, ctc) (a, cgc) (a, cca)

(a, cgc) (a, ata)

EPI_ISL_940626 – – – – – – – – 0 0.00571

(a, cgc)

notions of Definition 2.3, the separation between the original version of the virus MN908947, collected in 2019 and, the P.1 variant of the virus, a year after the beginning of the crisis.

72

J. E. García et al.

Fig. 5.1 Average type dendrogram build from the dmax values, see Definition 2.3-ii., using the complete sequences of SARS-CoV 2, P.1 variant, listed in Table 5.1

To obtain a more detailed comparison between the members of P.1 variant and the original sequence, we construct Table 5.5. Table 5.5 records the five highest values of dr and the states where each of them has occurred, on top of each table (related to each sequence) is given the dmax value (in bold letter). Table 5.5 also records the five states where the dr metric takes the highest values. The idea is to compare the magnitudes of the values of this dr with those recorded only in the P.1 variant, see Tables 5.3 and 5.4. When comparing Tables 5.3, 5.4 with Table 5.5, we see that the magnitude of dr undergoes an increase from approximately 0.0134 to approximately 0.1, which produces the separation of the sequence MN908947 in relation to the clusters of P.1 variant (see Fig. 5.2 and also Fig. 5.1). It is still found that the greatest discrepancy between each member of the group of sequences of P.1 and MN908947 is produced in the state (a, aaa) and, that this is not a state frequently indicated as causing the greatest value of dr (dmax), in the study using only the sequences of P.1 variant (Tables 5.3 and 5.4). Table 5.6 records the transition probabilities from the state (a, aaa) to any element of the alphabet A = {a, c, g, t}, calculated using the Eq. (5.3) in two situations, (1) considering the set of P.1 sequences (listed in Table 5.1) and (2) considering only the sequence MN908947. From the results of Table 5.6, we see how the transition probability for the element a (t) falls (rises) around in 20% in the P.1 variant in relation to the original version of SARS-CoV 2 (Wuhan). The

5 Stochastic Comparison Between the Original SARS-CoV 2 Genetic. . .

73

Fig. 5.2 Dendrogram build from the dmax values, see Definition 2.3-ii, using the complete sequences of SARS-CoV 2, P.1 variant, listed in Table 5.1 (in blue) plus the sequence MN908947, from Wuhan 2019 (in red)

transition probabilities for t and g in P.1 variant rise slightly when comparing with the probability recorded for the original version of SARS-CoV 2 (Wuhan).

5.4 Conclusion In this paper, it is proposed to identify the most appropriate model within the G-model family (see Definition 2.1) to represent the stochastic performance of complete genetic sequences of SARS-CoV 2 - P.1 variant (Brazil). The results point to the same model previously reported as the most indicated for the original sequence of SARS-CoV 2, MN908947 (Wuhan). This evidence shows that certain variants can preserve the original stochastic structure reported in García et al. (2020). Through a more detailed study, using the identified G-model, it is possible to show that in fact the P.1 variant is separated from the original version (see, for

74

J. E. García et al.

Table 5.5 Values of the metric dr and state r (Definition 2.3-i) between the sequence MN908947 and the sequence EPI_ISL_x, where x = 1034306, 1068281, 1068282, 1068283, 1068284, 906071, 906080, 906081, 940626, 940627, for the states where the metric takes the 5 highest values of dr . in bold the dmax values EPI_ISL_1034306

EPI_ISL_1068282

EPI_ISL_1068284

EPI_ISL_906080

EPI_ISL_940626

State r (a, aaa) (a, agg) (t, cga) (c, acc) (c, ccc) (a, aaa) (c, acc) (c, ccc) (g, gcc) (c, ttc) (a, aaa) (c, acc) (c, ccc) (c, ttc) (t, ttt) (a, aaa) (c, acc) (c, ccc) (g, acc) (g, tcc) (a, aaa) (c, acc) (c, ccc) (a, cca) (g, acc)

Table 5.6 Transition probabilities from the state (a, aaa) to any element of the alphabet · ∈ A = {a, c, g, t}

dr 0.10282 0.01678 0.01587 0.01367 0.01233 0.10280 0.01632 0.01232 0.00997 0.00694 0.09543 0.01632 0.01232 0.00694 0.00619 0.09542 0.01632 0.01232 0.00564 0.00551 0.09543 0.01632 0.01232 0.00914 0.00763

EPI_ISL_1068281

EPI_ISL_1068283

EPI_ISL_906071

EPI_ISL_906081

EPI_ISL_940627

· element of A a c g t

State r (a, aaa) (a, agg) (c, acc) (t, cga) (c, ccc) (a, aaa) (c, acc) (c, ccc) (c, ttc) (t, ttt) (a, aaa) (c, acc) (c, ccc) (g, tcc) (g, ggg) (a, aaa) (c, acc) (c, ccc) (g, tcc) (g, ggg) (a, aaa) (c, acc) (c, ccc) (a, cca) (a, tag)

P.1 variant (2021) Pˆ (·|(a, aaa)) 0.28621 0.20071 0.22293 0.29015

dr 0.10282 0.01678 0.01632 0.01587 0.01233 0.09543 0.01632 0.01232 0.00694 0.00619 0.07531 0.01632 0.01232 0.00551 0.00543 0.07532 0.01632 0.01232 0.00551 0.00543 0.08170 0.01632 0.01232 0.00914 0.00629

MN908947 (Wuhan, 2019) Pˆ (·|(a, aaa)) 0.35610 0.20030 0.20400 0.23960

example, Fig. 5.2). Such evidence indicates a movement (from the original version to the P.1 variant) in the direction of change of the transition probabilities associated with the underlying processes generating the sequences (see Eq. (5.2)). This is a consequence of considering that the estimation of the transition probabilities impacts the behavior of the metric (Definition 2.3) pointing changes, see Fig. 5.2.

5 Stochastic Comparison Between the Original SARS-CoV 2 Genetic. . .

75

The methodology used in this paper, detailed in Sect. 5.2, also allows identifying the states of the process where the sequences differ the most in relation to the transition probabilities. In this sense, we see that unanimously the state indicated by the set of sequences related to SARS-CoV 2 - P.1 variant is (a, aaa), and it is in this state that the greatest discrepancy occurs between variant P.1 and the sequence MN908947. That is, the transition probabilities {Pˆ (·|(a, aaa)), · ∈ A} differ from MN908947 to P.1 variant (see Table 5.6) and this finding justifies what was discovered in Fig. 5.2.

References Cordeiro, M. T. A., García, J. E., González-López, V. A., & Mercado Londoño, S. L. (2020). Partition Markov model for multiple processes. Mathematical Methods in the Applied Sciences, 43(13), 7677–7691. Hoffmann, M., Arora, P., Groß, R., Seidel, A., Hörnich, B., Hahn, A., Krüger, N., Graichen, L., Hofmann-Winkler, H., Kempf, A., Winkler, M. S., Schulz, S., Jäck, H.-M., Jahrsdörfer, B., Schrezenmeier, H., Müller, M., Kleger, A., Münch, J., & Pöhlmann, S. (2021). SARS-CoV-2 variants B.1.351 and P.1 escape from neutralizing antibodies. Cell, 184(9), 2384–2393. García, J. E., Gholizadeh, R., & González-López, V. A. (2018). A BIC-based consistent metric between Markovian processes. Applied Stochastic Models in Business and Industry, 34(6), 868– 878. García, J. E., & González-López, V. A. (2011). Minimal markov models. In Fourth Workshop on Information Theoretic Methods in Science and Engineering (p. 25). García, J. E., & González-López, V. A. (2017). Consistent estimation of partition Markov models. Entropy, 19(4), 160. García, J. E., González López, V.A., & Tasca, G. H. (2020). Partition Markov model for Covid-19 virus. 4open, 3, 13. García, J. E., González-L´pez, V. A., & Tasca, G. H. (forthcoming). A stochastic inspection about genetic variants of Covid-19 circulating in Brazil during 2020. In Conference Proceedings of ICNAAM 2020. AIP Conference Proceedings. Schwarz, G. (1978). Estimating the dimension of a model. The Annals of Statistics, 6(2), 461–464. Wu, F., Zhao, S., Yu, B., Chen, Y.-M., Wang, W., Song, Z.-G., Hu, Y., Tao, Z.-W., Tian, J.-H., Pei, Y.-Y., Yuan, M.-L., Zhang, Y.-L., Dai, F.-H., Liu, Y., Wang, Q.-M., Zheng, J.-J., Xu, L., Holmes, E. C., & Zhang, Y.-Z. (2020). A new coronavirus associated with human respiratory disease in China. Nature, 579, 265–269.

Chapter 6

Epidemic Management in the Emergency. Protection Measures, Cost and Compliance with Safety Protocols of the Employees of the Health Units. The Case of the General University Hospital of Heraklion “Venizeleio”, the Management and Pandemic of SARS-CoV-2 Anna Kefalaki and George Matalliotakis

6.1 Introduction The SARS-CoV-2 pandemic has plagued humanity since 2020, testing not only health systems around the world but also existing infectious disease mechanisms and the ability of states to deal with it. common challenge (Bachelet & Grandi, 2020). According to the World Health Organization every year, more and more states spend a lot of money on dealing with epidemics, natural disasters and other emergencies, but much less money is spent on prevention (Papadimitriou, 2020). In the face of the pandemic challenge, the need to create new approaches and mentalities, which will result in the transformation of healthcare, is considered more necessary than ever. (Skopelitis, 2020). Health organizations need to work together to develop safety standards and regulations that will work directly in any future pandemic. The purpose of this study is to examine how the issue of the SARS-CoV-2 pandemic was addressed at the level of the health units of the Greek state, with reference to the General Hospital “Venizeleio” in Heraklion at Crete, carried out, sought to investigate whether and to what extent “Venizeleio” complied with the protection measures and safety protocols to deal with the pandemic. The specific purpose of the research was to investigate the views of employees on how to manage the pandemic and its cost.

A. Kefalaki · G. Matalliotakis () Hellenic Open University, Patras, Greece © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 C. H. Skiadas, C. Skiadas (eds.), Quantitative Methods in Demography, The Springer Series on Demographic Methods and Population Analysis 52, https://doi.org/10.1007/978-3-030-93005-9_6

77

78

A. Kefalaki and G. Matalliotakis

6.2 Theoretical Framework The pandemic, according to the WHO, refers to a situation where an extremely pathogenic subtype of the virus affects people, most of whom do not have immune resistance, resulting in the virus spreading rapidly (WHO, 2011). As Professors Mark Woolhouse and Sylvia Gowtage-Sequeria (2010) have stated in their research, there are a number of factors that have contributed to the occurrence of pandemics: (1) The change in agricultural practices and the way people use the land, (2) environmental, climate, social and human geographical changes, (3) persistent pollution of water and food, (4) changing development of international trade and travel, (5) failure to implement public health programs, (6) poor hygiene conditions, (7) the resistance of pathogens, (8) the change of medical practices etc. In recent decades, most pandemics that have occurred have been linked to the respiratory system, often leading patients to acute respiratory syndromes. Typical cases are the H1N1 and H5N1 flu, the Middle East respiratory syndrome, Ebola (Gostin et al., 2016) and SARS (Bootsma & Ferguson, 2007). The SARS-CoV-2 belongs to the family of coronaviruses that led to a pandemic in 2019, which exists to this day (Benvenuto et al., 2020). A key feature of the virus is the high degree of transmissibility between humans and the fact that it has many similarities with SARS at a rate approaching 79% (Chan et al., 2020). The virus is transmitted through the respiratory system, with droplets from sneezing, exhalation or coughing (WHO, 2020). The disease has symptoms such as sore throat, cough, high fever, myalgias and arthralgias, difficulty breathing, physical fatigue, symptoms of gastroenteritis, limited sense of smell and taste. The majority of patients present with the above-mentioned symptoms in a mild form, however, a significant number of patients develop pneumonia, with the consequence that hospitalization is often necessary. In the extreme version of the disease, the sufferer may suffer septic shock and lead to death (Harris, 2000). People who are in the high-risk group are the oldest, those who suffer from heart disease, diabetes, liver or lung problems, cancer or hypertension. From the moment a person becomes ill, the estimated time of onset of the disease ranges from 2 to 14 days (ELINYAE, 2019). Healthcare professionals must strictly adhere to all preventive hygiene measures to reduce HIV infection, and all states have issued guidelines on how to treat those suspected cases. In particular, it has been pointed out that frequent hand washing should be done, people should keep a distance of at least 2 m, especially those who have a cough, to avoid contact with the face. In addition, there should be no sharing of personal hygiene items and household items (Xiao et al., 2020). The WHO has declared SARS-CoV-2 a pandemic, as the virus has spread worldwide, with health systems unable to deal with it. This is the 6th pandemic since 2009. The pandemic has led to social distancing of individuals, so that the infection can be controlled and its spread limited. Most states have declared quarantine measures, travel has almost ceased, schools, workplaces, theaters, shopping malls, stadiums have closed.

6 Epidemic Management in the Emergency. Protection Measures, Cost. . .

79

The Greek state according to ministerial decision no. Y1.G.P.114971-FEK 388/18-2-2014 emphasizes that every hospital unit should have drafted an internal regulation, so that it can control the level of microbial resistance and nosocomial infections. The internal regulation consists of a set of scientifically documented procedures applied by each hospital, in order to prevent and control infections (Ministry of Health, 2014), but also to protect both employees and patients, applying specific health safety standards. Greece in an effort, to deal with the pandemic of SARS-CoV-2 has proceeded to the publication of instructions concerning employees in every sector, especially those working in hospitals. The guidelines emphasize the restriction of the introduction of the virus in the health structure, the cleaning and disinfection of surfaces and objects, the immediate isolation of symptomatic patients, the protection of workers in health facilities (EODY, 2020). According to the available data, it is estimated that Greece from May 2020 to May 2021 had spent 1,000,000,000 euros for hospitalizations, tests and informing the public about the SARS-CoV-2 pandemic. The above amount must be taken into account that during the same period 1,000,000,000 euros were spent in terms of indirect social costs, which was associated with the loss of production and productivity, due to the absence of people from work (Fortune Greece, 2021).

6.3 Purpose of the Research The purpose of the investigation was to demonstrate whether the General Hospital of Heraklion “Venizeleio” complied with the protection measures and safety protocols to deal with the pandemic (General Hospital Venizeleio-Pananeio, 2021).

6.4 Methodology The present research was carried out through the application of quantitative research (Robson, 2010). The research tool used was the structured questionnaire, which consisted of 22 questions, of which the first 7 concerned the demographic data of the sample and the remaining 24 questions referred to the issue of the SARS-CoV2 pandemic. The investigation involved 120 employees of the General Hospital of Heraklion “Venizeleio”, who came from all levels of the hospital hierarchy and from all departments. The research was conducted electronically through Google Forms, due to the SARS-CoV-2 pandemic, in order to limit the possibility of transmitting the virus in the case of the printed form of the questionnaires. The period of the survey was from 14 May 2021 to 7 June 2021. The data were analyzed according to the statistical package SPSS-27 (Sahlas & Bersimis, 2017).

80

A. Kefalaki and G. Matalliotakis

6.5 Research Results The respondents were 120, of whom 67.5% were women and 32.5% were men. 47.5% of the sample were people aged 36–45 years, 48.3% have graduated from higher education TEI, 80% are married people, 48.3% are people employed as nursing staff, while 55% have from 16 to 24 years of service (Graphs 6.1, 6.2, 6.3, 6.4, 6.5, and 6.6). According to the results of the research, it was found that 87.5% of the sample had been informed about the internal regulations for the prevention and control of infections available to the hospital unit that works (Table 6.1). 42.5% of the participants in the research reported that they are regularly informed about the developments in the field of control and prevention of infections by the hospital administration, although in the vast majority (85%) the employees stated that in the hospital where they work does not provide training on a regular basis on issues related to the control and prevention of infections (Tables 6.2 and 6.3). According to the average of the answers, the sample claims that to a great extent, the human resources of the hospital unit were informed and trained, for the safety protocols that had to be observed, for the pandemic of SARS-CoV-2. Regarding the answers of the participants in the survey for the Personal Protection Measures provided to the employees by the hospital unit during the pandemic of SARS-CoV2, 88.3% answered that they were provided with a face mask, gloves, protective uniforms, and glasses, however the 30% of the survey sample reported that the hospital during the pandemic has experienced shortages in face shields (Graph 6.7).

GENDER

32,50%

female male 67,50%

Graph 6.1 Gender

6 Epidemic Management in the Emergency. Protection Measures, Cost. . .

81

SAMPLE AGE 3,33% 5,00% 14,17%

0, bx > 0, kt > 0

(15.7)

Its cumulative distribution function is given by;  F(y) = Φ

y − (ax + bx kt ) σ

 (15.8)

The probability density function of the Gamma-normal Lee-Carter is derived by substituting equations (15.7) and (15.8) into (15.5) to give; & − 12 1 1 g(y) = √ e Γ αx σ 2π

[y−(ax +bx kt )]2 σi 2

'

& &  ' 'αx −1 y − (ax + bx kt ) − ln 1 − Φ σi (15.9)

its likelihood function is given by: L (θ) =

n  i=1

=

 1

−2 1 1 √ e Γ αx σi 2π

 [yi −(ax +bx kt )]2  σi 2

&  ' αx −1 yi − (ax + bx kt ) − ln 1 − Φ σi

 &  'αx −1 n  − 1 2 [yi −(ax +bx kt )]2 n yi − (ax + bx kt ) e 2σi (σi Γ αx )−n (2π)− 2 − ln 1 − Φ σi

i=1

(15.10) The log-likelihood function is expressed as: lnL (θ ) = −n (lnσi + lnΓ αx ) −

n n 1  [yi − (ax + bx kt )]2 + (ln2π) − 2 2σi 2 i=1

(αx − 1)

n  i=1

 &  ' yi − (ax + bx kt ) ln − ln 1 − Φ σi

(15.11)

234

O. E. Adegbilero-Iwari and A. U. Chukwu

Therefore with respect to (15.10), the components of the score vector U(θ ) of the Gamma-normal Lee-Carter are finally of the forms: & ' 1 yi − (ax + bx kt ) (α − 1) + t t σi σi σi $ % x kt ) φ yi −(axσ+b i $

% $

% × yi −(ax +bx kt ) x kt ) 1 − Φ yi −(axσ+b ln 1 − Φ σi i

(15.12)

 &  ' yi − (ax + bx kt ) ln − ln 1 − Φ − nψ (α) σi

(15.13)

' n & n n 1  yi − (ax + bx kt ) 2 (α − 1)  + + Uσi (θ ) = σi σi σi σi i=1$ %

 i=1 yi −(ax +bx kt ) yi −(ax +bx kt ) φ σi σi

% $

% ×$ yi −(ax +bx kt ) yi −(ax +bx kt ) 1−Φ ln 1 − Φ σi σi

(15.14)

& ' kt (α − 1) kt  yi − (ax + bx kt ) Ubx (θ ) = + σi t σi σi % $ yi −(ax +bx kt ) φ  σi  $

%

× yi −(ax +bx kt ) yi −(ax +bx kt ) ln 1 − Φ 1−Φ t σi σi

(15.15)

& ' bx (α − 1)  bx  yi − (ax + bx kt ) Ukt (θ ) = + σi x σi σ x % i $ yi −(ax +bx kt ) φ σi

 $

% × yi −(ax +bx kt ) yi −(ax +bx kt ) 1−Φ ln 1 − Φ σi σi

(15.16)

Uax (θ ) =

Uα (θ ) =

 t

Setting these expressions to zero and solving them simultaneously yields the maximum likelihood estimates (MLEs) of the parameters.

15 Modelling Nigerian Female Mortality: An Application of Four Stochastic. . .

235

15.3 Results 15.3.1 Data Source and Structure The data set used for the study is the age-specific mortality data of Nigerian females. It was obtained from the Global Health Observatory, an arm of the WHO Indicator and Measurement Registry (IMR) (World Health Organisation, 2017). For the mortality data set, the age distribution of the population ranges between less than 1 year old to 85 years and above for the time period 2000–2015. Throughout the study, the number of deaths (dxt ), the central exposures or the population exposed to risk (Ext ) and the mortality rates (mxt ) are arranged in a rectangular array format comprising ages (on the row) x = x1 , x2 , . . . , xk and calendar years (on the columns) t = t1 , t2 , . . . , tn .

15.3.2 Normality Test To test for normality, two non-parametric procedures were used; the KolmogorovSmirnov and Shapiro-Wilks normality tests. While the Kolmogorov-Smirnov procedure was significant at 1%, confirming non-normality with P-values lesser than or equal to 0.000645 across the 19 age-groups, the Shapiro-Wilks procedure confirmed non-normality with P-values lesser than or equal to 0.07518 in 15 age-groups at 10% significance level. Moreover, across the time periods, the Kolmogorov-Smirnov and Shapiro-Wilks procedures were significant at 1%, confirming non-normality with P-values lesser than or equal to 4.18×10−13and 1.58×10−5. This confirms the need for a model allowing for a non-Gaussian distribution structure.

15.3.3 Estimate of Parameters The Gamma-Normal Lee-Carter (GNLC) is compared with three other variants; the Lee and Carter model (LC), the Brouhns model (BR) and the Renshaw-Haberman model (RH). The packages ilc, StMoMo and bbmle in R software were used in obtaining the maximum likelihood estimate of the parameters (Haberman & Butt, 2010; Villegas et al., 2015; Bolker, 2017). Results are displayed in Tables 15.1, 15.2, 15.3, and 15.4.

236

O. E. Adegbilero-Iwari and A. U. Chukwu

Table 15.1 Comparison of the MLEs of Parameter ax under the four models |t|

[95% confidence interval]

10.07468 10.98402

13.73 21.52

0.000 0.000

8.635521 11.51384 9.982885 11.98515

.7338821 .5105169

Source: The Czech Household Panel Survey, 2017 Table 18.5 Alcohol consumption by partnership (females)

Partner Without partner With partner

Delta-method Margin Std. Error

t

P > |t|

[95% confidence interval]

3.227768 4.08223

18.23 25.67

0.000 0.000

2.880593 3.574943 3.770414 4.394046

.177056 .1590233

Source: The Czech Household Panel Survey, 2017 Table 18.6 Binge drinking and partnership (males)

Partnership Without partner With partner

Odds ratio 1 (reference) 1.36

95% confidence interval (1.06–1.75)

Source: The Czech Household Panel Survey, 2017 Table 18.7 Binge drinking and partnership (females)

Partnership Without partner With partner

Odds ratio 1 (reference) 1.23

95% confidence interval (0.93–1.65)

Source: The Czech Household Panel Survey, 2017

In the next step Two-way ANOVA was used to determine whether the difference between alcohol consumption to those with and without a partner is significant. Results are separated for males and females (Tables 18.4 and 18.5). Findings: the difference is significant only for females. Next method that was used was the logistic regression to see the associations between binge drinking and partnership. Binge drinking means in this case consuming more than 60 grams of ethanol at a single occasion. Results are presented in Tables 18.6 and 18.7. As reference group we set ‘without partner’. Unlike expected, binge drinking is more prevalent among those living with a partner (although the difference is not significant among females). Logistic regression was used also for the analysis of associations between being an abstainer and partnership. Being an abstainer means in this case consuming less than 1 gram of ethanol daily. Please find results in Tables 18.8 and 18.9.

288 Table 18.8 Being an abstainer and partnership (males)

K. Svaˇcinová et al. Partnership Without partner With partner

Odds ratio 1 (reference) 0.44

95% confidence interval (0.31–0.65)

Source: The Czech Household Panel Survey, 2017 Table 18.9 Being an abstainer and partnership (females)

Partnership Without partner With partner

Odds ratio 1 (reference) 0.67

95% confidence interval (0.52–0.85)

Source: The Czech Household Panel Survey, 2017

18.5 Conclusion Several demographic trends have created a multiplicity of family structures that complicate the study of family processes (Bumpass, 2004; Leonard & Eiden, 2007). Marital status groups are known to differ in terms of health and mortality in different societes, with non-married persons being in a disadvantaged position compared with married persons (Hu & Goldman, 1990; Joung et al., 1996; Martikainen et al., 2005; Joutsenniemi et al., 2007). Alcohol is a major global contributing factor to death, disease and injury. Alcohol consumption affects not only the individuals, but also their families, partners and society at large (WHO, 2011). Alcohol was found a significant predictor of domestic violence and frequent cause of marriage breakdown in Czechia (377 divorces in 2018). We analyzed two different sources of information: SHARE database (version easySHARE rel. 7.0.0) and the Czech Household Panel Survey 2017. Our findings are identical to results in literature review: the association between alcohol consumption and living without a partner was not confirmed. Unlike expected, living with partner increases alcohol consumption. Results (p-values) from SHARE database using multinomial logistic regression show a statistically significant association. In other outcomes (e.g. data obtained from the Czech Household Panel Survey) significance wasn’t confirmed. The results contradict a common belief that living alone is associated with heavier or riskier drinking. The findings need to be confirmed using mortality data. Acknowledgments This article was supported by the Czech Science Foundation, Grant No. GA ˇ 19-23183Y, on a project titled ‘Alcohol burden in the Czech Republic: mortality, morbidity and CR social context’.

References Bachman, J. G., Wadsworth, K. N., O’Malley, P. M., Johnston, L. D., & Schulenberg, J. E. (1997). Smoking, Drinking, and Drug Use in Young Adulthood: The Impacts of New Freedoms and New Responsibilities (1st ed.). Psychology Press. https://doi.org/10.4324/9780203763797 Börsch-Supan, A., & Gruber, S. (2019). easySHARE. Release version: 7.0.0. SHARE-ERIC. Dataset. https://doi.org/10.6103/SHARE.easy.700

18 Alcohol Consumption and Marital Status in the Czech Republic

289

Brennan, P. L., Moos, R. H., & Kelly, K. M. (1994, December). Spouse of late life problem dmkers. Functioning, coping, responses and family contexts. Journal of Family Psychology, 8(4), 447–457 Bumpass, L. L. (2004). Social change and the American family. Annals of the New York Academy of Sciences. Academy of Sciences, 1038, 213–219. Canham, S. L., Mauro, P. M., Kaufmann, C. N., & Sixsmith, A. (2016). Association of alcohol use and loneliness frequency among middle-aged and older adult drinkers. Journal of Aging and Health, 28(2), 267–284. https://doi.org/10.1177/0898264315589579. Available at: https:// www.ncbi.nlm.nih.gov/pmc/articles/PMC4681688/ Crisp, B. R., & Barber, J. G. (1995, June). The Drinker’s partner distress scale: An instrument for measuring the distress causal by drinkers to their partners. International Journal of the Addictions, 30(8), 1009–1017. Czech Statistical Office. (2019). Divorces: By cause of marriage breakdown on the part of male and female and by educational attainment of divorces (Demographic yearbook 2010–2018). CZSO. (In Czech). Available at: https://www.czso.cz/csu/czso/demographic-yearbook-of-theczech-republic-2010-v1lnbg8yv8 Dohnal, D., Hokr Miholová, P., Šprincová, V., & Domesová, S. (2017). Analysis of incidence and latency of domestic violence in relationships. Prague: Office of the Government of the Czech Republic Herttua, K., Martikainen, P., Vahtera, J., & Kivimäki, M. (2011). Living alone and alcohol-related mortality: A population-based cohort study from Finland. PLoS Medicine, 8(9), e1001094. https://doi.org/10.1371/journal.pmed.1001094. Available at: https://www.ncbi.nlm.nih.gov/ pmc/articles/PMC3176753/ Hu, Y., & Goldman, N. (1990). Mortality Differentials by Marital Status: An International Comparison. Demography, 27(2), 233–250. https://doi.org/10.2307/2061451 Joung I. M., Glerum J. J., van Poppel F. W., et al. The contribution of specific causes of death to mortality differences by marital status in the Netherlands, European Journal of Public Health, 1996, vol. 6(pg. 142–149) Joutsenniemi, K., Martelin, T., Kestilä, L., Martikainen, P., Pirkola, S., & Koskinen, S. (2007, September). Living arrangements, heavy drinking and alcohol dependence. Alcohol and Alcoholism, 42(5), 480–491. https://doi.org/10.1093/alcalc/agm011 Leigh Hunt, N. (2017). An overview of systematic reviews on the public health consequences of social isolation and loneliness.https://www.sciencedirect.com/science/article/abs/ pii/S0033350617302731?via%3Dihub Leonard, K. E., & Rothbard, J. C. (1999). Alcohol and the marriage effect. Journal of Studies on Alcohol Supplement, 13, 139–146. Leonard, K. E., & Eiden, R. D. (2007). Marital and family processes in the context of alcohol use and alcohol disorders. Annual Review of Clinical Psychology, 3, 285–310. https://doi.org/ 10.1146/annurev.clinpsy.3.022806.091424 Marshal, M. P. (2003). For better or for worse? The effects of alcohol use on marital functioning. Clinical Psychology Review, 23(7), 959–997. https://doi.org/10.1016/j.cpr.2003.09.002 Martikainen, P. et al. (2005). Socioeconomic Position and Self-Rated Health: The Contribution of Childhood Socioeconomic Circumstances, Adult Socioeconomic Status, and Material Resources. American Journal of Public Health 95(8), 1403–1409. https://doi.org/10.2105/ AJPH.2004.047969 Martinková, M., Slavˇetínský, V., & Vlach, J. (2014). Selected problems in the field of domestic violence in the Czech Republic. Prague: Institute for Criminology and Social Prevention. McCrady, B. S., & Epstein, E. E. (1995). Marital therapy in the treatment of alcohol problems. In N. S. Jacobson & A. S. Gurman (Eds.), Clinical handbook of couple therapy (Vol. 1995, pp. 369–393). Guilford Press. Moskalewicz, J., & Sieroslawski, J. (2010). Drinking Population Surveys-Guidance Document For Standardized Approach. European Commission Health Consumer Protection Directorate General

290

K. Svaˇcinová et al.

ˇ Mravˇcík, V., Chomynová, P., Nechanská, B., Cerníková, T., & Csémy, L. (2019a). Alcohol use and its consequences in the Czech Republic. Central European Journal of Public Health, 27(Supplement), S15–S28. https://doi.org/10.21101/cejph.a5728. PubMed PMID: 31901189. Available at: https://cejph.szu.cz/artkey/cjp-201988-0003_alcohol-use-and-its-consequencesin-the-czech-republic.php Mravˇcík, V. et al. (2019b). Annual Report on Drug Situation in the Czech Republic in 2019. Prague: Office of the Government of the Czech Republic Nešpor, K., & Csémy, L. (2005). How many pathological players are there in the Czech Republic? Czech and Slovak Psychiatry, 101(8), 433–435. Paolino, T. J., & McCrady, B. S. (1977). The alcoholic marriage: Alternative perspectives. Grune & Stratton. Pikálková, S., Podaná, Z., & Buriánek, J. (2015). Women as victims of partner violence. Prague, Elephant Richard, A., Rohrmann, S., Vandeleur, C. L., Schmid, M., Barth, J., et al. (2017). Loneliness is adversely associated with physical and mental health and lifestyle factors: Results from a Swiss national survey. PLoS One, 12(7), e0181442. https://doi.org/10.1371/journal.pone.0181442 Saxena, S., Sharma, R., & Maulik, P. (2003). Impact of alcohol use on poor families: A study from North India. Journal of Substance Use, 8(2), 78–84. Schonfeld, L., & Dupree, L. W. (1991). Antecedents of drinking for early- and late-onset elderly alcohol abusers. Journal of Studies on Alcohol, 52(6), 587–592. The Czech Household Panel Survey (2017). Czech Social Science Data Archive. Available at: https://archiv.soc.cas.cz/en/ Varga, S., & Piko, B. F. (2015). Being lonely or using substances with friends? A cross-sectional study of Hungarian adolescents’ health risk behaviours. BMC Public Health, 15, 1107. https:// doi.org/10.1186/s12889-015-2474-y. Available at: https://www.ncbi.nlm.nih.gov/pmc/articles/ PMC4637146/ WHO. (2011). Global status report on alcohol and health. WHO. ISBN 978-92-4-156415-1. Available at: https://www.drugsandalcohol.ie/14675/1/Global_status_report_on_alcohol_and_ health.pdf

Chapter 19

Drug Addiction Mortality Among Young Muscovites: Official Rates and Actual Scale G. Semyonova Victoria, E. Ivanova Alla, P. Sabgayda Tamara, V. Zubko Aleksandr, S. Gavrilova Natalia, N. Evdokushkina Galina, and G. Zaporozhchenko Vyacheslav

Narcotics and drug addiction mortality are a global challenge: according to the latest World Drug Report published by the United Nations Office on Drugs and Crime (UNODC), about 35 million people worldwide suffer from substance use disorders and need treatment. Approximately 271 million people, or 5.5% of the world’s population aged 15– 64, used drugs in 2017. While these figures are close to the 2016 estimates, a longerterm analysis shows that the number of people using drugs is currently 30% higher compared to 2009. The most dangerous consequences are caused by opioid use, and, according to the Report, the situation looks more alarming than previously thought: for example, the number of opiate users is estimated at 53 million – 56% more than previous

G. Semyonova Victoria · E. Ivanova Alla · P. Sabgayda Tamara · N. Evdokushkina Galina Institute for Demographic Research – Branch of the Federal Center of Theoretical and Applied Sociology of the Russian Academy of Sciences, Moscow, Russia Federal Research Institute for Health Organization and Informatics of Ministry of Health of the Russian Federation, Moscow, Russia V. Zubko Aleksandr () Institute for Demographic Research – Branch of the Federal Center of Theoretical and Applied Sociology of the Russian Academy of Sciences, Moscow, Russia S. Gavrilova Natalia Institute for Demographic Research – Branch of the Federal Center of Theoretical and Applied Sociology of the Russian Academy of Sciences, Moscow, Russia Academic Research Centers, NORC, University of Chicago, Chicago, IL, USA G. Zaporozhchenko Vyacheslav Federal Research Institute for Health Organization and Informatics of Ministry of Health of the Russian Federation, Moscow, Russia © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 C. H. Skiadas, C. Skiadas (eds.), Quantitative Methods in Demography, The Springer Series on Demographic Methods and Population Analysis 52, https://doi.org/10.1007/978-3-030-93005-9_19

291

292

G. Semyonova Victoria et al.

estimates. Of particular note, 2/3 out of 585,000 drug users died due to opiates in 2017. A special focus should be put on the synthetic opioid overdose crisis in the North America: in 2017, the United States registered over 47,000 opioid overdose deaths, 13% over the previous year, and 4000 opioid overdose deaths in Canada, a 33% increase since 2016 (World Drug Report. UNODC, 2020). Please, remember that drug addiction mortality is not limited to deaths due to poisoning by narcotics: the injecting drug addiction is associated with HIV/AIDS and Hepatitis C. Furthermore, drug addiction dramatically increases the risk of suicide (World Drug Report. UNODC, 2020). Besides direct demographic consequences, there is an inevitable social cost of drug addiction, resulting in indirect demographic loss. Indirect loss due to drug addiction is caused by an imminent criminal lifestyle of drug users, which is associated with both illegal drug distribution, and increased risks of violence both inside the community and among the inner circle (family and friends), as well as random people at risk of becoming victims of the drug addict (Report on Drug Situation in the Russian Federation, State Anti-drug Committee, 2019). The following circumstances are also highly alarming in the demographic context: first, the multiplier effect of drug addiction: according to researchers, every addict engages about 15 people in using drugs. Second, this is a noticeable “rejuvenation” of the drug use debut: debut at 15–17 years 10 years ago versus 12– 17 years now. Third, life expectancy of drug addicts is reduced by 3–4 years (Report on Drug Situation in the Russian Federation, State Anti-drug Committee, 2019). Most researchers consider young people (persons aged 16–30) the largest segment of drug users: they account for about 60% of this community, which makes young people the main risk group for the formation and further spread of drug addiction in a particular region (World Drug Report. UNODC, 2020). The situation in Russia looks quite contradictory: on the one hand, according to the UN data, in 2019 Russia was among the top three countries that accounted for almost half of the 11.3 million injecting drug users worldwide (World Drug Report. UNODC, 2020), on the other hand, according to the Russian Federal State Statistics Service (Rosstat) data, Russia registered 4585 overdose deaths that year (the total number of deaths from poisoning by narcotics and mental disorders due to psychoactive substance use). The situation in the capital looks rather similar: according to the 2019 Report on Drug Situation in the Russian Federation, the drug situation in Moscow and St. Petersburg is defined as “pre-crisis” (Report on Drug Situation in the Russian Federation, State Anti-drug Committee, 2019), on the other hand, according to the Rosstat data, the number of overdose deaths in the capital equaled to 608 in 2019 was, including 85 people aged 15–29. It should be noted here that, according to the Bureau of Forensic Medical Examination of the Moscow Healthcare Department, the number of deaths due to poisoning by narcotics and psychotropic substances in 2019 added up to1 239 people, which is more than doubles the Rosstat data,

19 Drug Addiction Mortality Among Young Muscovites: Official Rates. . .

293

transferring the drug situation in the capital into a crisis one (Report on drug situation monitoring in Moscow, 2019). The purpose of the study is to estimate the scale of drug addiction mortality among young people against the background of mortality among the Moscow working-age population. Hypothesis of the study. A mismatch between the expert evaluation of the drug situation in Moscow as a crisis, and official statistics on mortality indicating Ok-ness, suggests that some of the fatal loss due to drug-related causes is underestimated. “Cardiomyopathy, unspecified” and “Symptoms, signs and ill-defined conditions” are considered as a latent reservoir for the underestimated loss. Materials and methods. The authors used the Rosstat data on mortality in Russia calculated in the FAISS-Potential1 system, as well as RFU-EMIAS2 data (July–December 2018 – January–June 2019). Standardized mortality rates by causes of death in the selected age groups were calculated. The direct method of standardization, and European standard population were used. The deceased of unknown age were pre-distributed in proportion to the number of deaths by age groups in ages over 1 year of life. Results According to ICD-10, only two groups of death causes have a clearly defined drug etiology: Poisoning by and exposure to narcotics and psychodysleptics [hallucinogens], not elsewhere classified, both accidental and undetermined intent (X42, Y12) and Mental and behavioral disorders due to psychoactive substance use (F11, F12, F14, F16).

19.1 Poisoning by Narcotics and Mental Disorders Due to Psychoactive Substance Use Unfortunately, poisoning by narcotics, as a separate nosological unit, has been identified in the short nomenclature of causes of death adopted in Russia only since 2011. In addition to accidental poisoning by narcotics, ICD-10 includes Poisoning by and exposure to narcotics and psychodysleptics [hallucinogens], not elsewhere classified, undetermined intent (Y12). This section covers events where available information is insufficient to enable a medical or legal authority to make a distinction between accident, self-harm and assault. We’d like to note that out of all

1

The factographic information and reference system (FAISS-Potential) is a system of mortality analysis indicators in Russia and the subjects of the Russian Federation, calculated on the basis of the Rosstat official data on mortality. 2 “Death registration in the Moscow Unified Medical Information and Analytical System” (RFUEMIAS) database contains data on death events registered by medical organizations of the Moscow Healthcare Department.

294

G. Semyonova Victoria et al.

injuries classified as events, undetermined intent (Y10-Y34), inclusion of poisoning by narcotics in this category seems to be most objective: indeed, death from drug overdose can be the result of an accident, suicide, or homicide. Proceeding form the above mentioned, it is only logical to evaluate the situation mainly on the basis of the total mortality from poisoning by narcotics, which includes both accidental poisoning (X42) and poisoning with undetermined intent (Y12). Analyzing total mortality from poisoning by narcotics among young Muscovites, we’d like to point out that in the first half of the 2010s, the indicator decreased by 30% in males, and this decrease was due to accidental poisoning (by 36.5%) against the background of the stagnated mortality from poisoning with undetermined intent; the indicator fluctuated within the range of 0.1–0.2 per 100,000 in females which was determined by accidental poisoning. However, in the second half of the 2010s, the situation changed dramatically: for example, within 1 year (2015–2016), the total mortality from poisoning by narcotics among young Muscovites increased 4.6 fold in males and 2.5 fold in females, which was primarily due to accidental poisoning (a 5.6 and 4-fold increase, respectively), while mortality from poisoning with undetermined intent in males doubled. In females, increase in mortality from these causes was postponed for a year: in 2016–2017, the indicator increased five-fold. In general, in the second half of the 2010s, total mortality from poisoning by narcotics among young Muscovites increased almost 12-fold in males and 7.5-fold in females, with loss due to poisoning with undetermined intent growing faster in males (13.5 versus 11.2 times), and due to accidental poisoning in females (8 versus 7 times). It should be particularly noted here that trends in mortality from poisoning by narcotics among youth in Russia fundamentally differed from those in Moscow: following the 2014 peak, the male indicators declined, followed by stagnation after 2016 due to accidental poisoning against the background of a slight increase in deaths from poisoning with undetermined intent. In females, dynamics in mortality from poisoning by narcotics was characterized by immature tendencies, however, the overall trend in the 2010s was positive. Describing mortality from poisoning by narcotics, it is necessary to mention a common pattern in both the capital and Russia as a whole: not always consistent, but rather sustainable increase in the contribution of poisoning with undetermined intent to the total mortality from poisoning by narcotics: the current contribution in Moscow added up to 32.5% in males and 46.7% in females versus 20% and zero in 2011, respectively (in Russia, this indicator increased from 21.5% and 30.8% to 40% and 50%, respectively, during this period). In general, results of 2011–2019 turned out to be extremely negative for young Muscovites: while the Russian total mortality from poisoning by narcotics decreased by almost a quarter (23.1%) in males and over two-fold (2.2 times) in females, the Moscow mortality rates increased 8.3 fold and 15 fold, respectively, furthermore, these negative changes have develop during the last 5 years.

19 Drug Addiction Mortality Among Young Muscovites: Official Rates. . .

295

Males Russia

Moscow 9

8

8 7

6

6 5 4

4 3 2

2

1 0 2011

2013

2015

2017

2019

0 2011

2013

2015

2017

2019

2017

2019

Females Moscow

Russia

1,8 1,6 1,4 1,2 1 0,8 0,6 0,4 0,2 0 2011

2013

2015

2017

2019

2 1,8 1,6 1,4 1,2 1 0,8 0,6 0,4 0,2 0 2011

2013

2015

accidental poisoning

accidental poisoning

poisoning, undetermined intent

poisoning, undetermined intent

poisoning, total

poisoning, total

Fig. 19.1 Mortality from poisoning by narcotics among population aged 15–29 in Moscow and Russia in the 2010s (standardized rates per 100,000)

Such dynamics resulted in a 6.4 – and nine-fold advantage among young Muscovites in terms of total mortality from poisoning by narcotics in 2011 compared to the all-Russia rates, noted in 2011, which was followed by a 86.7% loss in males and 2.7-fold loss in females in 2019; these negative consequences for the capital started to develop after 2015 and were determined by both accidental poisoning and poisoning with undetermined intent (Fig. 19.1). To evaluate the loss scale due to poisoning by narcotics, we should bear in mind that according to the Rosstat data, 85 Muscovites aged 15–29 (74 men and 11 women) died from poisoning by narcotics in 2019. As already mentioned, in addition to poisoning, mental and behavioral disorders due to psychoactive substance use have a clear reference to drug etiology. Table 19.1 shows that mortality from these causes among young Muscovites is extremely

296

G. Semyonova Victoria et al.

Table 19.1 Dynamics in mortality from mental and behavioral disorders due to psychoactive substance use among working-age population in Moscow and Russia in the 2000s (standardized rates per 100,000)

Years 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019

Moscow 15–29 years m f 0.1 0.0 0.3 0.2 0.1 0.0 0.0 0.1 0.1 0.0 0.0 0.0 0.2 0.1 0.1 0.2 0.5 0.1 0.1 0.2 0.7 0.1 0.4 0.2 0.1 0.1 0.3 0.0 0.2 0.1 0.1 0.0 0.3 0.0 0.8 0.3 0.3 0.4 0.5 0.0

30–44 years m f 0.0 0.1 0.5 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.2 0.0 0.2 0.0 0.5 0.0 0.4 0.1 0.7 0.0 0.9 0.1 0.7 0.3 0.6 0.1 1.1 0.3 0.6 0.2 0.4 0.1 2.4 0.3 4.1 1.1 3.1 0.7 2.9 0.7

45–59 years m f 0.2 0.0 0.1 0.0 0.1 0.0 0.0 0.0 0.1 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.2 0.0 0.2 0.0 0.1 0.0 0.1 0.0 0.1 0.1 0.0 0.0 2.2 0.6 3.1 0.7 2.2 0.5 1.2 0.4

Russia 15–29 years m f 0.4 0.1 0.3 0.1 0.1 0.1 0.1 0.1 0.1 0.0 0.2 0.0 0.2 0.1 0.2 0.1 0.2 0.1 0.1 0.0 0.2 0.0 0.1 0.1 0.1 0.0 0.1 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1 0.0 0.0 0.1 0.0 0.0

30–44 years m f 0.4 0.1 0.5 0.1 0.1 0.1 0.1 0.1 0.2 0.0 0.2 0.1 0.4 0.1 0.3 0.1 0.4 0.0 0.4 0.1 0.5 0.1 0.4 0.1 0.4 0.1 0.3 0.1 0.3 0.1 0.2 0.1 0.4 0.1 0.6 0.1 0.4 0.1 0.3 0.1

45–59 years m f 0.1 0.0 0.1 0.0 0.1 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1 0.0 0.1 0.0 0.0 0.0 0.1 0.0 0.1 0.0 0.1 0.0 0.1 0.0 0.1 0.0 0.1 0.0 0.0 0.0 0.2 0.1 0.3 0.1 0.2 0.0 0.1 0.0

low (for example, in 2019 6 men aged 15–29 died from these causes). However, according to ICD-10, mental disorders cannot be documented as the underlying cause of death (Crump et al., 2013; Drapkina et al., 2019; Hällgren et al., 2019; Kakorina et al., 2015; Kakorina & Kazakovtsev, 2013; Lumme et al., 2016; Mackenbach et al., 2014; Sabgaida et al., 2014; Saxena 2018; Solovyov et al., 2015; Starace et al., 2018; Walker et al., 2015). However, it should be noted here that 2015 turned out to be a reference point for both mental disorders due to psychoactive substance use and poisoning by narcotics: over the next 3 years (2015–2017), mortality among young Muscovites increased eight-fold in males and from zero to 0.3 per 100,000 in females, followed by a decline in indicators.

19 Drug Addiction Mortality Among Young Muscovites: Official Rates. . .

297

19.2 Cardiomyopathy, Unspecified Dynamics in mortality from diseases of the circulatory system raises a lot of questions: within the last 5 years (2015–2019) against the background of all programs aimed at reducing cardiovascular mortality in Russia mortality from diseases of the circulatory system among young Muscovites increased 3.3-fold in males and almost by one third (31.8%) in females versus 25.3% and 21.2% reduction in Russia (Fig. 19.2). Please, note that a fairly sustainable positive trend in mortality among young Muscovites formed in 2007–2015, when Moscow rates decreased five-fold in males and 3.4-fold in females (versus a 38.8% and 27.5% reduction in Russia), was suddenly followed by a rapid growth: over the next 3 years (2015–2017) the Moscow rates increased 4.4-fold in males and three fold in females versus a 10.6%-fold decrease in mortality among the Russian youth. This was followed by a sudden drop in mortality among Muscovites with a significantly faster rate compared to the all-Russia indicators adding up to 26.3% in males and 2.3 fold in females versus 16.4% and 11.9%, respectively. It is rather significant that these fluctuations in mortality from diseases of the circulatory system among young Muscovites are accounted for by cardiomyopathy, unspecified (I42.9), the dynamics of which perfectly coincides with the dynamics in cardiovascular mortality (Please, compare Figs. 19.2 and 19.3). We’d like to emphasize it here that within 3 years (2015–2017), mortality from cardiopathy, unspecified among young Muscovites increased 15.2 and 19.5-fold, respectively, and even in 2019, following a sharp decline within the last years under study, Moscow is among the top ten outsiders in terms of youth mortality from cardiomyopathy, unspecified, ranking 3rd in males and 8th in females among the Russian regions. Therefore, Males

Females

50

12

40

10 8

30

6 20

4

10

2

0

0 2000 2003 2006 2009 2012 2015 2018 Moscow

Russia

2000 2003 2006 2009 2012 2015 2018 Moscow

Russia

Fig. 19.2 Mortality from diseases of the circulatory system among population aged 15–29 in Moscow and Russia in the 2000s (standardized rates per 100,000)

298

G. Semyonova Victoria et al. Males

Females 4,5

18 16

4

14

3,5

12

3

10

2,5

8

2

6

1,5

4

1

0

0,5

20 11 20 12 20 13 20 14 20 15 20 16 20 17 20 18 20 19

2

Moscow

Russia

0 2011

2013

2015

Moscow

2017

2019

Russia

Fig. 19.3 Mortality from cardiomyopathy, unspecified among population aged 15–29 in Moscow and Russia in the 2010s (standardized rates per 100,000)

Moscow is characterized by an abnormally high contribution of cardiomyopathy, unspecified to the overall mortality from diseases of the circulatory system among young people (60% in males and 62.1% in females in 2019 versus 15.7% and 11.5% in Russia, respectively). There is a logical question: what is behind these fluctuating processes, unpredictable by the previous dynamics in mortality, registered among young Muscovites? Do these processes reflect the actual health deterioration in young Muscovites (increased mortality from cardiomyopathy) coupled with deteriorated social environment in the capital (increased mortality from poisoning by narcotics)?

19.3 Symptoms, Signs and Ill-Defined Conditions It looks like evolution of mortality from “Symptoms, signs, and ill-defined conditions” that are so blurry and, at first glance, are not likely to have any burden on society and therefore are outside of the decision-makers’ focus can help answer this question. We’d like to point it out that in working-age groups, including young people, mortality from symptoms, signs and ill-defined conditions in both Russia and Moscow was primarily determined by a single diagnosis “Ill-defined and unknown cause of mortality” (R99). Please, note that, until 2015 Moscow was a permanent outsider in terms of mortality from ill-defined conditions among the Russian regions: thus, in 2015 the Moscow rate was the highest one in males, while the female worst rate was registered in the Chukotka Autonomous Region. However, after 2015, there was a dramatic drop in mortality from these blurred causes among young people, which continued until 2017: within 3 years, mortality from ill-defined conditions in the

19 Drug Addiction Mortality Among Young Muscovites: Official Rates. . .

Males

299

Females 16

60

14

50

12

40

10 8

30

6

20

4

10

2

0

0 2000 2003 2006 2009 2012 2015 2018 Moscow

Russia

2000 2003 2006 2009 2012 2015 2018 Moscow

Russia

Fig. 19.4 Mortality from ill-defined conditions among population aged 15–29 in Moscow and Russia in the 2000s (standardized rates per 100,000)

capital decreased over seven-fold in both males and females versus a 43.9% and 30.8% decrease in Russia, while during the last years under study (2017–2019) it increased five-fold in males and 5.3 fold in females versus a 25% and 16.7% rise in Russia, respectively (Fig. 19.4). It is rather important that it is 2015 and 2017 that turned out to be the reference points fundamentally changing trends for both ill-defined conditions, cardiomyopathy unspecified, and mental and behavioral disorders due to psychoactive substance use (please, note that mortality from mental and behavioral disorders due to psychoactive substance use continued to grow after 2017 as well). This timing makes it possible to assume with all statistical probability, that these processes are not the result of a sharp health deterioration in young Muscovites on the one hand, and a dramatically increased prevalence of drug addiction with fatal consequences, on the other, but rather a statistical procedure that makes it possible to specify the real causes of mortality from ill-defined conditions, namely: poisoning by narcotics and cardiomyopathy, unspecified.

19.4 Mortality from Confirmed and Suspected Drug-Related Causes in the Working-Age Population Another question that cannot but arise: are these processes in Moscow specific and typical of young people only? A detailed analysis showed that such transformations were registered throughout the entire working-age population in Moscow. For example, positive trends in total mortality from poisoning by narcotics among Muscovites aged 30–44 that were

300

G. Semyonova Victoria et al.

Males Russia

Moscow 25

22 20 18

20

16 14

15

12 10

10

8 6

5

4 2

0 2011

2013

2015

2017

2019

0 2011

2013

2015

2017

2019

2017

2019

Females Russia

Moscow 2,5

6 5

2

4 1,5

3 2

1

1

0,5

0 2011

2013

2015

2017

2019

0 2011

2013

2015

accidental poisoning

accidental poisoning

poisoning, undetermined intent

poisoning, undetermined intent

poisoning, total

poisoning, total

Fig. 19.5 Mortality from poisoning by narcotics among population aged 30–44 in Moscow and Russia in the 2010s (standardized rates per 100,000)

formed in Moscow during the first half of the 2010s against the background of the increased Russian indicators, were replaced by a dramatic rise in mortality, within a year (2015–2016) the male mortality among Muscovites increased 13.8 fold while the female mortality of their peers increased 52-fold. Indicators showed a certain decline after 2006, however, in general, mortality from poisoning by narcotics among Muscovites aged 30–44 in 2015–2019 increased 13.4-fold in males and 28fold in females versus a 17.1% increase in the Russian men and a 5% decrease in their peers (Fig. 19.5). It is characteristic that in this age group, the growth rate of mortality from poisoning with undetermined intent also outstrips the growth rate of mortality from accidental poisoning, adding up to 36-fold in 2015–2019 versus 10.1 fold in males and an increase from zero to 0.9 per 100,000 versus a 19-fold increase in females.

19 Drug Addiction Mortality Among Young Muscovites: Official Rates. . .

301

Similar to young people, the share of poisoning with undetermined intent in the total poisoning by narcotics among people aged 30–44 is increasing, adding up to 33.6% and 32.1%, respectively, in 2019. Similar patterns were observed among older working-age people (45–59 years). Particularly noteworthy is the 2015–2016 peak with a seven and nine-fold increase in mortality from poisoning by narcotics among young Muscovits within a year followed by stagnation processes in males and a dramatic decline in females. Mortality from poisoning with undetermined intent in males was advancing at an outstripping rate during the last 5 years (11-fold versus 5.3 fold). Mortality from accidental poisoning among females in 2019 returned to the 2015 level (0.1 per 100,000), while mortality from poisoning with undetermined intent increased from zero to 0.1 per 100,000. Therefore, the current share of poisoning with undetermined intent in the total poisoning by narcotics equals to 40.7% in males aged 45–59 and a half – in their peers (Fig. 19.6). Due to such dynamics, Moscow, which had a multiple gain in terms of mortality from poisoning by narcotics among people age 30–44 in the first half of the 2000s compared to the all-Russia rates, lost over a quarter in males and almost 1.5-fold in females by 2019. By 2019 Moscow also lost its gains in older working ages, reported in the first half of the 2010s: Moscow’s indicators equaled to the all-Russia ones. Please, note that according to the Rosstat data, 412 people aged 30–59 (362 men and 50 women) died from poisoning by narcotics in 2019 in Moscow. As to mental disorders due to psychoactive substances use, similar to young people, mortality from these causes among other age and gender groups of the Moscow working-age population was either zero or extremely low until the mid2010s, developing significant rates after 2015 only. It should be emphasized here that the maximum indicators in all gender and age groups were registered in 2017, followed by a reduction trend. In Russia, the female indicators were extremely low throughout the entire period under study (not exceeding 0.1 per 100,000), while the highest male indicators were registered in people aged 30–44 ranging from 0.2 to 0.4 per 100,000. Therefore, during the last years under study, a significant rise in the Moscow indicators compared to the Russian ones (Table 19.1) has been reported. In general, according to official statistics, 74 people aged 30–59 (57 men and 17 women) died from mental disorders due to psychoactive substance use in Moscow in 2019. It should be noted here that dynamics in mortality from cardiomyopathy unspecified among Muscovites of middle and older working ages (unlike young people, in these ages one can already refer to realization of the accumulated behavioral risks) does not differ from that among young people. Thus, the indicators in both people aged 30–44 and 45–59 were decreasing until 2015, followed by an abrupt growth: within a year (2015–2016), mortality among people aged 30–44 increased 6.9 and 5.3-fold, and 3.6 and 4.3-fold among people aged 45–59, respectively. Mortality among people aged 30–44 continued to rise until 2017, followed by a decline in recent years. Negative trends in Muscovites aged 45–59 persisted until 2018. (a 8.2

302

G. Semyonova Victoria et al.

Males Moscow

Russia

3,5

3

3

2,5

2,5

2

2 1,5

1,5 1

1

0,5

0,5

0 2011

2013

2015

2017

0

2019

2011

2013

2015

2017

2019

2017

2019

Females Moscow

Russia 0,3

1 0,8

0,2

0,6 0,4

0,1

0,2 0 2011

2013

2015

2017

2019

0 2011

2013

2015

accidental poisoning

accidental poisoning

poisoning, undetermined intent

poisoning, undetermined intent

poisoning, total

poisoning, total

Fig. 19.6 Mortality from poisoning by narcotics among population aged 45–59 in Moscow and Russia in the 2010s (standardized rates per 100,000)

and 9.6-fold increase in indicators, respectively) with the decrease registered during the last year under study only (Fig. 19.7). In General, over the last 5 years mortality from cardiomyopathy unspecified among Muscovites aged 30–44 increased 11.3-fold in males and 7.8-fold in females, and 5.8 and 7.7-fold among Muscovites aged 45–59, respectively, against the background of rather sustainable positive trends in the Russian population in similar ages. It is very revealing that mortality from ill-defined conditions in these age groups of the Moscow population mirrors dynamics in mortality from poisoning by narcotics, especially from cardiomyopathy, unspecified; similar to young people, 2015 and 2017 serve the reference points: for example, a drastic decline in mortality from ill-defined conditions resulted a 11.9 and 7.2-fold reduction among people

19 Drug Addiction Mortality Among Young Muscovites: Official Rates. . .

303

Males Moscow

Russia

70

35

60

30

50

25

40

20

30

15

20

10

10

5

0 2011

2013

2015

2017

2019

0 2011

2013

2015

2017

2019

2017

2019

Females Moscow

Russia 12

25

10

20

8 15 6 10

4

5

2

0

0 2011

2013 30-44 years

2015

2017

2019

45-59 years

2011

2013 30-44 years

2015

45-59 years

Fig. 19.7 Mortality from cardiomyopathy, unspecified among population aged 30–59 in Moscow and Russia in the 2010s (standardized rates per 100,000)

aged 30–44, and a 6.6 and 7-fold reduction among people aged 45–59 in 2015–2017 respectively, after 2017 the indicators simultaneously increased 4.6 and 2.7-fold in middle ages and 2.8 and 3-fold in older working ages (Fig. 19.8). Please, note that the overall trends in the Russian mortality had the same vector as in Moscow (an over 40% decline among people aged 30–44 in both males and females in 2015–2017 followed by a 28% and 17.9% increase in 2017–2019; a 27.4% and 32.6% decrease among people aged 45–59 followed by a 25.7% and 35.8% rise), however, the magnitude of these changes in Russia is far from being comparable with Moscow. Furthermore, the size of the Moscow population is close to 10% of the Russian population, therefore, the Moscow changes cannot but affect the Russian trends.

304

G. Semyonova Victoria et al.

Males Moscow

Russia

200

140

180

120

160

100

140 120

80

100

60

80 60

40

40

20

20 0 2000 2003

2006

2009 2012

0

2015 2018

2000 2003

2006

2009 2012

2015 2018

Females Russia

Moscow 30

50 45

25

40 35

20

30 15

25 20

10

15 10

5

5 0

0 2000

2003

2006

30-44 years

2009

2012

2015

2018

45-59 years

2000

2003

2006

30-44 years

2009

2012

2015

2018

45-59 years

Fig. 19.8 Mortality from symptoms, signs and ill-defined conditions among population aged 30– 59 in Moscow and Russia in the 2000s (standardized rates per 100,000)

Thus, a conclusion can be made that transformations in mortality among young Muscovites are not specific: these patterns are registered throughout the entire working-age population of Moscow and are characterized by increased mortality from poisoning by narcotics, mental disorders due to psychoactive substance use, and especially from cardiomyopathy unspecified, against the background of a mirror decrease in mortality from symptoms, signs, and ill-defined conditions.

19 Drug Addiction Mortality Among Young Muscovites: Official Rates. . .

305

19.5 Discussion Discussion of the study results is focused on the following three fundamental issues: first, identification of cardiomyopathy unspecified in the context of the drug addiction consequences; second, peculiarities of death registration in working ages (including young people) in Moscow; third, general approaches to registering deaths related to drug addiction, stipulated by ICD-10. Discussing etiology of cardiomyopathy unspecified, one should remember that cardiomyopathy unspecified is the vaguest out of all cardiovascular diagnoses: the term “cardiomyopathy” per se was proposed in 1957 only for myocardial diseases of unknown origin (Blagova & Nedostup, 2017), and classification of pathologies included in this group is still the subject of acute discussions among cardiologists and clinicians (Gorgaslidze et al., 1993; Moiseev & Kiyakbaev, 2009; Blagova & Nedostup, 2017; Kaktursky, 2000). Discussion of this problem is far beyond the scope of this research, however it should be pointed out that experts, discussing the etiology of cardiomyopathies, agree that its primary forms are determined by endogenous (genetic) factors, while secondary – by exogenous factors, with alcohol being the most common exogenous factor. However, alcoholic cardiomyopathy is allocated into a separate nosological unit (ICD-10 code “I42. 6”). Mortality from this cause among young Muscovites (like in Russia) was steadily decreasing in the 2010s, and its dynamics, in contrast to the dynamics in mortality from alcohol poisoning, was not associated with mortality from ill-defined conditions (Semyonova et al., 2019). However, numerous studies in recent years have shown that it is drug use that leads to the formation of cardiovascular pathologies, especially cardiomyopathies, both infectious (due to drug injections) and non-infectious ones (Brigden, 1957; Frustaci et al., 2015; Oh et al., 2019; Radunski et al., 2017; Sorokina, 2018; Sorrentino et al., 2018; Stankowski et al., 2015). The second fact suggesting drug-related etiology behind the increased mortality from cardiomyopathy unspecified in Moscow is hospitalization due to the relevant causes. If the increase in mortality from cardiomyopathy unspecified was due to natural causes, this catastrophic deterioration in the health of Muscovites, especially in young people, would have resulted in their intensive treatment, especially against the background of programs aimed at reducing cardiovascular mortality in the capital, which, in turn, would have been reflected in hospitalization of patients with this severe pathology. However, according to the Moscow RFU-EMIAS database, hospitalization within a month prior to death accounted for only 1.3% of cases of cardiomyopathy unspecified in working ages among males and 1.1% of cases in females (versus 66.7% of case of dilated cardiomyopathy (I42.0), 62.5% of case of obstructive hypertrophic cardiomyopathy (I42.1), 25% of case of other hypertrophic cardiomyopathy (I42.2), all cases of other restrictive cardiomyopathy (i42.5), i.e. those forms of cardiomyopathy that suggest a natural formation of the pathology).

306

G. Semyonova Victoria et al.

Please, note that according to the RFS-EMIAS database, the share of hospitalizations due to alcoholic cardiomyopathy within a month prior to death added up to 80.9% in males and 75% in females. A special attention should be paid to the fact that alcohol consumption, unlike using drugs, is not criminal in nature. Therefore, as to cardiomyopathy unspecified, it is about a sudden death of people of working age, including young people. The lack of treatment can suggest, first of all, a rather isolated group of the population on the one hand, and a cause that is not approved by society, moreover, a cause of a criminal nature, on the other hand. It seems that it is narcotics that can now cause this pathology (Semyonova et al., 2019). The third indirect evidence of the drug-related etiology of cardiomyopathy unspecified is a similar social profile of those Muscovites of working age (15– 59 years) who died due to the causes of either confirmed or suspected drug-related etiology (poisoning by narcotics and cardiomyopathy unspecified, respectively). Thus, both male and female mortality from these causes was mainly accounted for by the following three groups by educational level – individuals with general complete secondary education, with secondary vocational and higher education (Table 19.2). People with general secondary education (11 grades) ranked first with both male and female contribution of this educational group being very similar to those who died from cardiomyopathy unspecified and poisoning by narcotics, adding up to 38.3% and 38.8% in males and 34.4% and 35.9% in females, which seems highly revealing. The group of secondary vocational education ranked second in terms of their contribution to the male mortality from both cardiomyopathy unspecified and poisoning by narcotics, however, the share of this group among those who died from cardiomyopathy was a bit lower than among those who died from poisoning by narcotics (26.1% versus 31%). The female indicators were quite similar (23.3% and 24.7%). The contribution of people with higher education is extremely interesting in the social context: first, in both males and females, this share is higher among those who died from cardiomyopathy rather than from poisoning by narcotics (22.7% versus 14.2% and 30.1% versus 22.2%, respectively); second, unlike males, the female contribution of this group ranks second among those who died from cardiomyopathy, remaining third among those who died from poisoning by narcotics. In our opinion, this only reconfirms the above-mentioned taboo diagnosis with reference to drug-related etiology, especially among fully socially adapted women. By and large, in the context of this study, we can verify a similar social profile of those who died from the confirmed poisoning by narcotics and cardiomyopathy unspecified. Finally, the fourth fact in favor of the drug-related etiology of cardiomyopathy unspecified in Moscow is a similar age profile of those who died from this cause and from poisoning by narcotics, that developed after 2015 (Figs. 19.9 and 19.10).

Level of education No primary education General primary education General basic education General secondary education, complete Vocational primary education Vocational secondary education Higher education, incomplete Higher education n/a Total

Cardiomyopathy, unspecified (I42.9) Males Females Number of deaths % Number of deaths 9 0, 3 6 5 0.2 2 176 5.6 61 1212 38.3 381 3 0.1 0 824 26.1 258 84 2.7 27 718 22.7 333 131 4.1 39 3162 100.0 1107 % 0, 5 0.2 5.5 34.4 0.0 23.3 2.4 30.1 3.5 100.0

Poisoning by narcotics (¸42, Y12) Males Females Number of deaths % Number of deaths 1 0.1 0 10 0.6 6 135 7.7 14 683 38.8 71 24 1.4 0 545 31.0 49 38 2.2 8 250 14.2 44 74 4.2 6 1760 100.0 198

0.0 3.0 7.1 35.9 0.0 24.7 4.0 22.2 3.0 100.0

%

Table 19.2 Social profile of those died from poisoning by narcotics (X42, Y12) and cardiomyopathy, unspecified (I42.9) with a breakdown on educational level (Moscow, July 2018 – September 2020)

19 Drug Addiction Mortality Among Young Muscovites: Official Rates. . . 307

308

G. Semyonova Victoria et al. males

females

25

3,5 3

20

2,5

15

2 1,5

10

1

5

0,5

0

2015

2017

2018

2019

2015

2017

2018

85 +

75 -7 9

65 -6 9

55 -5 9

45 -4 9

35 -3 9

25 -2 9

15 -1 9

85 +

75 -7 9

65 -6 9

55 -5 9

45 -4 9

35 -3 9

-5

25 -2 9

15 -1 9

0

2019

Fig. 19.9 Age profile of total deaths from poisoning by narcotics (¸42, Y12) among adult Muscovites, 2015–2019 (per 100,000) males

females 30

90 80

25

70 20

60 50

15

40 10

30 20

5

10 0

0 1519

2529 2015

3539

4549

2017

5559

6569

2018

7579

85+

2019

1519

2529 2015

3539

4549

2017

5559

6569

2018

7579

85+

2019

Fig. 19.10 Age profile of total deaths from cardiomyopathy, unspecified (I42.9) among adult Muscovites, 2015–2019 (per 100,000)

Thus, in 2015, the age profile of adult Muscovites died from poisoning by narcotics was characterized by a gradual increase in age-related indicators to a maximum in 30–34 years with a further decrease, whoever, age-related indicators in population over 45 equaled to zero (according to official data, there were no deaths from poisoning by narcotics among Muscovites over 45 in 2015). In females, the 2015 age profile of those died from poisoning by narcotics looks completely unformed, which is not surprising; according to the Rosstat data only five female

19 Drug Addiction Mortality Among Young Muscovites: Official Rates. . .

309

Muscovites aged over 15 died from poisoning by narcotics in 2015. However, after 2017, the age profile changes dramatically: a distinctive peak is being clearly developed with the maximum in people aged 35–39 in both males and females. There is a gradual decrease with age in males and their age-related indicators after the age of 65 years mainly equal to zero, while the age-related indicators in females over 50 are fluctuating (Fig. 19.9). The age profile of Muscovites died from cardiomyopathy unspecified in 2015 was characterized by a smooth increase in age-related mortality rates with the maximum in people aged 50–54 and an equally smooth drop to the minimum in the oldest. In Muscovites, the age curve reaches the plateau in people over 40 and gradually decreases with age with a certain increase in people aged 80–84. However, after 2017, the age profile becomes different: as in case with poisoning by narcotics, a clear peak is being developed in both males and females, however, its maximum is registered in males aged 40–44, i.e. 5 year “older” men. In females, the maximum is registered in women aged 45–59 with its shape much resembling the age curve in Muscovites died from poisoning by narcotics, up to a certain increase in 2018, clearly expressed in older female Muscovites (Fig. 19.10). The data obtained make it possible to advance a very cautious, yet to be confirmed hypothesis that “cardiomyopathy, unspecified” diagnosis in older ages can disguise suicides or accidents due to misuse of narcotic medications. It is rather illustrative that the shape of the 2015 and 2017–2019 age curves in Russia is fundamentally similar for poisoning by narcotics on the one hand, and cardiomyopathy unspecified on the other, but differ significantly: in case of poisoning by narcotics it is a peak with the maximum in either 30–34 or 35–59 years with a further reduction in age-related indicators, while in case of cardiomyopathy unspecified it is a smooth rise up to the maximum in 45–49 years, followed by a decline with age in males and increased age-specific mortality in females over 70. The 2015 age profile is characterized by the same patterns, differing only in the maximum in older ages (60–64 years) (Figs. 19.11 and 19.12). Thus, the age profile of Muscovites died from poisoning by narcotics and cardiomyopathy unspecified, first, changed dramatically in the second half of the 2010s, second, the age curve of mortality from these causes after 2015 becomes unmistakably similar, which was not observed in Russia. Therefore, referring to the question about etiology of unspecified cardiomyopathy in Muscovites of working age and a suddenly increased mortality from this cause, drug-related etiology is confirmed, first, by similarity in mortality dynamics, second, by hospitalization rate within a month prior to death, third, by similarity in social profile, and fourth, by similarity in age profile developed after 2015. Another important issue is death registration among working ages, including young people in Moscow, that cannot but follow from the analysis of mortality from poisoning by narcotics and cardiomyopathies unspecified, as well as a clear mismatch between the Rosstat data and the Moscow Bureau of forensic medical examination regarding the number of overdose deaths mentioned in the beginning of the study.

310

G. Semyonova Victoria et al. males

females 3

25

2,5

20

2 15 1,5 10 1 5

0,5 0

0 1519

2529

3539

2015

4549

5559

2017

6569

7579

2018

85+

1519

2019

2529 2015

3539

4549

2017

5559

6569

2018

7579

8599

2019

Fig. 19.11 Age profile of total deaths from poisoning by narcotics (¸42, Y12) among the Russian adult population, 2015–2019 (per 100,000) males

females

30

12

25

10

20

8

15

6

10

4

5

2 0

0 1519

2529 2015

3539

4549

2017

5559

6569

2018

7579

85+

2019

1519

2529 2015

3539

4549

2017

5559

6569

2018

7579

85+

2019

Fig. 19.12 Age profile of total deaths from cardiomyopathy, unspecified (I42.9) among the Russian adult population, 2015–2019 (per 100,000)

The authors have already dwelled upon the timing of these changes and mortality from symptoms, signs, and ill-defined conditions – the vaguest class of death causes. In terms of mortality from these causes among the working-age population, especially young people, Moscow used to be a permanent outsider among the Russian regions in 2000–2015. As already mentioned, mortality from ill-defined conditions in both Moscow and Russia is currently determined by the diagnosis “Ill-defined and unknown cause

19 Drug Addiction Mortality Among Young Muscovites: Official Rates. . .

311

of mortality” (R99). Such diagnostics is undoubtedly opportunistic in nature, as suggested by the very possibility of sharp changes in mortality trends from these causes, registered both in 2015–2016, and 1999–2000, when indicators decreased multiple times during the year (Semyonova & Evdokushkina, 2003; Semyonova et al., 2004; Gavrilova et al., 2008). According to ICD-10, external causes should not be included in this class, however a detailed analysis showed that, along with the collapse in mortality from ill-defined conditions, Moscow reported increased mortality from both poisoning by narcotics and cardiomyopathy unspecified, as well as accidental alcohol poisoning (X45), and event of undetermined intent (Y10-Y34), which includes both latent suicides and latent homicides by condition (Semyonova et al., 2019). This episode is not a unique one. Change in mortality trends from ill-defined conditions registered in Moscow in 1999–2000 mirrored itself in 2015–2016: this period is marked by a dramatic increase in mortality from ill-defined conditions among working ages in Moscow against the background of a sharp decline in mortality (Gavrilova et al., 2008; Semyonova et al., 2004, 2019; Semyonova & Evdokushkina, 2003). This timing of changes in mortality among working-ages in Moscow, including young Muscovites, makes it possible to conclude that the Moscow mortality from ill-defined conditions in the 2000s were determined by both external causes (alcohol and drug poisoning, event of undetermined intent) and causes due to external factors (cardiomyopathy of drug-related etiology). We’d like to emphasize it here that this diagnosis can be specified if all filling rules for death certificate adopted in Russia are respected: if the death cause is cardiomyopathy, it is suffice to note the drug status of the deceased in item 19II (other significant conditions contributing to death but not related to the disease or pathological condition leading to it) making at least indirect references to the underlying cause of death. Please, keep in mind, that ICD-10 allocates Cardiomyopathy due to drug and external agent (I42.7) as a separate nosological unit with the note: “use an additional code of external causes (class XX) if identification of the cause is needed”. By definition, drugs are among “other external factors”. However, according to the database of medical death certificates in Moscow, this diagnosis is a rare exception and cannot account for the explosive increase in mortality from cardiomyopathy unspecified in the working-age population (especially among young people). However, against the background of the decreased mortality from ill-defined conditions, increased mortality from such causes as poisoning by narcotics and cardiomyopathy unspecified among young Muscovites as well as working ages should be considered as a statistical artifact suggesting some improvement in diagnosis in medical terms and quality of data – in statistical terms rather than a sign of health deterioration. In this regard, the developed within last 2 years reduction in mortality from cardiomyopathy unspecified among Muscovites in working ages, including young people, against the background of the overlapping rise in mortality from ill-defined conditions cannot be regarded as a positive thing: it is a common way of transferring

312

G. Semyonova Victoria et al.

mortality due to drug addiction – a socially-determined and socially-sound cause, in the most latent form, kind of a “black box”, “Ill-defined and unknown cause of mortality” (R99). Is not it ironic that the answer to the second question about mismatch in data on deaths from poisoning by narcotics between the Moscow Bureau of forensic medical examination, on the one hand, and Rosstat, on the other, is the simplest one: the analysis of biological material of people allegedly died from drugs and, consequently, making the final diagnosis often requires extensive research, resulting in issuing a preliminary death certificate to the deceased’ relatives and the diagnosis listed in the certificate (in Moscow it is often “Ill-defined and unknown cause of mortality”) is registered at the Registry office, that in its tern submits data on to statistical agencies that use such data to develop official reports on population mortality. A further research by forensic officers can identify narcotics as the underlying cause of death, but the Russian legal framework does not mandate changes in the cause of death in the statistical authorities. A similar mismatch in data between the Burau of forensic medical examination and official statistics have been earlier identified by domestic researchers with regard to mortality from alcohol poisoning (Vaysman et al., 2006). In the context of this study we’d like to stress it out that the decreased mortality from ill-defined conditions in working ages, including young people was accompanied by the increased mortality from both poisoning by narcotics and cardiomyopathy unspecified, as well as alcohol poisoning (Semyonova et al., 2019). However, a mere possibility of such statistical manipulations suggests a much more fundamental problem – the problem of diagnosing drug addiction and its consequences within the framework of ICD-10. Paradoxically, ICD-10 provides for more than a dozen pathologies of alcoholic etiology, including a rather funny diagnosis as “Poisoning by and exposure to alcohol, undetermined intent” (Y15): according to the ICD-10 logics, the expert fails to figure out whether poisoning by alcohol is an accident or an exotic suicide. As to drug addiction and its consequences, as mentioned above, ICD-10 provides for only two causes of death of drug-related etiology – poisoning by narcotics (X42 and Y12) and mental and behavioral disorders due to psychoactive substance use (F11, F12, F14, F16). Since mental disorders cannot be the underlying cause of death, drug users, according to ICD-10, can die only from overdose; ICD-10 does not provide for somatic pathologies caused by narcotics (the only exception is the above-mentioned “Cardiomyopathy due to drug and external agent” (I42.7)). Unfortunately, this problem is not restricted to Moscow or even Russia alone: despite the fact that drug addiction is a global challenge, the European database on mortality does not contain either indicators on mortality from “cardiomyopathy due to drug and external agents”, or poisoning by narcotics. In fact, this WHO database3 lists only “drug dependence and toxicomania” and “psychoactive substance use” related to mental disorders as causes of death of drug-related etiology. This approach

3

https://gateway.euro.who.int/ru/datasets/european-mortality-database/

19 Drug Addiction Mortality Among Young Muscovites: Official Rates. . .

313

to the WHO mortality database makes it impossible to conduct comparative crosscountry analysis of deaths caused by narcotics. Thus, the global approaches to registering drug-related deaths implemented by WHO in ICD-10 as the basic document for coding causes of diseases and deaths, as well as Eurostat as a tool for comparative analysis and assessment of mortality, question the level of drug-related mortality either in Moscow and Russia or in other countries as well.

19.6 Conclusions To finalize the analysis of drug addiction mortality in Moscow, we’d like to focus on the following circumstances. The 2011–2019 results turned out to extremely negative for young Muscovites: while the Russian total mortality from poisoning by narcotics decreased by almost a quarter (23.1%) in males and over two-fold (2.2 times) in females, the Moscow one increased 8.3 and 15-fold, respectively, and the negative changes took place within the last 5 years. Despite difference in trends, there is a common pattern shared by both the capital and Russia as a whole: not necessarily consistent, but rather sustainable increase in the contribution of poisoning with undetermined intent to the total mortality from poisoning by narcotics: it currently adds up to 32.5% in Muscovite males and 46.7% in Muscovite females, compared to 20% and zero in 2011, respectively. (in Russia, this indicator during this period, increased from 21.5% and 30.8% to 40% and 50%, respectively). Poisoning by narcotics is far from being the complete loss due to causes associated with drug use. It is proved that cardiomyopathy unspecified is also considered a death cause of drug-related etiology. In addition to the available clinical data it is confirmed by similarity in mortality dynamics, hospitalization rates within a month prior to death, similarity in social profile, and similarity in age profile developed after 2015. Some deaths due to drug-related causes of death are disguised as ill-defined conditions, as suggested by simultaneous mirror changes in mortality from poisoning by narcotics, mental disorders due to psychoactive substance use, cardiomyopathy unspecified, and ill-defined conditions. Changes in mortality among young Muscovites are not specific: these patterns are registered throughout the entire working ages in the capital.

314

G. Semyonova Victoria et al.

References Blagova, O. V., & Nedostup, A. V. (2017). Classification of non-coronary heart diseases: Our views on the issue. Russian Journal of Cardiology, 2(142), 7–21. https://doi.org/10.15829/1560-40712017-2-7-21. (in Russian). Brigden, W. (1957). Uncommon myocardial diseases; The non-coronary cardiomyopathies. Lancet, 273(7007), 1179–1184. Crump, C., Ioannidis, J. P., Sundquist, K., Winkleby, M. A., & Sundquist, J. (2013). Mortality in persons with mental disorders is substantially overestimated using inpatient psychiatric diagnoses. Journal of Psychiatric Research, 47(10), 1298–1303. https://doi.org/10.1016/ j.jpsychires.2013.05.034. Epub 2013 Jun 25. Drapkina, O. M., Samorodskaya, I. V., Starinskaya, M. A., Masyakin, A. V., Kazakovtsev, B. A., & Stupakov, I. N. (2019). ´omparison of mortality from diseases of the circulatory system and nervous and mental disorders in Russia in 2013 and 2017. Preventive Medicine, 22(4), 7–13. Frustaci, A., Russo, M., Morgante, E., et al. (2015). Oxidative myocardial damage in human cocainerelated cardiomyopathy. European Journal of Heart Failure, 17, 283–290. Gavrilova, N. S., Semyonova, V. G., Dubrovina, E. V., Evdokushkona, G. N., Ivanova, A. E., & Gavrilov, L. A. (2008). Russian mortality crisis and the quality of vital statistics. Population Research and Policy Review, 27, 551. Gorgaslidze, A. G., Sayfullaeva, M. A., & Kuzmina, M. M. (1993). Heart rhythm and heart muscle contractility disturbances in opium and ephedrine addiction. Cardiology, 33(1), 14–16. (in Russian). Hällgren, J., Ösby, U., Westman, J., & Gissler, M. (2019). Mortality trends in external causes of death in people with mental health disorders in Sweden, 1987-2010. Scandinavian Journal of Public Health, 47(2), 121–126. https://doi.org/10.1177/1403494818758912. Epub 2018 Mar 1. Kakorina, E. P., & Kazakovtsev, B. A. (2013). Peculiar features of mental mortality indicators in the Russian Federation. Mental Health, 11(83), 3–6. (in Russian). Kakorina, E. P., Kazakovtsev, B. A., & Golland, V. B. (2015). Mortality from mental disorders and mortality in case of mental disorders. Mental Health, 13(106), 22–27. (in Russian). Kaktursky, L. V. (2000). Sudden cardiac death (Clinical morphology). Medicine Publ; 127 p. (in Russian). Lumme, S., Pirkola, S., Manderbacka, K., & Keskimäki, I. (2016). Excess mortality in patients with severe mental disorders in 1996-2010 in Finland. PLoS One, 11(3), e0152223. https:// doi.org/10.1371/journal.pone.0152223 Mackenbach, J. P., Karanikolos, M., & Looman, C. W. (2014). The rise of mortality from mental and neurological diseases in Europe, 1979-2009: Observational study. BMC Public Health, 14, 840. https://doi.org/10.1186/1471-2458-14-840 Moiseev, V. S., & Kiyakbaev, G. K. (2009). The problem of classification of cardiomyopathies. Cardiology, 49(1), 65–70. (in Russian). Oh, T. K., Song, I.-A., Lee, J. H., Lim, C., Jeon, Y.-T., Bae, H.-J., et al. (2019). Preadmission chronic opioid usage and its association with 90-day mortality in critically ill patients: A retrospective cohort study. British Journal of Anaesthesia. https://doi.org/10.1016/j.bja.2019.03.032 Radunski, U. K., Fuger, U., Bohnen, S., Lund, G. K., Stehning, C., Zeller, T., et al. (2017). Asymptomatic cocaine abuse – Myocardial tissue characterization using cardiac biomarkers and cardiovascular magnetic resonance imaging. Circulation Journal, 81(5), 701–708. https:// doi.org/10.1253/circj.cj-16-0941 Report on Drug Situation in the Russian Federation, State Anti-drug Committee. (2019). https:// ÔÒÃÓÕÃ74.Ó×/upload/iblock/0b5/Doklad_GAK_2019_KMM.pdf. (in Russian). Report on drug situation monitoring in Moscow. (2019). https://123ru.net/moscow/239457689/ (in Russian). Sabgaida, T. P., Semyonova, V. G., Evdokushkina, G. N., Sekrieru, E. M., & Nikitina, S. Y. (2014). Death cause modification in death registration. Social Aspects of Public Health, 3(37), 2. (in Russian).

19 Drug Addiction Mortality Among Young Muscovites: Official Rates. . .

315

Saxena, S. (2018). Excess mortality among people with mental disorders: A public health priority. The Lancet Public Health, 3(6), e264–e265. https://doi.org/10.1016/S2468-2667(18)30099-9 Semyonova, V. G., & Evdokushkina, G. N. (2003). “Ill-defined” epidemiological crisis. In V. I. Starodubov, V. Mikhailova Yu, & A. E. Ivanova (Eds.), Health of the Russian population in the social context of the 90’s: Problems and prospects (pp. 85–94). Medicine Publ. (in Russian). Semyonova, V. G., Gavrilova, N. S., Evdokushkina, G. N., & Gavrilov, L. A. (2004). Quality of medical and statistical data as a mirror of the current crisis in the Russian healthcare. Public Health and Disease Prevention, 3, 11–18. (in Russian). Semyonova, V. G., Ivanova, A. E., Zubko, A. V., Sabgayda, T. P., Zaporozhchenko, V. G., Evdokushkina, G. N., & Gavrilova, N. S. (2019). Risk factors for increased mortality and features of their registration in Moscow. Healthcare of the Russian Federation, 63(6), 322– 330. (in Russian). Solovyov, A. G., Vyazmin, A. M., Mordovsky, E. A., & Nikulichev, V. V. (2015). Structure of mortality among people under medical follow-up for mental and behavioral disorders caused by alcohol use. Korsakov Journal of Neurology and Psychiatry, 115(11), 97–102. (in Russian). Sorokina, V. V. (2018). Dynamics in the heart morphological changes in chronic drug addiction depending on the length of drug use. Siberian Medical Journal, 1, 11–15. (in Russian). Sorrentino, A., Trotta, S., Colucci, A. P., Aventaggiato, L., Marzullo, A., & Solarino, B. (2018). Lethal endomyocarditis caused by chronic “Krokodil” intoxication. Forensic Science, Medicine and Pathology, 14(2), 229–235. https://doi.org/10.1007/s12024-018-9967-9 Stankowski, R. V., Kloner, R. A., & Rezkalla, S. H. (2015). Cardiovascular consequences of cocaine use. Trends in Cardiovascular Medicine, 25, 517–526. Starace, F., Mungai, F., Baccari, F., & Galeazzi, G. M. (2018). Excess mortality in people with mental illness: Findings from a Northern Italy psychiatric case register. Social Psychiatry and Psychiatric Epidemiology, 53(3), 249–257. https://doi.org/10.1007/s00127-017-1468-8 Vaysman, D. S., Dubrovina, E. V., & Redko, A. N. (2006). Information support for research on mortality in Russia. Public Health and Disease Prevention, 6, 31–38. (in Russian). Walker, E. R., McGee, R. E., & Druss, B. G. (2015). Mortality in mental disorders and global disease burden implications: A systematic review and meta-analysis. JAMA Psychiatry, 72(4), 334–341. https://doi.org/10.1001/jamapsychiatry.2014.2502 World Drug Report. UNODC. (2020). https://www.unodc.org/unodc/press/releases/2020/June/ media-advisory-global-launch-of-the-2020-world-drug-report.htm

Chapter 20

Factors Reducing Child Mortality from Congenital Heart Defects in Russia A. V. Zubko, T. P. Sabgayda, and V. G. Semyonova

Despite progress in technologies for care delivery to low birth weight preterm babies and prenatal care, congenital anomalies (malformations) (CA) remain an urgent public health issue. Worldwide, about 10% of newborn deaths are associated with congenital anomalies (Rothman et al., 2008; WHO/CDC/ICBDMS, 2014), in Europe they are responsible for 25% of newborn deaths, including 2.5% of deaths that occurred during the first week of life (Dolk et al., 2010). Of the 5.1 million births in the European Union (EU) each year, approximately 104,000 (2.5%) will be born with congenital anomalies (Kinsner-Ovaskainen et al., 2020). In Russia, congenital anomalies rank second among causes of death in the first year of life following conditions originating in the perinatal period, and rank third among causes of child disability (Postoyev et al., 2017). Complications caused by these anomalies can affect health throughout the life, having a significant impact on life expectancy and life quality (Glinianaia et al., 2012). Some congenital anomalies can be corrected either by therapy or surgery, while others are untreatable and concise life expectancy to 10 years. In countries with limited resources, the share of children with congenital anomalies is much higher compared to the rich regions of the world. The original version of the chapter has been revised. A correction to this chapter can be found at https://doi.org/10.1007/978-3-030-93005-9_35 A. V. Zubko ()· T. P. Sabgayda · V. G. Semyonova Institute for Demographic Research Branch of the Federal Center of Theoretical and Applied Sociology of the Russian Academy of Sciences (IDR FCTAS RAS), Moscow, Russia Federal Research Institute for Health Organization and Informatics of Ministry of Public Health of Russian Federation, Moscow, Russia A.N. Bakulev National Medical Research Center of Cardiovascular Surgery, Moscow, Russia © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022, corrected publication 2022 C. H. Skiadas, C. Skiadas (eds.), Quantitative Methods in Demography, The Springer Series on Demographic Methods and Population Analysis 52, https://doi.org/10.1007/978-3-030-93005-9_20

317

318

A. V. Zubko et al.

Factors increasing the risk of birth defects include socio-economic conditions (poor sanitary conditions, maternal infections, maternal exposure to certain medications, malnutrition, lack of essential vitamins and minerals in the diet), sociodemographic (parental age (Berg et al., 2015; Garne & de Walle, 2009)), parental lifestyle (smoking (Eiriksdottir et al., 2015; Jawad et al., 2009), heavy drinking and using drugs (O’Leary et al., 2010), insufficient exposure to fresh air, inadequate rest), maternal health (diabetes mellitus (Dart et al., 2015; Gabbay-Benziv et al., 2015), kidney and urinary system infections (Zhu et al., 2016), viral infections (Gilboa et al., 2017; Hall et al., 2017; Potera, 2018), obesity or overweight (Correa & Marcinkevage, 2013)), environment (exposure to pesticides and other chemicals (Selyutina et al., 2014), exposure to high radiation, living near a large metallurgical production, waste sites or mines (Tverskaya & Verzilina, 2018)). However, many causes of congenital anomalies are yet to be identified. Researchers increasingly agree that the development of congenital anomalies is multifactorial in nature, therefore, along with genetic predisposition it is necessary to consider all aspects of the parental life, including influence of environmental factors to explore risk factors for congenital anomalies. World Health Organization classifies congenital anomalies as a group of environmental diseases that serve as markers of environmental problems (WHO, 2014). According to EUROCAT (European network of population-based registries for the epidemiological surveillance of congenital anomalies), congenital heart defects (CHD) were the most common non-chromosomal subgroup of all anomalies: 79.76 per 10,000 births among 255.28 anomalies in 2011 to 2018 in Europe (EUROCAT, 2020). In the United States, Japan, Sweden, Finland, Canada, and Russia, an average of 0.7% of children with CHD are born annually. In North America, congenital heart defects account for 37% of infant mortality, and 45% – in Western Europe (Rosano et al., 2020). In Russia, CHD in comparison with other congenital anomalies also leads in prevalence (Baibarina et al., 2011). A meta-analysis of the prevalence of congenital heart defects in 1930–2009 showed a tendency towards increasing prevalence (Van der Linde et al., 2011). Increased CHD prevalence, including severe forms, is likely to be associated with improved methods of diagnosis and prevention of antenatal and infant mortality (Saperova & Vakhlova, 2017). According to randomized studies conducted in the United States and Great Britain, over 70% of babies with untreated CHD die in their first year of life (Kim et al., 2003). According to the Ukrainian authors, without surgery, 90% of babies with CHD die in their first year of life (Knyshov, 2003; Beshlyaga & Lazorishinets, 2005). However, 20% of babies with CHD without surgical correction become hardly operable or absolutely beyond surgery by the end of the first year of life due to irreversible changes in organs and systems (Zinkovsky et al., 2003). With a highly developed system of heart surgery and its timely provision, the mortality rate in patients equals to 7–10% by the end of the first year of life, and under 14% by the 16-th year of life (Kim et al., 2003). This indicates the importance of diagnosing CHD in children as early as the first year of life to ensure timely heart surgery and reduce mortality rates.

20 Factors Reducing Child Mortality from Congenital Heart Defects in Russia

319

40 35

SDR, per 100000

30 25 20 15 10 5

19

89 19 91 19 93 19 95 19 97 19 99 20 01 20 03 20 05 20 07 20 09 20 11 20 13 20 15 20 17 20 19

0

Russia

Kazakhstan

Belarus

Bulgaria

Czech Republic

Estonia

France

Italy

Japan

Fig. 20.1 Standardized deaths from Ôongenital malformations, deformations and abnormalities, both sexes, 0–14 years old (per 100,000 population)

Since the beginning of the century, most European countries have seen a decline in the mortality rate from congenital anomalies (WHO Mortality Database, 2020; WHO. Congenital anomalies. Information Bulletin, 2020). The relevant Russian mortality rate exceeds the European ones and rates of the most Post-soviet countries (Fig. 20.1). It should be noted here that, unlike other Post-soviet countries, the Russian child mortality increased as a result of socio-political transformations and the country collapse in 1991. In the current century, only Kazakhstan reports the increased child mortality from congenital anomalies while other countries register its reduction. From the beginning of the century and until 2015, Estonia was the only country to report a higher than Russian decline in mortality (reduction by 72.3% versus 60.5%). Against the decreased mortality from congenital anomalies, the incidence of congenital anomalies increased by 9.3% in the Russian children under 15 from 2009 to 2019 (Fig. 20.2). The growth rate of the incidence of congenital anomalies of the circulatory system in children is most notable (by 48.1% over the same period), accounting for 37.2% of all anomalies in 2009 and adding up to 50.4% in 2019. Dynamics in the incidence of congenital anomalies of the circulatory system shows increased growth rates after 2012, when the Russian Federation initiated active implementation of the modern technologies for active management for preterm babies born from 22 weeks’ gestation and/or with low and extremely low weight at birth.

320

A. V. Zubko et al. 4000 3500 3000 2500 2000 1500 1000 500 0 2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

2019

Congenital malformations, deformations and abnormalities Congenital malformations of the circulatory system Fig. 20.2 Overall incidence of Ôongenital malformations, deformations and abnormalities and congenital malformations of the circulatory system, both sexes, 0–14 years old (per 100,000 population), Russian Federation

Over a decade, the share of congenital anomalies in the structure of child morbidity increased from 1.4% to 1.7%, while in the structure of causes of death they decreased from 19.6% to 16.3%. However, over a 30-year period, the contribution of congenital anomalies to the mortality in children under 15 has hardly changed (16.7% in 1989 and 16.3% in 2019). Unlike morbidity, the contribution of congenital heart defects in the structure of causes of death from congenital anomalies decreased from 39.8% in 1989 and 37.6% in 2009 to 31.1% in 2019 (Fig. 20.3). It does raise the question of what factors have the greatest impact on reducing child mortality in the current century. Socio-economic conditions and the environment have approximately the same impact on the residents of a Russian region. Their indicator can be environmental indexes measured four times a year by the public organization “Green Patrol” [Environmental Rating of Regions]. To determine factors for reducing mortality rate, it is advisable to analyze relationship between both child morbidity and mortality from congenital anomalies of the circulatory system and corresponding environmental indexes. The level of mortality from congenital anomalies is affected by the health system performance. Success in treating CHD is directly dependable upon quality of health care rather than timely detection alone (Cao et al., 2017). Owing to improvements of diagnostic and operational methods, in recent years we have witnessed reduction in the mortality rates in this group of patients and increased survival rate in patients operated for CHD. Therefore, it is wise to consider indicators of the health system performance as factors for reducing mortality.

20 Factors Reducing Child Mortality from Congenital Heart Defects in Russia

321

40.0

SDR, per 100000

35.0 30.0 25.0 20.0 15.0 10.0 5.0

09 20 11 20 13 20 15 20 17 20 19

07

20

03

05

20

20

01

20

99

20

97

19

95

19

91

93

19

19

19

19

89

0.0

Congenital malformations, deformations and abnormalities Congenital malformations of the circulatory system Fig. 20.3 Standardized deaths from congenital malformations, deformations and abnormalities and congenital malformations of the circulatory system, both sexes, 0–14 years old (per 100,000 population), Russian Federation

20.1 Material and Methods Mortality from congenital cardiovascular anomalies in children under 15 was analyzed on the basis of the official statistics of the Russian Federal State Statistic Service (Rosstat) for 1989–2019. Standardized deaths from congenital heart defects (the European Standard Population) were calculated for the following age groups for both sexes: 0–14 years, 0–1 year, 1–4 years and 5–14 years. The age structure of the deceased was calculated. To determine impact of the health system on mortality under study, the authors have analyzed the decline rates of mortality from congenital heart defects in children over the periods characterized by different levels of resources of the Russian health system. First, in 1998, a new building of the V. I. Burakovsky Research Institute of Cardiac Surgery (hereinafter referred to as the Burakovsky Center) was commissioned. It is one of the three institutes of the A.N. Bakulev National Medical Research Center for Cardiovascular Surgery performing about 85% of all heart surgeries in infants. Second, since 2012, the Russian Federation has initiated implementation of the modern technologies for active management for preterm babies born from 22 weeks’ gestation and/or with low and extremely low weight at birth. Therefore, the authors compared rates of mortality reduction in children across different ages during 1989–1998, 1998–2012 and 2012–2019. To compare mortality rates in individual regions of the Russian Federation, the authors have calculated average mortality rates 3 years prior to the commission of

322

A. V. Zubko et al.

the Burakovsky Center (1996–1998), 3 years prior to the introduction of the practice of active management for low birth preterm babies (2009–2011) and in 2017–2019. Surgical activity of heath care facilities was evaluated according to the Statistical collections of the A.N. Bakulev National Medical Research Center for Cardiovascular Surgery from 1995 until 2019, developed on the basis of statistical records obtained from health care facilities that provide specialized cardiac surgery (Bokeria & Gudkova, 2006–2019). To determine impact of the environmental and socio-economic factors on mortality from congenital heart defects, the authors have compared dynamics in mortality from congenital heart defects in 85 regions of Russia and different levels of socio-economic development and environmental pollution. In this section the authors have also analyzed disease incidence in children under 15. Data from Collections of the Ministry of Health of the Russian Federation for 2009–2019 were used. The correlation analysis of mortality and morbidity indicators has been undertaking with the corresponding factors expressed by corresponding indexes. Since 2009, the public organization “Green Patrol” has been annually publishing Environmental rating of the Russian regions for the purpose of their comparative assessment from the standpoint of environmental safety and environmental protection. The rating model is developed on the basis of the conceptual frames for the noosphere origin, proposed by V.I. Vernadsky. Each region is evaluated according to the following three criteria: ecosphere (environmental index), technosphere (industrial environmental index), and society (socio-ecological index). Each index has seven indicators; depending on the nature of the event, a certain indicator or several indicators are assigned numerical values of +1/−1, where +1 is a positive rating, while −1 – a negative one. The most environmentally advantaged regions (oblast, territory) get the highest score, while the most polluted ones- the lowest score. The Environmental rating of regions of the Russian Federation is being developed as information materials come from various sources, including the media, government authorities, public organizations, expert organizations, economic entities and initiative groups of citizens. Information materials are current messages, publications or documents that describe the state of objects and processes, as well as situations, happenings and events in the field of ecology and environmental protection on-line. The significance of the event is being evaluated by a group of experts, whose recruitment and performance is guided by the principle of “jury trial”. Ratings are calculated using a computer program. The ratio between positive and negative ratings is automatically converted into a 100-score scale. The authors have used the published environmental, industrial environmental and socio-ecological indexes estimated for the spring period for 2009–2019. Following identification of the statistically significant correlations, the levels of and dynamics in the incidence of CHD in children under 15 and infant mortality from CHD were compared in groups of the regions with the permanent highest and lowest environmental and industrial environmental indexes.

20 Factors Reducing Child Mortality from Congenital Heart Defects in Russia

323

20.2 Results 20.2.1 Impact of the Health System In the last years of the Soviet period, the CHD mortality in children under 15 remained unchanged (Fig. 20.4a) reaching its 30-year maximum in 1997 (12.7 per 100,000 population), followed by a steady decline up to 2.6 in 2019. Dynamics in child mortality is mainly determined by infant mortality, which decreased from 155.0 in 1997 to 33.4 in 2019 (Fig. 20.4b). Dynamics in mortality in children aged 1–4 years hardly differs from dynamics in infant mortality, being a tenfold lower (Fig. 20.4c). During the period under study, changes in mortality in younger children are best approximated by a thirddegree polynomial, while mortality from CHD in older children was monotonically decreasing during the analyzed period with dynamics in their mortality being well approximated by a straight line. A notable decrease in child mortality from CHD began in 1999, i.e. after the new building of the Burakovsky Center was put into operation in 1998. Its commission resulted in a significant increase in the share of infants among CHD surgeries: 6.7% in 1995 versus 37.2% of all CHD corrective surgeries in 2019. Children in the first year of life account for the vast majority of all deaths from CHD in children under 15, and this share is growing (Table 20.1). However, 180.0

14.0

A

12.0

B

160.0 140.0

10.0

120.0

8.0

100.0

6.0

80.0 60.0

4.0

40.0

y = 0.0115x 3 - 0.6653x 2 + 6.2656x + 137.23

0.0

8.0

C

7.0 6.0

89 19 91 19 93 19 95 19 97 19 99 20 01 20 0 20 3 05 20 07 20 09 20 11 20 13 20 15 20 17 20 19

R 2 = 0.9912

19

95 19 97 19 99 20 01 20 03 20 05 20 07 20 09 20 11 20 13 20 15 20 17 20 19

19

19

19

19

93

0.0

91

20.0

89

2.0

1.4

D

1.2 1.0

5.0

0.8

4.0 0.6

3.0

0.4

2.0 3

2

0.2

y = -0.0364x + 1.3239 2

R = 0.9494

0.0

0.0

19 89 19 91 19 93 19 95 19 97 19 99 20 01 20 03 20 05 20 07 20 09 20 11 20 13 20 15 20 17 20 19

R = 0.9419

19 89 19 91 19 93 19 95 19 97 19 99 20 01 20 03 20 05 20 07 20 09 20 11 20 13 20 15 20 17 20 19

1.0

2

y = 0.0006x - 0.037x + 0.4316x + 4.5307

Fig. 20.4 Standardized deaths from congenital heart defects, 0–14 years old (a), 0 year (b), 1– 4 years (c) and 5–14 years (d), (the European Standard Population, per 100,000 corresponding population), 1989–2017, Russian Federation

324

A. V. Zubko et al.

Table 20.1 Number and share (%) of deaths from congenital health defects in different ages, 1989, 1997 and 2019, Russian Federation Age (years) 0 1 2 3 4 5–9 10–14 0–14

1989 Number 3224 244 114 75 68 183 81 3989

% 80.8 6.1 2.9 1.9 1.7 4.6 2.0 100.0

1997 Number 2002 191 68 51 44 129 100 2585

% 77.4 7.4 2.6 2.0 1.7 5.0 3.9 100.0

2019 Number 514 34 12 8 11 18 18 615

% 83.6 5.5 2.0 1.3 1.8 2.9 2.9 100.0

the share of children aged 5–9 in the structure of child mortality is significantly decreasing. In the period from 1998 to 2019, the number of health care facilities that perform heart surgery in children with cardiopulmonary bypass (CPB) increased by 43.8%, while the total number of surgeries increased 3.1-fold (including a 2.9-fold increase in surgery with CPB), and the number of CHD surgeries in infants increased 10.6 times, including a 13.9-fold increase in surgery with CPB. In 2019, 50.6% of surgeries for CHD were performed with cardiopulmonary bypass, adding up to 62.5% in infants. A strong negative correlation was found between the number of CHD and the number of deaths from CHD under 1 year of life (−0.980) and in children under 15 in general for the period from 1995 to 2017 (−0.982) and 1995–2019, respectively. The correlation coefficient of the number of CHD surgery in infants with CPB and the CHD mortality in children under 15 equals to −0.89. The same correlation coefficient between child mortality and the number of such surgeries in children aged 1–3 added up to −0.909. Until 1998, mortality in children aged 1–4 was growing more rapidly than infant mortality, while mortality in children aged 5 and over was decreasing (Table 20.2). In the second period the mortality rate in children aged 1–4 was decreasing most rapidly (by 7.0% per year on average), reflecting increased cure rate for CHD by

Table 20.2 Average annual rate of increase in mortality from CHD in different ages prior to and after commission of the Buracovsky Center and implementation of active management for low birth weight preterm babies (%) Periods 1989–1998 1998–2012 2012–2019 1989–2019

Age Under 1 0.9 −5.2 −9.8 −6.6

1–4 3.9 −7.0 −7.2 −7.1

5–14 −1.9 −5.3 −6.0 −5.5

0–14 1.0 −5.4 −9.8 −6.8

20 Factors Reducing Child Mortality from Congenital Heart Defects in Russia

325

surgery because of advances in medical technologies and improved quality of care delivery. The average annual rate of decline in mortality since 2012 has hardly changed in children over 1 year of life and significantly increased in infants. In other words, the cure rates for CHD by surgery in low weight preterm babies who previously failed to survive until 1 year of life and did not affect the survival rate in children aged 1–4, increased. Since child mortality from CHD is mainly determined by infant deaths, the authors have analyzed regional variations in child mortality based on the mortality in children in the first year of life averaged for 2017–2019. For 85 regions of the country, a direct correlation between infant mortality from CHD and availability of doctors in the regions during this period (0.22) was identified including obstetricians and gynecologists (0.31), but it is not relevant to cardiovascular surgeons, revealing the problem of insufficient quality of diagnosis of congenital malformations. As to morbidity, a direct correlation between the availability of doctors in the regions and the incidence of congenital anomalies in children under 15 (0.24) and the incidence of congenital heart defects (0.30) was also identified but it does not relate to the availability of obstetricians and gynecologists or cardiovascular surgeons. The authors have analyzed impact of the comission of the new Burakovsky Center and introduction of active management for low birth weight babies on the regional variations in child mortality from CHD based on infant mortality, averaged over the following three 3-year periods: 1996–1998, 2009–2011 and 2017–2019. Table 20.3 shows maximum and minimum in the current period and corresponding values in previous periods.

Table 20.3 Regions with maximum and minimum infant mortality from congenital heart defects and averaged deaths in 1996–1998, 2009–2011 Ë 2017–2019 (The European Standard Population, per 100,000 population) Regions of Russia Russian Federation Chukotka Autonomous Region Republic of Tyva Republic of Dagestan Kaliningrad Region Jewish Autonomous Region Moscow Stavropol Territory Khabarovsk Territory Republic of Karelia Chuvash Republic Republic of Kalmykia Sakhalin Region Kamchatka Region Belgorod Region Kurgan Region

1996–1998 151.5 314.6 296.1 151.8 121.5 66.1 142.0 79.6 68.6 100.2 173.4 102.8 103.9 162.9 144.9 143.1

2009–2011 73.9 94.9 115.6 174.9 47.5 138.2 97.3 98.6 52.4 22.0 34.1 129.8 72.9 94.8 36.2 74.1

2017–2019 36.4 153.3 90.9 88.6 76.2 72.4 58.6 56.2 52.7 15.4 13.3 11,4 10.7 9.3 9.0 7.4

326

A. V. Zubko et al.

Regional variations in infant mortality from CHD are significant: in the first analyzed period, infant mortality varied 32.1-fold (325.9 per 100,000 in the Republic of Khakassia versus 10.2 in the Republic of Ingushetia), in the second period −8.0-fold (174.9 per 100,000 in the Republic of Dagestan versus 22.0 in the Republic of Karelia), and in the period 2017–2019, the CHD infant mortality varied 20.7-fold. Over the entire period under study, infant mortality from CHD in Russia decreased 4.5-fold. The decline was registered in the majority of the Russian regions, with the exception of the Jewish autonomous region, where infant mortality increased.

20.2.2 Impact of Environmental Factors The matrices of correlation coefficients between the total morbidity with congenital anomalies of the circulatory system in children under 15 and environmental, industrial environmental and socio-ecological indexes for each year in the period 2009–2019 were calculated. Negative are only correlation coefficients with the environmental index. Statistically significant values of the coefficients are observed only for the relationship between morbidity in 2019 and environmental index in 2010 (r = −0.24) and 2011 (r = −0.21). I.e. morbidity in children under 15 is affected by the 10-year ago ecological situation. On the other hand, a direct correlation was found between this morbidity (as well as morbidity in 2018 and 2017) and the industrial environmental index in 2013 (r = −0.22), suggesting changes in detection of congenital anomalies of the circulatory system in 2013 mainly in the industrialized regions. No correlations between morbidity and socio-ecological index for any year were identified. The authors have defined the impact of environmental pollution and economic development on morbidity by comparing dynamics in morbidity in relatively successful and disadvantaged regions according by the corresponding indexes (Fig. 20.5). Out of the top 20 regions in 2009 according to the environmental index, only 11 have maintained their position in the top 20 by 2020. The authors have combined these regions into Group 1. Out of the 20 regions from the worst 20 in terms of environmental index, 12 regions have remained polluted by 2020. These regions have combined into Group 2. Out of the top 20 regions in terms of industrial environmental index in 2009, only 9 have kept their position in the top 20 by 2020 (Group 3), 11 regions from the worst 20 have remained in the rating by 2020 (Group 4). It turned out that the highest average incidence for the period 2009–2019 was observed in Group 3 (the most industrialized regions, 1981.7 per 100,000 population), while the lowest average incidence was registered in Group 4 (the least industrialized regions, 1178.4 per 100,000 population). The impact of environmental pollution is less pronounced: the average incidence in Group 1 is little higher compared to Group 2 (1674.0 versus 1553.3 per 100,000 population).

Morbidity, per 100000

20 Factors Reducing Child Mortality from Congenital Heart Defects in Russia

327

3000 2500 2000 1500 1000 500 0 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019

Group 1

Group 2

Group 3

Group 4

Fig. 20.5 Dynamics in morbidity with congenital anomalies of the circulatory system in children aged 0–14 (per 100,000 population) in groups with the maximum (Group 1) and minimum (Group 2) environmental index, maximum (Group 3) and minimum (Group 4) industrial environmental index, Russian Federation

The maximum increase in morbidity over the period 2009–2019 was registered in the contaminated areas of Group 2 (by 59.0%). Most industrialized regions of Group 3 rank second (a 41.8% increase). In regions with high environmental index (Group 1) and low industrial environmental index (Group 4), the incidence increased to a lesser extent (by 26.3% and 23.8%, respectively). A similar correlation analysis of the relationship between environmental factors and mortality from congenital heart defects has been undertaken for ages under 1 year of life, under 5 years of life, 5–9 and 10–14 years. A statistically significant correlation was found only between the environmental index in 2009 and mortality in children under 5 in 2014 (r = −0.35). No other cases of correlation between mortality in children of different age groups and environmental index have been detected suggesting a random nature of the identified relationship. In several cases, a correlation between infant mortality from congenital heart defects in 2013, 2015 and 2016 and industrial environmental index in the current (r = −0.28, r = −0.26, r = −0.25, respectively) and previous years (r = −0.28 for the 2012 index, r = −0.22 for the 2014 index, r = −0.26 for the 2015 index, respectively) has been identified. This finding is in line with the hypothesis that the better equipped perinatal centers in the economically developed regions the lower mortality in preterm babies Bokeria (2014), Bokeria & Gudkova (2013), Bokeria & Gudkova (2014), Bokeria & Gudkova (2015), Bokeria & Gudkova (2016). Correlations between infant mortality and socio-ecological index have been detected for the same years, however, the relationship is weaker suggesting its indirect nature. Figure 20.6 shows dynamics in infant mortality across groups formed by the ranking of environmental and industrial environmental indexes. The highest average mortality for the period 2009–2019 is registered in Group 1 (65.1 per 100,000 population), followed by industrially underdeveloped regions of Group 4 (63.5 per 100,000 population). Mortality in the polluted regions of Group 2 is lower compared

328

A. V. Zubko et al.

SDR, per 100000

120.0 100.0 80.0 60.0 40.0 20.0 0.0

2009

2010

Group 1

2011

2012

Group 2

2013

2014

2015

2016

Group 3

2017

2018

2019

Group 4

Fig. 20.6 Dynamics in infant mortality from congenital anomalies of the circulatory system (per 100,000 population) in groups with the maximum (Group 1) and minimum (Group 2) environmental index and maximum (Group 3) and minimum (Group 4) industrial environmental index, Russian Federation

to the relatively clean regions (52.6 per 100,000 population). This situation is explained by the composition of Group 2: it includes higher industrially developed regions, although most of them are not included in Group 3. The minimum average mortality is registered in the most industrially developed regions of Group 3 (40.7 per 100,000 population). In groups by ranking of the industrial environmental index, mortality over the 10-year period decreased to a greater extent compared to groups by ranking of the environmental index. Decrease in mortality in Group 3 was less notable compared to Group 4 (by 62.4% and 64.0%, respectively), which is likely to be associated with the initially lower mortality. Decrease in mortality in Group 1 was almost the same as in the polluted regions of Group 2 (by 59.1% versus 58.5%). In Groups 1 and 4, there was an increase in mortality after 2012, while in Groups 2 and 3, there was a marked decrease in mortality. The average annual mortality growth rates before and after 2012 equaled to −1.3% and 0.6% in Group 1, −3.5% and − 8.5% in Group 2, −4.9% and − 9.9% in Group 3, and − 8.5% and − 0.5% in Group 4, respectively.

20.3 Discussion At the peak of the CHD child mortality in 1997, the share of infants in child mortality was lower compared to the current period, due to higher mortality in primary school children (5–9 years), which is likely to be due to the lack of adequate nutrition and medicines during this period: children with pathological conditions that could have been treated in other context without deficits died. The contribution of infant mortality from CHD in the overall mortality in ages 0–14, the one that

20 Factors Reducing Child Mortality from Congenital Heart Defects in Russia

329

determines identity of the dynamics in infant and child mortality, only at different levels (Nigmirova et al., 2016). Approximation of infant mortality and mortality in ages 1–4 by a third-degree polynomial indicates a twofold change in the influence of risk factors for CHD deaths in children. The first change can be explained by the opening of the Burakovsky Center, which accumulates infants with CHD. The second change can be associated with the introduction of active management for low birth weight babies. The fact that dynamics in mortality in children aged 5–14 is described by a linear function against the background of approximating dynamics in mortality in children under 5 by a polynomial function can be considered as a result of longer survival of children with those undiagnosed heart defects that are less fatal due to advances in medical technologies (Marelli et al., 2014). The opinion about the leading influence of surgical methods of CHD treatment on child mortality is confirmed by the following facts. First, the share of infants among patients with surgery for CHD after 1998 increased and the number of surgeries with cardiopulmonary bypass in infants with CHD compared to older patients notably increased as well. The revealed correlation between the number of surgeries in infants with CHD and the number of deaths from CHD in children under 15 indicates an extremely high medical and social effectiveness of high-tech methods of surgical treatment. Even in the second half of the twentieth century, it was shown that best early and long-term outcomes can be achieved by interventions in neonates and infants, even treating complex, cyanostic congenital heart defects (Podzolkov, 2017). The maximum average annual rate of decline in child mortality over the entire analyzed period is observed in children aged 1–4, suggesting the higher rate of effective CHD surgery not only in this age group, but also in infants who used to survive beyond 1 year of life. It should be emphasized here that preventing infant deaths makes the greatest contribution to increasing life expectancy of the population. Second, since 1998, the average annual rate of decline in the CHD mortality in all age groups has significantly increased. The maximum rates of reduction in mortality from congenital heart defects in children are registered in children aged 1–4, suggesting the increased rate of effective CHD surgery not only in children of this age group, but also in infants who used to survive beyond 1 year of life. Third, regional variations in infant mortality from CHD have significantly decreased since 1998. On the one hand, a large variability in the CHD prevalence in Russia can be hardly explained by natural causes and uniqueness of the environmental situation, it is more likely to be accounted for difference in CHD registration methods, quality and principles of diagnostics in different regions (Aldasheva et al., 2018; Peredviguina, 2009). It is the relationship between regional variations in infant mortality from CHD and completeness of detection of congenital anomalies that made it possible to identify a positive correlation between infant mortality and availability of obstetricians and gynecologists in the regions who detect congenital malformations in newborns. However, no correlation with the availability of cardiovascular surgeons has been registered. A direct correlation

330

A. V. Zubko et al.

between the availability of doctors in the regions and incidence of congenital anomalies in children under 15 also highlights the problem of underdiagnosis of this pathology in children in the regions with low availability of doctors. On the other hand, to utilize capacity of the Burakovsky Center, its employees began to provide counselling services and methodological support to health care facilities in some regions of the Russian Federation on detection, as well as nursing, rehabilitation and further socialization of the operated patients rather than on surgical treatment alone (Bokeria et al., 2016). Since detecting patients with malformations, especially in early stages depends upon access to health care and availability of modern high-tech equipment and trained staff (Podzolkov, 2017), it is logical to make a conclusion that medical institutions in some regions face staffing problems and have a weak material and technical base. A more comprehensive detection of CHD in regions with the originally low mortality due to improper diagnostics and surgical interventions in children in the regions with high mortality and insufficient capacity have resulted in the reduced regional variations in the CHD infant mortality by 2012. Conclusion on the impact of new technologies for active management for preterm babies born from 22 weeks’ gestation and/or babies with low and extremely low birth weight in all regions of the Russian Federation on the decline rates of infant mortality from CHD is confirmed by almost a twofold increase in the average annual rate of decline in infant mortality from CHD since 2012. Early diagnosis of congenital anomalies made it possible to increase completeness of CHD detection and ensure timely surgical interventions with better postoperative infant care. Since new technologies of pediatric care are applicable for newborns, their introduction did not affect the postoperative survival of older children, but influenced only the increase in the share of children with CHD operated within the first weeks of life. Another aspect of care for low birth weight preterm babies was the increase in regional variations in the CHD mortality, which is associated with different efficiency of perinatal centers and their different accessibility in the country. In addition, in complicated cases of CHD, which are more common in premature babies, the postoperative survival rate depends upon quality of follow-up in health care facilities at the place of residence, which differs significantly across individual regions. This situation suggests significant reserves for reducing CHD mortality through improving provision of the remote regions with high-tech equipment and qualified specialists. The influence of the level of economic development of the regions on child mortality from CHD is confirmed by the fact that the minimum average mortality rate was observed in the group of the most industrialized regions. At the same time, those regions also report the maximum average CHD incidence reflecting the relationship between detection of congenital malformations and availability of medical personnel and the level of material and technical base of the health care facilities. The fact that, the average mortality rate in the industrially undeveloped regions is lower compared to the regions with the maximum environmental index can be explained by the problem of territorial accessibility of health care facilities,

20 Factors Reducing Child Mortality from Congenital Heart Defects in Russia

331

which is characteristic of the most “clean” regions (the Republics of Adygea, Kabardino-Balkaria and Karachay-Cherkessia, Altai and Kamchatka Territories, Jewish Autonomous Region, Kostroma and Magadan regions). Differences in the average annual rates of mortality growth before and after 2012 in the groups by ranking of industrial environment index and environmental index suggest a relationship between introduction of new high-tech methods and the level of economic development of the regions: in economically developed regions, the decrease in infant mortality from CHD following introduction of modern technologies of care had the highest rate. In underdeveloped regions and regions with limited territorial accessibility of care, child mortality from CHD increased in 2013. No effect of environmental pollution on child mortality from CHD has been detected, however this effect is relevant to morbidity: the maximum increase in morbidity over the period 2009–2019 was registered in the polluted regions. Despite better detection of CHD, industrialized regions rank only second in terms of the increase in morbidity. Under-detection of CHD is registered in the least industrialized regions with the minimum average incidence.

20.4 Conclusions Commission of the new building of the Burakovsky Institute of Cardiac Surgery had a crucial impact on mortality reduction in children with CHD in Russia due to a tenfold increase in the number of CHD surgeries in infants. The impact level of modern technologies of active management for preterm babies born from 22 weeks’ gestation and/or babies with low and extremely low birth weight on child mortality from CHD is low due to insufficient level of economic development of some Russian regions and limited territorial accessibility of health care facilities in the mountain republics and regions of Siberia and the Far East. Environmental protection measures contribute to reducing the incidence of congenital heart diseases in children, without affecting mortality rates. Conflict of Interests The authors declare no conflict of interest.

References Aldasheva, N. M., Lobzova, A. B., & Kuznetsova, T. V. (2018). Influence of environmental factors on the frequency of fetal congenital malformations. Physiology, Morphology and Pathology of Humans and Animals in Kyrgyzstan, 8, 381–386. Baibarina, E. N., Degtyarev, D. N., & Kucherov, Y. I. (2011). Improving early surgical care for children with congenital malformations. Russian Bulletin of Perinatology and Pediatrics, 56(2), 12–19. Berg, E., Lie, R. T., Sivertsen, A., & Haaland, O. A. (2015). Parental age and the risk of isolated cleft lip: A registry-based study. Annals of Epidemiology, 25(12), 942–947.

332

A. V. Zubko et al.

Beshlyaga, V., & Lazorishinets, V. (2005). Echocardiography in newborn heart surgery. Doctor, 2, 52–55. Bokeria, L. A. (Ed.). (2014). Clinical guidelines for managing children with congenital heart defects (p. 342). Bakulev National Research Center. (In Russ). Bokeria, L. A., & Gudkova, R. G. (2006). National heart surgery (1995–2004) (p. 42). Bakulev National Research Center. (In Russ). Bokeria, L. A., & Gudkova, R. G. (2013). Heart surgery-2012. Diseases and congenital anomalies of the circulatory system (p. 210). Bakulev National Research Center. (In Russ). Bokeria, L. A., & Gudkova, R. G. (2014). Heart surgery-2013. Diseases and congenital anomalies of the circulatory system (p. 220). Bakulev National Research Center. (In Russ). Bokeria, L. A., & Gudkova, R. G. (2015). Heart surgery-2014. Diseases and congenital anomalies of the circulatory system (p. 226). Bakulev National Research Center. (In Russ). Bokeria, L. A., & Gudkova, R. G. (2016). Heart surgery-2015. Diseases and congenital anomalies of the circulatory system (p. 208). Bakulev National Research Center. (In Russ). Bokeria, L. A., Kim, A. I., Tumanyan, M. R., Berishvili, D. O., & Bespalova, E. D. (2016). Evolution of treatment of congenital heart defects in newborns and infants at the A. N. Bakulev Scientific Center for Cardiovascular Surgery. Bulletin of the A. N. Bakulev National Research Center of the Russian Academy of Medical Sciences, 17(3), 4–17. Cao, H., Wang, J., Li, Y., et al. (2017). Trend analysis of mortality rates and causes of death in children under 5 years old in Beijing, China from 1992 to 2015 and forecast of mortality into the future: An entire population based epidemiological study. BMJ Open, 7(9), 1–11. https:// doi.org/10.1136/bmjopen-2017-015941 Correa, A., & Marcinkevage, J. (2013). Pregnancy obesity and the risk of birth defects: An update. Nutrition Reviews, 71(Suppl.1), S68–S77. Dart, A. B., Ruth, C. A., Sellers, E. A., Au, W., & Dean, H. J. (2015). Maternal diabetes mellitus and Congenital Anomalies of the Kidney and Urinary Tract (CAKUT) in the child. American Journal of Kidney Diseases, 65(5), 684–691. Dolk, H., Loane, M., & Garne, E. (2010). The prevalence of congenital anomalies in Europe. Advances in Experimental Medicine and Biology, 686, 349–364. Eiriksdottir, V. H., Valdimarsdottir, U. A., Asgeirsdottir, T. L., Gisladottir, A., Lund, S. H., Hauksdottir, A., & Zoega, H. (2015). Smoking and obesity among pregnant women in Iceland 2001-2010. European Journal of Public Health, 25(4), 638–643. EUROCAT. Prevalence tables. (2020). http://www.eurocat-network.eu/AccessPrevalenceData/ PrevalenceTabl. Date of the search: 28.10.2020. Gabbay-Benziv, R., Reece, E. A., Wang, F., & Yang, P. (2015). Birth defects in pregestational diabetes: Defect range, glycemic threshold and pathogenesis. World Journal of Diabetes, 6(3), 481–488. Garne, E., & de Walle, H. (2009). Parental age and congenital malformations. Paediatric and Perinatal Epidemiology, 23(3), 279. Gilboa, S. M., Mai, C. T., Shapiro-Mendoza, C. K., Cragan, J. D., Moore, C. A., Meaney-Delman, D. M., Jamieson, D. J., Honein, M. A., & Boyle, C. A. (2017). Population-based pregnancy and birth defects surveillance in the era of Zika virus. Birth Defects Research, 109(5), 372–378. Glinianaia, S. V., Embleton, N. D., & Rankin, J. A. (2012). Systematic review of studies of quality of life in children and adults with selected congenital anomalies. Birth Defects Research. Part A, Clinical and Molecular Teratology, 94(7), 511–520. Hall, N. B., Broussard, K., Evert, N., & Canfield, M. (2017). Notes from the field: Zika virusassociated neonatal birth defects surveillance - Texas, January 2016-July 2017. MMWR. Morbidity and Mortality Weekly Report, 66(31), 835–836. Jawad, A., Patel, D., Brima, N., & Stephenson, J. (2009). Alcohol, smoking, folic acid and multivitamin use among women attending maternity care in London: A cross-sectional study. Sexual & Reproductive HealthCare, 22, 100461. Kim, A. I., Bokeria, L. A., Podzolkov, V. P., Ilyin, V. N., & Tumanyan, M. R. (2003). Cardiovascular diseases in newborns: Cardiological and surgical problems. Bulletin of the Russian Academy of Medical Sciences, 12, 77–80. (In Russ).

20 Factors Reducing Child Mortality from Congenital Heart Defects in Russia

333

Kinsner-Ovaskainen, A., Morris, J., Garne, E., Loane, M., & Lanzoni, M. (2020). European monitoring of congenital anomalies. JRC-EUROCAT report on statistical monitoring of congenital anomalies (2008–2017). EU Science Hub, 75. https://doi.org/10.2760/575186 Knyshov, G. V. (2003). Cardiac surgery: Objectives and prospects. Doktor, 2, 9–11. (In Russ). Marelli, A. J., Ionescu-Ittu, R., Mackie, A. S., et al. (2014). Lifetime prevalence of congenital heart disease in the general population from 2000 to 2010. Circulation, 130, 749–756. Nimgirova, A. S., Naberezhnaya, Z. B., & Serdyukov, A. G. (2016). Main trends in infant mortality from congenital anomalies in the Astrakhan region. Journal of Scientific Articles on Health and Education in the XXI Century, 8(3), 52–55. (In Russ). O’Leary, C. M., Nassar, N., Kurinczuk, J. J., de Klerk, N., Geelhoed, E., Elliott, E. J., & Bower, C. (2010). Prenatal alcohol exposure and risk of birth defects. Pediatrics, 126(4), e843–e850. Peredviguina, A. V. (2009). Frequency, structure and risk factors for fetal and newborn congenital malformations in the Udmurt Republic: Author’s abstract . . . Candidate of Medical SciencesPerm, p. 27. Podzolkov, V. P. (2017). Congenital heart defects. Disorders of Blood Circulation and Heart Surgery, 21(35), 26–27. Potera, C. (2018). One in seven babies exposed to Zika virus in utero has birth defects. The American Journal of Nursing, 118(11), 12. Rosano, A., Botto, L. D., Botting, B., et al. (2020). Infant mortality and congenital anomalies from 1950 to 1994: An international perspective. Journal of Epidemiology and Community Health, 54, 660–665. Rothman, K. J., Greenland, S., & Lash, T. L., 3rd. (2008). Modern epidemiology. Wolters Kluwer Health/Lippincott Williams & Wilkins, 758 p. Saperova, E. V., & Vakhlova, I. V. (2017). Congenital heart defects in children: Incidence, risk factors, mortality. Questions of Modern Pediatrics, 16(2), 126–133. https://doi.org/10.15690/ vsp.v16i2.1713. (In Russ). Selyutina, M. Y., Evdokimov, V. I., & Sidorov, G. A. (2014). Congenital malformations as a marker of ecological status of the environment. Nauchnye vedomosti BelGU. Ser. Meditsina. Farmatsiya, 11(182), 173–177. (In Russ). Tverskaya, A. V., & Verzilina, I. N. (2018). Study of the influence of air pollutants on neonatal incidence of congenital anomalies in the Belgorod region. Nauchnye vedomosti BelGU. Ser. Meditsina. Farmatsiya, 41(2), 297–304. (In Russ). Van der Linde, D., Konings, E. E., Slager, M. A., et al. (2011). Birth prevalence of congenital heart disease worldwide: A systematic review and meta-analysis. Journal of the American College of Cardiology, 58(21), 2241–2247. https://doi.org/10.1016/j.jacc.2011.08.025 WHO mortality Database. (2020). https://www.who.int/data/data-collection-tools/who-mortalitydatabase WHO. Congenital anomalies. Information Bulletin. (2020). https://www.who.int/news-room/factsheets/detail/congenital-anomalies. Date of the search: 20.12.2020. WHO/CDC/ICBDMS. (2014). Birth defects surveillance: A manual for programme managers. World Health Organization, 115 p. Zhu, Z., Cheng, Y., Yang, W., Li, D., Yang, X., Liu, D., Zhang, M., Yan, H., & Zeng, L. (2016). Who should be targeted for the prevention of birth defects? A latent class analysis based on a large, population-based, cross-sectional study in Shaanxi Province, Western China. PLoS One, 11(5), e0155587. Zinkovsky, M., Lazorishinets, V., & Rudenko, N. (2003). Principles of treatment of children with congenital heart defects. Doktor, 2, 23–25. (In Russ).

Part IV

Special Methods

Chapter 21

America’s Zika Virus and Its Similarities with African and Asian Lineages Jesús E. García and V. A. González-López

21.1 Introduction In this paper, we look for evidence that allows us to discard or support the following geographical assumption about the Zika genetic structure, the sequences coming from America are colser to the sequences coming from Asia than those coming from Africa. There is a range of studies in the field of the evolution of Zika that reveals a possibility of transformation of the genetic structure. It has been postulated that the virus originated in East Africa and then spread into both West Africa and Asia around 80–100 years ago. And that the Asian genotype has been gradually evolving and spreading geographically throughout Asia and the Pacific Islands. Here the problem is placed since this last place is considered a possible entrance door of the genetic version that prevails in the sequences of America. The theoretical elements that allow giving an answer to these questions come from the area of stochastic processes. Among such notions, which we will address later, we highlight three. The first of them is a metric that allows deciding whether two samples, coming from stochastic processes, follow the same stochastic law or not (see García et al. 2018). The second is an indicator that, based on a collection of samples from independent stochastic processes, it allows selecting the sample that best represents the collection (see Fernández et al. 2020). In relation to the third notion, this is constituted by a model that is build using a collection of samples from stochastic processes and which allows us to reveal what the samples have in common and what they do not have in common, in terms of units of the state space (see Cordeiro et al. 2020). In this article, we consider each genomic and complete sequence of Zika (in FASTA format) as a sample coming from a stochastic process

J. E. García () · V. A. González-López Department of Statistics, University of Campinas, Campinas, São Paulo, Brazil e-mail: [email protected]; [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 C. H. Skiadas, C. Skiadas (eds.), Quantitative Methods in Demography, The Springer Series on Demographic Methods and Population Analysis 52, https://doi.org/10.1007/978-3-030-93005-9_21

337

338

J. E. García and V. A. González-López

of finite order and arranged in a finite alphabet, which is the genetic alphabet. Such abstraction has already been used in the case of sequences coming from America, but with the purpose of modeling the profile of Zika in America (see Cordeiro et al. 2020). The organization of this paper is as follows; Sect. 21.2 introduces the notions we will use as well as the notation. Section 21.3 describes the data and its source. Section 21.4 shows the results and Sect. 21.5 shows the conclusion.

21.2 The Markovian Model The section gives the theoretical framework of the paper. The notions of proximity between processes represented by samples and the criterion of classification of samples are introduced. In the latter case, a sample that best represents the process is identified from a collection of samples. Such a representative sample, by Zika’s lineage, can be used to compare Zika’s lineages. The section concludes by showing the model that allows us to identify, from a collection of samples (a sample by provenance), the states that operate stochastically equivalently (in terms of the transition probabilities). Let (Xt ) be a discrete time Markov chain on a finite alphabet A with finite order n = a a o. Let us call S = Ao the state space, denote am m m+1 . . . an where ai ∈ A, m ≤ i ≤ n. For each a ∈ A and s ∈ S define the transition probability P (a|s) = −1 Prob(Xt = a|Xtt −o = s). If x1n is a sample coming from the stochastic process (Xt ), the number of occurrences of s in the sample x1n is denoted by Nn (s) and the number of occurrences of s followed by a in the sample x1n is denoted by Nn (s, a). Then, NNnn(s,a) (s) is the estimator of the transition probability P (a|s). Consider now, two Markov chains (X1,t ) and (X2,t ), of order o, arranged on the finite alphabet A with state space S. Given s ∈ S denote by {P (a|s)}a∈A and {Q(a|s)}a∈A the sets of transition probabilities of (X1,t ) and (X2,t ) respectively. The local metric ds , introduced by García et al. (2018), is given now, and it allows us to establish how far or near the samples are. Definition 2.1 Consider two Markov chains (X1,t ) and (X2,t ), of order o, with n1 n2 finite alphabet A, state space S = Ao and independent samples x1,1 , x2,1 respectively. i. For a string s ∈ S, n1 n2 , x2,1 ) ds (x1,1

= cα,n1 ,n2

⎧ ⎨  a∈A

⎩

 Nnj (s, a) ln

j =1,2

−Nn1 +n2 (s, a) ln



Nn1 +n2 (s, a) Nn1 +n2 (s)

Nnj (s, a) Nnj (s)

 ,



21 America’s Zika Virus and Its Similarities with African and Asian Lineages

339

ii. n1 n2 n1 n2 dmax(x1,1 , x2,1 ) = max{ds (x1,1 , x2,1 )}, s∈S α with cα,n1 ,n2 = (|A|−1) ln(n , Nn1 +n2 (s, a) = Nn1 (s, a) + Nn2 (s, a), 1 +n2 ) Nn1 +n2 (s) = Nn1 (s) + Nn2 (s), where Nn1 and Nn2 are given as usual, computed n1 n2 from the samples x1,1 and x2,1 respectively. With α a real and positive value.

In García et al. (2018) is proved that ds is a metric, this is ds is such that Nn (s,a) Nn (s,a) n1 n2 a) ds (x1,1 , x2,1 ) ≥ 0 with equality ⇔ N1n (s) = N2n (s) ∀a ∈ A; b) 1

2

n

n

n1 n2 n2 n1 n1 n2 n1 n2 3 3 , x2,1 ) = ds (x2,1 , x1,1 ) and c) ds (x1,1 , x2,1 ) ≤ ds (x1,1 , x3,1 ) + ds (x3,1 , x2,1 ). ds (x1,1 The two notions introduced by Definition 2.1 are statistically consistent, then, by increasing the min{n1 , n2 } grows their ability to detect discrepancies and similarities. In the application we use α = 2 (see Definition 2.1-i.), with this value (α = 2), to decide that the sequences follow the same law when ds < 1, is equivalent to use the Bayesian Information Criterion, see Schwarz (1978) and García et al. (2018). To follow is introduced a notion that makes possible the classification of sequences that belong to a group of sequences. n

j m Definition 2.2 Given a finite collection {xj,1 }j =1 of independent samples from m independent processes {(Xj,t )}j =1 with probabilities {Pj }m j =1 , over the finite alphabet A, with state space S = Ao (o < ∞). For a fixed i ∈ {1, 2, . . . , m} define

n

n

n

j V (xi,1i ) = median{dmax(xi,1i , xj,1 ) : j = i, 1 ≤ j ≤ m}.

Where, given a sequence {zj }lj =1 , median{zj , 1 ≤ j ≤ l} = z(k+1) if l = 2k + 1 z

+z

and median{zj , 1 ≤ j ≤ l} = (k) 2 (k+1) if l = 2k, for k an integer and z(j ) denoting the j th order statistic of the collection {zj }lj =1 . With the V values attributed to each sample, we can proceed to order the samples, from lowest to highest value of V , in order to identify their classification. As we can perceive from the Definition 2.2, low values of V indicate that these samples represent the whole group better, while high values of V indicate little representativeness. The next result, proved in Fernández et al. (2020), gives an adequate tool to classify sequences, according to their underlying laws. According to Theorem 1 of Fernández et al. (2020), under the assumptions of Definition 2.2, for each i, 1 ≤ i ≤ m, set ξi = |{j : 1 ≤ j ≤ m, Pj = Pi }|, i. n

V (xi,1i )

−→

min{n1 ,··· ,nm }→∞

∞ if, and only if, ξi ≤ 

m . 2

340

J. E. García and V. A. González-López

ii. n

V (xi,1i )

−→

min{n1 ,··· ,nm }→∞

0, if, and only if, ξi > 

m . 2

Where x is the smallest integer greater than or equal to x. The result guarantees that if at least 50% of the samples of the set follow the same law; each of them receives a value of V close to zero. And, if this does not happen, V takes arbitrarily large values identifying discrepancies in the generating laws of the sequences. We now show an extension of the previous notions, specifically developed for a collection of independent stochastic processes, see Cordeiro et al. (2020). Consider j p F = {(Xt )}j =1 , a collection of p independent, discrete time, Markov chains on the same finite alphabet A. To simplify the notation, we will assume that all the processes have the same memory o. S = Ao is the state space of each Markov chain in the collection. For each j ∈ J = {1, 2, · · · , p}, a ∈ A and s ∈ S, t −1 t −1 j P j (s) = Prob(Xj t −o = s) and P j (a|s) = Prob(Xt = a|Xj t −o = s). Now we define a space, on which the model is created, this space considers all the p independent processes and the state space, which is unique to all of them. Define M = J × S. To continue, we introduce the notion that defines the model. j

p

Definition 2.3 Consider a collection of p independent processes F = {(Xt )}j =1 of p independent, discrete time, Markov chains on the same finite alphabet A, with memory o. M = J × S, J = {1, · · · , p}, S = Ao , i. the elements (i, s), (j, r) ∈ M, are equivalent if P i (a|s) = P j (a|r) for all a ∈ A; ii. the collection F has Markov partition L = {L1 , L2 , . . . , L|L| } if L is the partition of M defined by the relationship introduced in i. The partition L is minimal, that is, it has the smallest cardinal |L|. Moreover, there are no two parts in the partition that share all the transition probabilities. The following notation, applies the principle that for each part of the partition j p (Definition 2.3), all the elements share the same probability. Then, if F = {(Xt )}j =1 has Markov partition L = {L1 , L2 , . . . , L|L| }, for any L ∈ L, we will denote for all a ∈ A, PL (a) = P i (a|s) for any (i, s) ∈ L.

(21.1)

Then, the model is completely specified once L is estimated and also once it is estimated the set of probabilities conditioned to the structure L. The estimation procedure is developed in Cordeiro et al. (2020) consisting of a metric on the space M that consistently identifies the partition given by Definition 2.3. The model allows us to reveal what a specific group of sequences have in common, thus revealing what is different about them.

21 America’s Zika Virus and Its Similarities with African and Asian Lineages

341

In the next section, we describe the data and its source. We also detail the alphabet, order of the processes necessary for the computation of the notions 2.1, 2.2 and for building the model given by Definition 2.3.

21.3 Data Set All the sequences are in format FASTA obtained from the source of the National Center for Biotechnology Information Support Center (NCBI), available in https:// www.ncbi.nlm.nih.gov/nuccore/. We consider each genomic sequence as a sample, that is to say, that x1n is composed by the concatenation of elements of the alphabet A. As each sequence takes values in the genomic alphabet we define A = {a,c,g,t}, composed by the four bases: adenine (a), cytosine (c), guanine (g) and thymine (t). So, A has cardinal |A| = 4. In Table 21.1, we give the identifier codes of the complete genomic sequences of Asia and Africa. The list of complete sequences coming from America is given in Table 21.2, separated by country (a total of m = 153 samples). The list of cases recorded in Table 21.2 has been used to determine a unique model for the sequences of America, see Cordeiro et al. (2020). The memory o allowed is such that o < log|A| (n) − 1, where n is the sample size coming from the sequence, in this case, n ≥ 10,000 for all the sequences. In the modeling problem of genomic sequences, the elements of A are organized in triples, so o = 3, 6 are the recommended orders. In the present study we use o = 3, then S = Ao = {a, c, g, t}3 . The next section shows the results, bearing in mind that we seek an answer to the conjecture of greater proximity between the sequences of America (all of Table 21.2) and Asia to that found between the sequences of America and Africa. Table 21.1 Complete genomic sequences of Zika by lineage

Lineage Asian

African

Sequence KU312312.1 KJ776791.2 EU545988.1 KF268948.1 KF268949.1 KF268950.1 AY632535.2 LC002520.1 DQ859059.1

Year 2015 2013 2007 2013 2013 2013 2009 2014 2006

Origin Suriname French Polynesia Yap Island Central African Republic Central African Republic Central African Republic Uganda Uganda Uganda

342

J. E. García and V. A. González-López

Table 21.2 Complete genomic sequences of Zika by country: Brazil (BRA), Colombia (COL), Cuba (CUB), Dominican Republic (DOM), Honduras (HND), Jamaica (JAM), Martinique (MTQ), Mexico (MEX), Nicaragua (NIC), Puerto Rico (PRI), United States (USA), Venezuela (VEN) Country Brazil (44)

US (34)

DOM (23)

MEX (19)

HND (13) NIC (7) JAM (4) COL (3) PRI (2) VEN (2) CUB (1) MTQ (1)

Sequences KX197192.1, KY014296.2, KY014297.2, KY014301.2, KY014307.2 KY014308.2, KY014313.2, KY014317.2, KY014320.2, KY558999.1 KY559001.1, KY559003.1, KY559004.1, KY559005.1, KY559006.1 KY559007.1, KY559009.1, KY559010.1, KY559011.1, KY559012.1 KY559013.1, KY559014.1, KY559015.1, KY559017.1, KY559018.1 KY559019.1, KY559021.1, KY559023.1, KY559024.1, KY559027.1 KY559031.1, KY559032.1, KY785410.1, KY785426.1, KY785427.1 KY785429.1, KY785433.1, KY785437.1, KY785439.1, KY785450.1 KY785455.1, KY785456.1, KY785479.1, KY817930.1 KX832731.1, KX842449.2, KX922703.1, KX922704.1, KX922705.1 KX922706.1, KX922707.1, KY014295.2, KY014298.1, KY014316.2 KY014325.2, KY014326.1, KY075932.1, KY075933.1, KY075934.1 KY075935.1, KY075936.1, KY325464.1, KY325465.1, KY325466.1 KY325467.1, KY325468.1, KY325469.1, KY325471.1, KY325472.1 KY325473.1, KY325476.1, KY325477.1, KY325479.1, KY785412.1 KY785445.1, KY785457.1, KY785459.1, KY785474.1 KY014300.2, KY014302.3, KY014303.2, KY014304.2, KY014305.2 KY014314.2, KY014318.3, KY014321.2, KY785413.1, KY785415.1 KY785420.1, KY785423.1, KY785435.1, KY785441.1, KY785447.1 KY785449.1, KY785453.1, KY785463.1, KY785465.1, KY785470.1 KY785475.1, KY785476.1, KY785484.1 MF801391.1, MF801395.1, MF801396.1, MF801398.1, MF801402.1 MF801403.1, MF801404.1, MF801406.1, MF801407.1, MF801408.1 MF801409.1, MF801410.1, MF801412.1, MF801413.1, MF801414.1 MF801417.1, MF801418.1, MF801420.1, MF801423.1 KY014306.2, KY014310.2, KY014312.2, KY014315.2, KY014319.2 KY014327.2, KY785414.1, KY785418.1, KY785442.1, KY785444.1 KY785448.1, KY785452.1, KY785461.1 MF434516.1, MF434517.1, MF434518.1, MF434520.1, MF434521.1 MF434522.1, MF801426.1 KY785419.1, KY785424.1, KY785430.1, KY785432.1 KY785417.1, KY785466.1, KY785469.1 KY785462.1, KY785464.1 KX702400.1, KX893855.1 MF438286.1 KY785451.1

21 America’s Zika Virus and Its Similarities with African and Asian Lineages

343

21.4 Results Since most of the sequences are from Brazil, we first compared the 44 sequences from Brazil with the African and Asian lineages. Section 21.4.1 is intended for such a comparison. Section 21.4.2 the comparison is made between the American sequences and the Asian and African sequences. Finally, in Sect. 21.4.2 and through Definition 2.3 we expose the differences and similarities between the sequences properly identified by their origins: America, Asia, and Africa.

21.4.1 Comparison Between Brazilian, Asian and African Sequences In Fig. 21.1, we show the dmax values organized in a dendrogram, see Definition 2.1-ii. This graphic includes only the sequences of Brazil and those of Asian and African lineages. We see that the sequences from Asia are shown as the closer. And, the closest to the Brazilian ones are KU312312.1 (Suriname) and KJ776791.2 (French Polynesia). Such a conclusion is verified by Table 21.3, where we record the values of V , according to Definition 2.2 and considering as the whole set of sequences the 44 sequences from Brazil, the 3 sequences from Asia, and the 6 sequences from Africa. V offers us an excellent notion of how Asian sequences can be seen as much closer to Brazilian in comparison with the African ones (with the highest values of V ). Furthermore, it is possible to point out that sequence KU312312.1 (Suriname) could be considered the closest to the Brazilian set. The dendrogram (Fig. 21.1) reveals that there are clusters of Brazilian sequences that are far from the majority, for example (1) composed by KY785439.1, KY559004.1 and KY817930.1 and (2) composed by KY559001.1, KY559009.1, KY559010.1, KY014308.2 and KY559014.1. Consistently, we also note that the 8 sequences previously mentioned are those that receive the highest values of V (among the Brazilian ones), according to the records in Table 21.3. KY785439.1 stands out with the highest value V = 0.19626.

21.4.2 Comparison Between Sequences Coming from America, Africa and Asia The general comparison can be visualized using the dendrogram of Fig. 21.2 constructed from the dmax values (see Definition 2.1-ii). We see (Fig. 21.2) that among the clusters furthest from the majority are two that contain the African sequences (1) composed of DQ859059.1 (Uganda), AY632535.2 (Uganda) and LC002520.1 (Uganda) and (2) KF268949.1 (Central African Republic), KF268948.1 (Central

I

RA

.B .B

KY RA 78 54 39 .1. KY BR 55 A 90 04 KY .1 81 .B RA 79 30 .1 .B R A

RA

.1.B

8.2.B

014

1430

KY

559

KY0

RA

A 20.2.BR

KY559001.1.BRA

KY0143

RA

RA

.1

54

AS

B .1.

92 .1

78

.2 .

55

71 10

.1.

26

KF

A

297

9 68

2

KF

.2.B

143

01.2

K

.BR

KY78

FR

.1.A

950

68 F2

RA

KY0

R

AF

.1.

48

BR

014

FR

.A

.1

9 94

8

RA

50

KY

91

54

54

6.2.B 1429 KY0

67

78

19 78

KY

.2.B 317 014

KY 77 KJ

KY

KX KY

KY559 010.1.B RA

J. E. García and V. A. González-López KY559009.1.BRA

344

DQ8

5479.1

R

.AF

59.1

590

A

FR

35.2.A

.BRA

5 AY632

KU312312

.1.ASI

AFR

LC002520.1.

KY014307.2.BRA EU545988.1.ASI .1.BRA

KY785433

KY014

313.2.B

RA

27.1.B

54 KY78

KY7

A

.BR

56.1

854

559

R 1.B

.

26

54

KY

55

RA

.B 7.1

55

A

.1

KY559023.1.BRA

RA

29.1.BR KY7854

RA .1.B

1.1.B

027 559

5902

.1. 99 89 55

A

A BR

.1. 15

KY

A

RA

KY

RA

.B

BR

.1.

12

90

A

A

.BR

.BRA

KY5

13 90

R

.1

90

.1.B

024

55

A

RA

.B

06

55

BR

.B

.1

90

55

559

KY

03

55

KY

KY

18.1

590

019.1

RA

KY559017.1.B

KY559

KY5

KY

.1

90

KY

55

KY

32

55

.B

90

.1.

90

KY

RA

. 05

31

55

A

BR

1.

KY

A

.1.B

RA

90

KY

0

90

.BR

011

A

78

KY

37.1

KY

BR

KY7

RA

854

Fig. 21.1 Dendrogram built from dmax values, see Definition 2.1-ii, for the sequences of Brazil (Table 21.2), African and Asian lineages (Table 21.1). Brazil (BRA), Asia (ASI), Africa (AFR)

African Republic), KF268950.1 (Central African Republic), KY785449.1 (Dominican Republic), KY785439.1 (Brazil), KY785444.1(Honduras). Once again, the Brazilian sequence KY785439.1 is shown to be particularly distant from the Brazilian ones and, also, from the American ones. In relation to the Asian sequences, we see that they are confused with the American ones, mainly KU312312.1 (Suriname) and KJ776791.2 (French Polynesia). We apply the notion V —Definition 2.2—to 3 scenarios (a) sequences from America, (b) sequences from Asia, and (c) sequences from Africa, separately, in order to identify that sequence with the lowest V in each case. The assumption that governs this computation is the following, each group (a), (b), or (c) shows a certain homogeneity in its composition, so we proceed to select that sequence by a group that can be considered a good representative of the group. Given the definition of V , the most representative sequence of the group is the one with the lowest V .

21 America’s Zika Virus and Its Similarities with African and Asian Lineages

345

Table 21.3 V values, see Definition 2.2, for the sequences of Brazil (Table 21.2), African and Asian lineages (Table 21.1). Brazil (BRA), Asia (ASI), Africa (AFR) Sequence KY559005.1.BRA KY559027.1.BRA KY559007.1.BRA KY558999.1.BRA KY559015.1.BRA KU312312.1.ASI KY785410.1.BRA KY014307.2.BRA KY559013.1.BRA KY785479.1.BRA KY559012.1.BRA KY785426.1.BRA KY785450.1.BRA KY785433.1.BRA KY014296.2.BRA KY014320.2.BRA KY785455.1.BRA KY785427.1.BRA KY559021.1.BRA KY014301.2.BRA KY559023.1.BRA KY559024.1.BRA KY014317.2.BRA KY559017.1.BRA KY785456.1.BRA KJ776791.2.ASI KY014297.2.BRA

Median of Dmax 0.02006 0.02058 0.02107 0.02151 0.02224 0.02237 0.02317 0.02347 0.02411 0.02588 0.02666 0.02702 0.02759 0.02795 0.02796 0.02796 0.02824 0.02856 0.02865 0.02866 0.02994 0.03034 0.03069 0.03152 0.03184 0.03204 0.03206

Sequence KY559003.1.BRA KX197192.1.BRA KY559006.1.BRA KY559019.1.BRA KY559018.1.BRA KY785429.1.BRA KY014313.2.BRA KY559031.1.BRA KY559032.1.BRA KY559011.1.BRA KY785437.1.BRA KY559010.1.BRA EU545988.1.ASI KY559009.1.BRA KY014308.2.BRA KY559001.1.BRA KY559014.1.BRA KY817930.1.BRA KY559004.1.BRA KY785439.1.BRA KF268948.1.AFR KF268950.1.AFR KF268949.1.AFR LC002520.1.AFR AY632535.2.AFR DQ859059.1.AFR

Median of Dmax 0.03332 0.03631 0.03671 0.03693 0.03694 0.03754 0.04061 0.04452 0.04885 0.05050 0.05109 0.05490 0.05673 0.07528 0.07719 0.08371 0.10132 0.14787 0.15016 0.19626 0.20933 0.20936 0.23211 0.36991 0.38538 0.41088

Table 21.4 shows the classifications in cases (b) and (c), and for (a) the indicated sequence is KY014318.3 (Dominican Republic). Thus, among the Asian sequences, the most representative is KU312312.1 (Suriname), and among the African ones, the most representative is KF268948.1 (Central African Republic). Table 21.5 exposes the dmax values between the most representative sequences. With these elements given by the procedure of choosing the most representative sequence by a group and the value obtained by dmax between such sequences, we can conclude that there is greater proximity between the America’s set and the Asian’s set, see Table 21.5. Now we look at the constitution of each of those 12 parts. We see in Table 21.7, their compositions. Let’s think about the configurations, that is to say in the structures xyzi , where i registers to which genomic sequence the configuration belongs. i = 1 refers to sequence KF268948.1 (Central African Republic), i =

J. E. García and V. A. González-López KY785462.1.PRI KY785429.1.BRA DOM KY785463.1. .BRA .1 KY559012 .1.BRA 06 KY5590 .BRA 9003.1 .BRA KY55 .1 RA 9019 KY55 018.1.B RA .B 59 KY5 9024.1 .BRA 5 23.1 RA KY5 590 .1.B A KY5 017 .BR X 559 1.1 E KY 5902 3.1.M OM 5 L KY 0142 0.1.D.CO L 8 7 1 O MF 7854 69. 1.C RA 4 . KY 785 466 .1.B BRA X KY 785 456 .1. ME A 6 . KY 785 42 3.1 .US 5 KY 8 141 32.1 7 KY 80 59 F M Y07 K

M JA EX 1. 2. 1.M A 43 20. 1.US SA 5 78 014 66. 1.U RA KY F8 254 74. .1.B OL M Y3 854 14 .C M K Y7 90 7.1 DO K Y55 541 7.1. SA K Y78 544 .1.U D K 78 457 1.HN KY 785 461. .BRA KY 785 04.1 EX KY 5590 9.1.M A KY 80140 .1.BR 0 F 3 M 8179 .NIC KY 34520.1 FR MF4 68949.1.A R KF2 8948.1.AF KF26 50.1.AFR KF2689 .1.DOM KY785449 BRA KY785439.1. KY785444.1.HND

DQ859059.1.AFR AY632535.2.AFR LC002520 MF43451 .1.AFR 8.1.NIC KY01 MF4 4301.2.BRA KY7838286.1.C KY7 5465.1 UB KY 85423 .DOM KY 78545 .1.DOM KY 78545 3.1.DO M KY 7854 9.1. KY 785 52.1 USA KY 785 410.1 .HND 4 0 .B K 4 1 M Y78 429 8.1. RA M F8 54 7.2 HND K F8 01 50. .BR M Y78 014 403. 1.BR A KY F8 54 10 1.M A 01 014 55 .1.M EX 43 06 .1. E 17 .1 BR X .2 .M A .B EX R A

A R .B A .2 BR M 96 .2. O 42 20 .2.D EN 01 43 03 1.V EN . 1 KY Y0 143 400 .1.V IC K Y0 02 55 1.N C K X7 938 16. .NI K 8 45 .1 ND KX F43 4522 .2.H D M F43 312 2.HN X M 014 319. .ME KY 014 12.1 IC KY 8014 17.1.N EX MF 4345 2.1.M MF 0140 .NIC MF8801426.1 .USA MF 25465.1 SA KY3 5479.1.U KY32 310.2.HND KY014 1.1.NIC MF43452 .2.ASI KJ776791 A KX197192.1.BR KY785419.1.JAM

K KY Y01 KY 32 43 K 32 54 16. KY Y32 547 77.1 2.U KY 014 546 3.1 .US SA 3 4 . 0 KY 143 21. .1.U USA A 2 KY 3254 04.2 .DO SA KY 3254 67.1 .DO M 014 76 .U M S . KX 9 302 1.U A KY 2270 .3.D SA 0 O 5 KY7 14314 .1.US M 854 .2.D A 35.1 OM KX8 .D 3 KY0 2731.1.U OM 14 KX92 325.2.U SA S 2704 .1.US A KX922 A 70 KY7854 6.1.USA 20.1.DO M KY014318 .3.DO KY075936.1.U M SA KX922707.1.USA KY785484.1.DOM

M M F80 M F80 13 M F80 139 98. F MF 80 141 5.1 1.M 1 MF 801 39 8.1 .ME EX 1 . KY 801 417 .1.MME X X 4 .1 7 1 8 MF 54 4.1 .ME EX 8 4 MF 014 2.1 .ME X 0 8 . X KY 0139 8.1.MHND 5 6 KY 5900 .1.M EX 5 7 E KY5 59005 .1.BR X 590 .1.B A KY5 13.1 RA 5 .B KY5 9015.1.B RA 58 KY78 999.1.B RA RA 54 KY014 79.1.BRA 30 KY7854 7.2.BRA 33.1.BR A KY785464 .1.P KY075934.1.U RI SA KY075933.1.USA KY785418.1.HND KY014315.2.HND KU312312.1.ASI BRA KY559027.1..HND 06.2 KY0143 .1.BRA 27 KY7854 27.2.HND M 43 KY01 76.1.DO M O 54 KY78 415.1.D SA 5 .U 8 .1 7 5 SA 4 KY 854 .1.U A KY7 2703 .2.US A 2 49 9 X 4 .1.US M K O 5 842 KX 7593 5.1.D OM 0 47 1.D M Y K 85 1. .DO SA 4 7 KY 7854 05.2 .1.U SA 3 KY 014 472 .2.U USA A 5 . KY 325 429 9.1 .US M KY 01 546 8.1 .DO KY 32 546 0.2 KY 32 430 KY 01 KY

KY785412.1.USA KY014298.1.USA KY014313.2. BRA KY7854 KY7854 14.1.HND KY55 24.1.JAM KY3 9001.1.BR EU5 25471.1.U A MF8 45988.1 SA .ASI KY 0140 KY 78543 4.1.ME KY 7854 0.1.J X KY 0143 37.1. AM KY 785 26. BRA K 55 413 1.US K Y55 903 .1.D A KY Y55 903 1.1.B OM K 5 90 2.1 R M Y5 59 10 .B A K F 59 01 .1. RA KY Y0 801 009 1.1.B BRA 78 143 407 .1.B RA 54 08 .1 R 51 .2 .ME A .1 . B R X .M A TQ

346

Fig. 21.2 Dendrogram built from dmax values, see Eq. (2.1)-ii, for the sequences of America (Table 21.2), African and Asian lineages (Table 21.1)

Table 21.4 V values of each set of Asian sequences (left), African sequences (right). In bold letter the most representative sequence by set

Asian sequence KU312312.1 KJ776791.2 EU545988.1

V 0.02777 0.03778 0.04613

African sequence KF268948.1 AY632535.2 KF268950.1 KF268949.1 DQ859059.1 LC002520.1

V 0.14306 0.14309 0.14309 0.14450 0.15173 0.16114

21 America’s Zika Virus and Its Similarities with African and Asian Lineages

347

Table 21.5 dmax values between the most representative sequences of America, Africa and Asia. In bold letter the least value

KY014318.3 (Dominican Republic) KU312312.1 (Suriname)

KU312312.1 (Suriname) 0.01364

KF268948.1 (Central African Republic) 0.17449

0

0.19697

2 refers to sequence KU312312.1 (Suriname), and i = 3 refers to sequence KY014318.3 (Dominican Republic). If the law is, in fact, the same for the 3 sequences (i = 1, 2, 3) we would expect that all the configurations xyz and independent of i are in the same part, which in fact we verify from Table 21.7 that does not occur. In Table 21.8, we show the states that, having the same configuration, xyz, have been allocated in different parts. The states that are not mentioned in Table 21.8 have been allocated in the same part regardless of the sequence, that is, for example, the state aac for the 3 sequences is found in part L2 . We can clearly see how the xyz states from sequences 2 (KU312312.1 (Suriname)) and 3 (KY014318.3 (Dominican Republic)) stay together, except in the case of cta. This fact strengthens the evidence given by Table 21.5, reporting the most pronounced proximity between sequences from America and Asia.

21.5 Concluding Remarks In this paper we use three powerful tools, originating from stochastic processes to verify an epidemiological conjecture, we give a magnitude to the validity of such conjecture, and we show the reasons that allow it to be verified. We use a metric between processes (see García et al. 2018), an indicator of representativeness between samples of processes (see Fernández et al. 2020) and a model of communality (see Cordeiro et al. 2020) to identify the greatest proximity between the genetic structure of sequences originating in America and those from Asia when comparing them with the proximity between the sequences from America and those from Africa. The ordering between the sequences of Brazil, Africa, and Asia, see Fig. 21.1 and Table 21.3, already shows that the last ones positioned are the African ones, exposing the genetic diversity announced by the literature, see Enfissi et al. (2016), Lanciotti et al. (2016), Zanluca et al. (2015). These pieces of evidence are confirmed in the general comparison, which positions the African sequences as the most distant (see Fig. 21.2). With the help of the metric ds , after applying the index V , (Definitions 2.1 and 2.2), we can concretely quantify this

348 Table 21.6 Transition Probabilities—Eq. (21.1) PˆLi (·), · ∈ A = {a, c, g, t}. In bold letter the highest values

J. E. García and V. A. González-López i of Li 1 2 3 4 5 6 7 8 9 10 11 12

a 0.29821 0.39141 0.37231 0.22066 0.28877 0.19964 0.25857 0.12574 0.30381 0.17818 0.42098 0.13222

c 0.17092 0.23853 0.19614 0.19150 0.21828 0.27293 0.26475 0.20433 0.27060 0.25769 0.20546 0.22882

g 0.35854 0.14075 0.25889 0.42656 0.22811 0.25944 0.29012 0.41320 0.11439 0.36680 0.07040 0.46249

t 0.17233 0.22931 0.17266 0.16128 0.26484 0.26799 0.18656 0.25672 0.31119 0.19733 0.30316 0.17647

Table 21.7 The subscript i indicates the provenance of sequence i, where i = 1 refers to sequence KF268948.1 (Central African Republic), i = 2 refers to sequence KU312312.1 (Suriname), and i = 3 refers to sequence KY014318.3 (Dominican Republic) i of Li Elements 1 aaa1 , aaa2 , aaa3 , gaa2 , gaa3 , aga1 , gca1 , gca2 , gca3 , gga1 , gga2 , gga3 , tag2 , tag3 , aga2 , aga3 , cga1 , ttg2 , ttg3 , gaa1 , ttg1 2 aac1 , acc1 , acc2 , gtc1 , acc3 , agg1 , gtc2 , gtc3 , aac2 , aac3 , gcc2 , gcc1 , gcc3 , ttc3 , ttc2 , cac2 , cac3 , gac2 , gac3 , cac1 , ccc1 , ccc3 , ccc2 , cgc1 , ctc1 3 aag1 , tgg1 , aag2 , ggg3 , ggg2 , aag3 , cag1 , cgg1 , cgg2 , cgg3 , agg2 , agg3 , tgg2 , tgg3 , ggg1 , gag1 , gag2 , gag3 4 aat1 , aat2 , aat3 , atg1 , ctg2 , ctg3 , atg2 , atg3 , tat2 , tat3 , ctg1 , gcg1 , gtg1 , gtg2 , gtg3 5 aca1 , tag1 , aca2 , aca3 , tga1 , tga3 , tga2 , cgc2 , cgc3 , ttc1 , tca1 , tca2 , tca3 6 acg1 , gta2 , gta3 , taa1 , ccg2 , ccg3 , cta1 , tct1 , tta2 , tta3 , cca1 , cta2 , tcg1 , tct2 , tct3 , gta1 , tta1 , cca2 , cca3 , taa2 , tcg2 , tcg3 , taa3 7 acg2 , acg3 , ata2 , ata3 , caa1 , caa2 , caa3 , cag2 , cag3 8 act1 , ggt2 , ggt3 , act2 , act3 , ggt1 , agt1 , ttt2 , ttt3 , cgt1 , agt2 , agt3 , cct1 , cgt2 , att2 , cgt3 , att3 , ctt2 , ctt3 , att1 , cct2 , cct3 9 agc1 , agc2 , agc3 , tcc1 , ggc2 , ggc3 , tac2 , tac3 , tcc2 , tcc3 , ctc2 , ctc3 , ggc1 , tgc1 , tgc2 , tgc3 10 ata1 , ccg1 , cga2 , cga3 , gcg3 , gcg2 , gtt2 , gtt3 , cat1 , ctt1 , cat2 , cat3 , cta3 , gtt1 11 atc1 , gac1 , atc2 , atc3 , tac1 12 gat1 , tgt1 , tat1 , gct1 , gct2 , gct3 , ttt1 , gat2 , gat3 , tgt2 , tgt3

distance (see Table 21.5), and confirm the conjecture. Furthermore, through the model (Definition 2.3) we give meaning to it, since we find the states where the sequences of America and Asia are shown to be equivalent and we see that this occurs in most of the states of the state space, explaining the proximity identified (see Tables 21.6, 21.7 and 21.8).

21 America’s Zika Virus and Its Similarities with African and Asian Lineages

349

Table 21.8 To the right of each state, we report the part where it was included (see Table 21.7), according to the sub-index related to the sequence of origin. KF268948.1 (Central African Republic) is the sequence 1, KU312312.1 (Suriname) is the sequence 2 and KY014318.3 (Dominican Republic) is the sequence 3 State agg ata acg cag ccg cga cgc ctc

Sequences 2 and 3 L3 L7 L7 L7 L6 L10 L5 L9

State cta

Sequence 1 L2 L10 L6 L3 L10 L1 L2 L2

State ctt gac gcg tac tag tat ttc ttt

Sequences 1 and 2 L6

Sequences 2 and 3 L8 L2 L10 L9 L1 L4 L2 L8

Sequence 1 L10 L11 L4 L11 L5 L12 L5 L12 Sequence 3 L10

References Cordeiro, M. T. A., García, J. E., González-López, V. A. & Londoño, S. L. M. (2020). Partition Markov model for multiple processes. Mathematical Methods in the Applied Sciences, 43, 7677–7691. https://doi.org/10.1002/mma.6079 Enfissi, A., Codrington, J., Roosblad, J., Kazanji, M., & Rousset, D. (2016). Zika virus genome from the Americas. The Lancet, 387(10015), 227–228. Fernández, M., García Jesús, E., Gholizadeh, R., González-López, V. A. (2020) Sample selection procedure in daily trading volume processes. Mathematical Methods in the Applied Sciences, 43, 7537–7549. https://doi.org/10.1002/mma.5705 García, J. E., Gholizadeh R., & González-López, V.A. (2018). A BIC - based consistent metric between Markovian processes. Applied Stochastic Models in Business and Industry, 34(6), 868– 878. Lanciotti, R. S., Lambert, A. J., Holodniy, M., Saavedra, S., & Signor, L. D. C. C. (2016). Phylogeny of Zika virus in western hemisphere, 2015. Emerging infectious Diseases, 22(5), 933. Schwarz, G. (1978). Estimating the dimension of a model. The Annals of Statistics, 6(2), 461–464. Zanluca, C., Melo, V. C. A. D., Mosimann, A. L. P., Santos, G. I. V. D., Santos, C. N. D. D., & Luz, K. (2015). First report of autochthonous transmission of Zika virus in Brazil. Memórias do Instituto Oswaldo Cruz, 110(4), 569–572.

Chapter 22

A Relative Entropy Measure of Divergences in Labour Market Outcomes by Educational Attainment Maria Symeonaki

22.1 Introduction In recent years, there has been an increasing interest in early job insecurity and the labour market outcomes of young individuals in order to examine the labour market position of youth in Europe and recognise factors explaining divergences between member states aiming at informing social policy making and provide applicable knowledge. It is well accepted that young people are under greater risk of unemployment, involuntary part-time employment, and take on precarious and flexible jobs more easily in the process of moving from school to the labour market. Nevertheless, there are cross-national discrepancies in the patterns of transitions from school to employment. A considerable amount of literature has been published on school-to-work transitions and the labour market outcomes of young individuals: Quintini et al. (2007) discovered that there is much turnover between labour market categories in all OECD countries, but the average length of young graduates’ transition differs considerably amongst these countries. A more recent study (Quintini & Martin, 2014) studied the school-to-work transitions for sixteen emerging and advanced countries and observed that the employability of young individuals is lower in emerging economies, where school leavers of a young age have a longer transition to the labour market, characterised by a higher percentage of the NEET rate (individuals that are Not in Employment, Education or Training) and informal employment. School to work transition was investigated in a following ILO report (Mathys, 2019) where it was suggested that there is a large variation across sixty countries that were studied, based on

M. Symeonaki () Department of Social Policy, School of Political Sciences, Panteion University of Social and Political Sciences, Athens, Greece e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 C. H. Skiadas, C. Skiadas (eds.), Quantitative Methods in Demography, The Springer Series on Demographic Methods and Population Analysis 52, https://doi.org/10.1007/978-3-030-93005-9_22

351

352

M. Symeonaki

their level of development and their income, while education was found to be a strong, positive factor that influenced young individuals’ transition, especially in developed countries. Early job insecurity and the dynamic process of transiting into the labour market is addressed by multiple indicators and models such as time elapsed between graduation and the first job, rate of transitions from employment to unemployment or inactivity, transition probabilities from school to the labour market (Bosch and Maloney (2007), Brzinsky-Fay (2007, 2014), Christodoulakis and Mamatzakis (2009), Eurofound (2014), Flek and Mysikova (2015), Karamessini et al. (2016a, b, 2019a, b), Symeonaki and Stamatopoulou (2020a, b, c), Gallie et al. (2017), among others). Other studies have considered the construction of a multidimensional index of early job insecurity: Symeonaki et al. (2019a, b, c), in Karamessini et al. (2016a, b, 2019a, b), Symeonaki and Stamatopoulou (2015) and Symeonaki et al. (2018). It has conclusively been shown that higher educational attainments is a major contributing factor for a smother transition into the labour market. The present study examines cross-country differences in labour market outcomes for young individuals aged between 15 and 29 in relation to educational attainment, using raw data drawn from the European Union’s Labour Force Survey (EU-LFS) and three different Kullback–Leibler divergence indicators Kullback and Leibler (1951), Kullback (1959, 1987)). The Kullback-Leibler divergence measure has been extensively used since its definition in various fields of studies including economics, engineering, statistics, physics, psychology, etc. More precisely, we measure the direct divergence between the distributions of employed (and unemployed) young individuals to the educational categories (i.e. Low, Medium and High) and the discrete Uniform distribution where the elements of the finite set of educational categories are equally likely. The divergence between the distributions of employed and unemployed individuals is also explored. Countries are ranked according to their relative entropy values of these measures for the latest at the time available raw data for the year 2016. The Kullback–Leibler relative entropy (or divergence) measure is suggested as a very practical way to measure equality of opportunities of young individuals in employment in respect to their educational achievements and the differences in the distributions of employed and unemployed individuals to the educational levels. The paper is structured as follows: Sect. 2 provides information on the data and the indicators used in the subsequent analysis, while Sect. 3 presents the results for the countries under study, draws conclusions and makes further research suggestions.

22.2 Preliminaries, Data and Measurement In the present study the Kullback–Leibler divergence measure is suggested as a means to measure equality of opportunities of young individuals in employment and the differences in employment and unemployment in respect to the different educational levels. Raw data from the EU-LFS survey is used for implementing

22 A Relative Entropy Measure of Divergences in Labour Market Outcomes. . .

353

the suggested methodology for 25 European countries. The EU-LFS is a unique data source, providing detailed information on labour market participation and the working conditions in European countries. It allows multivariate analysis by sex, age, educational attainment and other socio-demographic characteristics, while common principles and guidelines are used to ensure cross-country comparability. For the purpose of the present research, the focus is on individuals that are aged between 15 and 29. Normally, in EUROSTAT’s definitions as well, a young person is an individual aged between 15 and 24. It was decided that the upper limit was extended to the age of 29 to incorporate more information on the postgraduation employment and simultaneously increase the country samples. Other studies have used the suggested definition of a young individual acknowledging the fact that school-to-work transition has been gradually belated in many countries and frequently is completed after the age of 25 Karamessini et al. (2016a, b, 2019a, b), Symeonaki and Stamatopoulou (2020a, b, c). The fact that some young individuals do remain in education beyond the age of 24 years is in general well-accepted (OECD, 1998, Chapter 3, p. 91). The Kullback–Leibler divergence measure, denoted by DKL and also known as relative entropy measure, is a measure of how one probability distribution is different from a second probability distribution, which is called the reference distribution. The Kullback–Leibler divergence was firstly introduced by Solomon Kullback and Richard Leibler in 1951 as the directed divergence between two distributions. However, Kullback preferred the term discrimination information measure Kullback (1987). More specifically, DKL is given by the following definition: Definition 1 For discrete probability distributions P and Q defined on the same probability space X, the Kullback–Leibler divergence from Q to P is defined as:  

(   P (x) ( . P (x) ln DKL P (Q = Q(x) x∈X

Apparently to:

 is equivalent

( this  ( DKL Q(P = − P (x) ln Q(x) P (x) . x∈X

Moreover, DKL can be interpreted as the expectation of the logarithmic difference between the probabilities P and Q, where the expectation is taken using the probabilities P. It is proven that, the Kullback–Leibler divergence is defined only if: ∀x, Q(x) = 0 ⇒ P (x) = 0 (absolute continuity) . The minimum value is equal to 0 (DKL = 0) when the two distributions are identical. In this study we use DKL in order to quantify the equality of opportunities of youth into the labour market. More specifically, we estimate the direct divergence

354

M. Symeonaki

between the distributions of employed individuals to the educational categories and the Uniform distribution, DKL (EUnif ) and unemployed individuals to the educational categories and the Uniform distribution, DKL (UUnif ), assuming the discrete Uniform distribution with three possible outcomes relating to the educational levels (i.e. Low, Medium and High), each having an equal probability of p = 1/3. E denotes the distribution of employed young individuals into the three educational categories, denoting low, medium and high educational attainment. U is the respective distribution for unemployed young individuals. The divergence between the distributions of employed and unemployed individuals DKL (EU) is also explored. The Kullback–Leibler divergence is a special case of a broader class of statistical divergences called f -divergences as well as the class of Bregman divergences. It is in fact the only such divergence over probabilities that is a member of both classes. Hobson (1971) proved that the Kullback–Leibler divergence is the only measure of difference between probability distributions that satisfies some desired properties, which are the canonical extension to those appearing in a commonly used characterization of entropy. The values of the educational attainment were recoded to reflect low, medium and high educational level and harmonised according to the latest version of International Standard Classification of Education (ISCED) and the divergence measures were estimated for the latest at the time available data for the year 2016.

22.3 Results, Interpretation and Future Work In this section the results of our analysis are presented. More specifically, Table 22.1 reveals the respective DKL values that were estimated using raw data drawn from EU-LFS for the year 2016. DKL (EUnif ) (DKL (EUnif )) is the relative entropy of E(or U) with respect to a distribution that reflects equal distribution of employed (unemployed) young individuals to the three educational categories. DKL (EU) measures the discrepancies between employed and unemployed individuals in relation to their distribution to the educational levels. Apparently, lower values would indicate smaller divergences. Figures 22.1, 22.2 and 22.3 present the respective relative entropy measures for the year 2016 for the European countries under study. Expressed in the language of Bayesian inference, DKL (EUnif ) measures the information gained by revising one’s beliefs from the prior probability distribution Unif to the posterior probability distribution E. In other words, it is the amount of information lost when an equal opportunity distribution Unif is used to approximate E. In the general case where the divergent measure DKL (A  B) is estimated, A represents the “actual” distribution of data, observations, or a precisely calculated theoretical distribution, while B typically represents a theory, a model, a description or an approximation of A. Consequently, it gives a measure of divergence of “reality from a model” and it provides estimation of how much the model has yet to learn.

22 A Relative Entropy Measure of Divergences in Labour Market Outcomes. . .

355

Table 22.1 The values of DKL (EUnif ), DKL (UUnif ) and DKL (E  U) Switzerland Estonia Czech Republic Austria Latvia Lithuania Denmark Sweden the Netherlands Hungary Norway Poland Slovakia Belgium Finland Slovenia Bulgaria France Romania Portugal Cyprus Croatia Spain Italy Greece

DKL (E  Unif ) 0.063 0.126 0.325 0.090 0.141 0.233 0.042 0.120 0.032 0.208 0.008 0.269 0.352 0.108 0.189 0.241 0.221 0.132 0.101 0.029 0.163 0.407 0.019 0.160 0.150

DKL (UUnif ) 0.059 0.106 0.140 0.080 0.126 0.222 0.061 0.111 0.168 0.185 0.101 0.233 0.158 0.047 0.126 0.155 0.077 0.084 0.169 0.029 0.158 0.396 0.052 0.149 0.130

DKL (E  U) 0.036 0.136 0.153 0.118 0.091 0.127 0.018 0.258 0.177 0.159 0.134 0.076 0.126 0.140 0.107 0.012 0.143 0.138 0.021 0.016 0.014 0.042 0.110 0.025 0.006

Source: EU-LFS, 2016

It is evident from the results exhibited in Table 22.1 and Fig. 22.1 that in countries belonging to the social democratic regime (e.g. Norway, the Netherlands, Switzerland, Denmark) the values of DKL (EUnif ) are very low, with Norway exhibiting the smallest value (equal to 0.008) denoting an almost identical distribution of E and Unif. Some southern European countries (Spain and Portugal) perform in a similar way, also exhibiting small values. However, it is important to note here that in first eight countries (Fig. 22.1) with the highest values of divergence measures are postsocialist countries (namely Croatia, Slovakia, Czech Republic, Poland, Slovenia, Lithuania, Bulgaria and Hungary). This means that in these countries the way employed young individuals are distributed to educational categories is influenced by educational attainment the most. We detect a similar behaviour when looking at the divergence between the unemployed distribution to educational categories and the Uniform distribution. Again, the first five countries with the highest scores are post-socialist countries (namely Croatia, Poland, Lithuania, Hungary and Romania),

356

M. Symeonaki

0.5

0.4 0.3 0.2

0

HR SK CZ PL SI LT BG HU FI IE CY IT GR DE LV FR EE SE BE RO UK AT CH DK LU NL PT ES NO

0.1

Fig. 22.1 DKL (EUnif ), EU-LFS, 2016

0.5 0.4 0.3 0.2

0

HR PL LT HU RO NL CY SK SI IT DE CZ GR LV FI SE EE IE NO LU FR AT BG DK CH UK ES BE PT

0.1

Fig. 22.2 DKL (UUnif ), EU-LFS, 2016

whereas no country in this welfare regime is seen in the right hand side of the graph (Fig. 22.2). When the divergences between the distributions of employed and unemployed young individuals to the three educational levels is examined one notices that there are three distinct clusters of countries (Fig. 22.3). In the right-hand side cluster Portugal, Cyprus, Slovenia and Greece are included showing very small differences between the distribution of employed and unemployed individuals to the educational categories. This means that educational attainment in these countries could not predict whether a young individual would be employed or unemployed and therefore would not increase or decrease the chances of being employed or unemployed.

22 A Relative Entropy Measure of Divergences in Labour Market Outcomes. . .

357

0.3

0.2

0

SE LU NL HU CZ DE BG BE FR EE NO LT SK IE AT ES FI LV UK PL HR CH IT RO DK PT CY SI GR

0.1

Fig. 22.3 DKL (EU), EU-LFS, 2016

The relationship to other probability distances (distance measures) could be explored as further research, while the values of the Kullback–Leibler divergence could be estimated for the following or preceding years using the EU-LFS raw data to detect changes or evolution through time.

References Bosch, M., & Maloney, W. (2007). Comparative analysis of labor market dynamics using Markov processes: An application to informality (Discussion paper no. 3038). IZA. Brzinsky-Fay, C. (2007). Lost in transition? Labour market entry sequences of school leavers in Europe. European Sociological Review, 23(4), 409–422. Brzinsky-Fay, C. (2014). The measurement of school-to-work transitions as processes. About Events and Sequences. European Societies, 16(2), 213–232. Christodoulakis, G., & Mamatzakis, C. (2009). Labour Market dynamics in EU: A Bayesian Markov Chain approach (Discussion paper series no. 2009-07). Department of Economics, University of Macedonia. Eurofound. (2014). Mapping youth transitions in Europe. Publications Office of the European Union. Flek, V., & Mysikova, M. (2015). Unemployment dynamics in Central Europe: A labour flow approach. Prague Economic Papers, 24(1), 73–87. Gallie, D., Felstead, A., Green, F., & Inanc, H. (2017). The hidden face of job insecurity. Work, Employment and Society, 31(1), 36–53. Hobson, A. (1971). Concepts in statistical mechanics. Gordon and Breach. ISBN 978-0677032405. Karamessini, M., Symeonaki, M., Stamatopoulou, G., & Papazachariou, A. (2016a). The careers of young people in Europe during the economic crisis: Identifying risk factors. Negotiate working paper no. D3.2. Retrieved from https://blogg.hioa.no/negotiate/files/2015/ 04/NEGOTIATE-working-paper-no-D3.2-The-careers-of-young-people-in-Eurpa-during-theeconomic-crisis.pdf

358

M. Symeonaki

Karamessini, M., Symeonaki, M., & Stamatopoulou, G. (2016b). The role of the economic crisis in determining the degree of early job insecurity in Europe. Negotiate working paper 3.3.https:/ /blogg.hioa.no/negotiate/files/2015/04/NEGOTIATE-workingpaper-D3.3.pdf Karamessini, M., Symeonaki, M., Parsanoglou, D., & Stamatopoulou, G. (2019a). Mapping early job insecurity impacts of the crisis in Europe. In B. Hvinden, T. Sirovatka, & J. O’Reilly (Eds.), Youth Unemployment and early job insecurity in Europe: Concepts, consequences and policy approaches? (pp. 24–44). Edward Elgar. Karamessini, M., Symeonaki, M., Stamatopoulou, G., & Parsanoglou, D. (2019b). Factors explaining youth unemployment and early job insecurity in Europe. In B. Hvinden, T. Sirovatka, & J. O’Reilly (Eds.), Youth Unemployment and early job insecurity in Europe: Concepts, consequences and policy approaches? (pp. 45–69). Edward Elgar. Kullback, S. (1959). Information theory and statistics. Wiley. Republished by Dover Publications in 1968; reprinted in 1978: ISBN 0-8446-5625-9. Kullback, S. (1987). Letter to the editor: The Kullback–Leibler distance. The American Statistician, 41(4), 340–341. https://doi.org/10.1080/00031305.1987, 10475510. JSTOR 2684769 Kullback, S., & Leibler, R. A. (1951). On information and sufficiency. Annals of Mathematical Statistics, 22(1), 79–86. https://doi.org/10.1214/aoms/1177729694. JSTOR 2236703. MR 0039968. Mathys, Q. (2019). From school to work: An analysis of youth labour market transitions (ILOSTAT spotlight on work statistics, no. 9). ILO. OECD. (1998). Getting started, settling in: The transition from education to the labour market, OECD Employment Outlook, OECD. Quintini, G., & Martin, S. (2014). Same same but different: School-to-work transitions in emerging and advanced economies (OECD Social, Employment and Migration Working Papers, No. 154). OECD Publishing. https://doi.org/10.1787/5jzbb2t1rcwc-en Quintini, G., Martin, J., & Martin, S. (2007). The changing nature of the school-to-work transition process in OECD countries (Discussion Paper No. 2582). Institute for the Study of Labor. Symeonaki, M., & Stamatopoulou, G. (2015). A Markov system analysis application on labour market dynamics: The case of Greece. Paper presented at IWPLMS, Athens, Greece, 22–24 June, 2015. Symeonaki, M., & Stamatopoulou, G. (2020a). Describing labour market dynamics through non homogeneous Markov system theory. The Springer Series on Demographic Methods and Population Analysis, 50, 359–376. Symeonaki, M., & Stamatopoulou, G. (2020b). Assessing labour market mobility in Europe. In Demography of population health, aging and health expenditures (The Springer series on demographic methods and population analysis). Springer. Symeonaki, M., & Stamatopoulou, G. (2020c). On the measurement of positive labour market mobility. SAGE Open, 10(3), 1–13. https://doi.org/10.1177/2158244020934489 Symeonaki, M., Stamatopoulou, G., & Karamessini, M. (2018). On the measurement of early job insecurity. In C. H. Skiadas & C. C. Skiadas (Eds.), Demography and health issues – Population aging, mortality and data analysis (pp. 275–288). Springer. https://doi.org/10.1007/978-3-31976002-5 Symeonaki, M., Parsanoglou, D., & Stamatopoulou, G. (2019a). The evolution of early job insecurity in Europe. SAGE Open, 9, 1–23. https://journals.sagepub.com/doi/pdf/10.1177/ 2158244019845187 Symeonaki, M., Karamessini, M., & Stamatopoulou, G. (2019b). Measuring school-to-work transition probabilities in Europe with evidence from the EU-SILC. In J. Bozeman, T. Oliveira, C. Skiadas, & S. Silvestrov (Eds.), Data analysis and applications: New and classical approaches (pp. 121–136). ISTE Science Publishing, to appear. Symeonaki, M., Karamessini, M., & Stamatopoulou, G. (2019c). Gender-based differences on the impact of the economic crisis on labour market flows in Southern Europe. In J. Bozeman & C. Skiadas (Eds.), Data analysis and applications: New and classical approaches (pp. 107–120). ISTE Science Publishing.

Chapter 23

Assessing the Intergenerational Educational Mobility in European Countries Based on ESS Data: 2002–2016 Maria Symeonaki and Paraskevi Tsinaslanidou

23.1 Introduction Although the principle of social justice is theoretically promoted in the prevailing system of economic and social organization, research on social mobility demonstrates the difficulty of the lower social classes for upward and especially long-distance mobility and their intergenerational stagnation. The phenomenon of social stratification, and therefore social inequality, is not a characteristic of our time, but has extensively preoccupied societies in the past. It is a fact that there is a very wide range of different and conflicting theories about the causes that give rise to social inequality and the impact it has on both the individual and society as a whole. The approaches/policies that have prevailed in advanced countries to interpret and address social inequalities are essentially wavering between equality of opportunity and equality of conditions. The approach of equality of opportunities stems from the liberal notion and asserts that policies such as public education equate existing inequalities and give individuals the same chances of achieving social growth. In this way, the development of each one is a result of his/her abilities and does not depend on monetary, racial, geographical or other barriers related to social origin. Equality of opportunities is usually studied in relation to the educational and professional success of the individual, compared to her/his social and demographic characteristics (Breen & Jonsson, 2005). On the contrary, equality of conditions concerns income inequality, wealth and, more broadly, material goods between individuals and concerns intervention policies such as high taxation (Breen & Jonsson, 2005). The approach of equality of opportunities seems to have prevailed

M. Symeonaki () · P. Tsinaslanidou Department of Social Policy, School of Political Sciences, Panteion University of Social and Political Sciences, Athens, Greece e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 C. H. Skiadas, C. Skiadas (eds.), Quantitative Methods in Demography, The Springer Series on Demographic Methods and Population Analysis 52, https://doi.org/10.1007/978-3-030-93005-9_23

359

360

M. Symeonaki and P. Tsinaslanidou

in the field of social policy in recent decades, while the approach of equality of conditions has been significantly sidelined. But to what extent are the policies pursued by European states sufficient to create equal opportunities? The research presented in the present paper is based on the above reflection, on the fact that the welfare state is one of the key institutional forces influencing stratification and mobility by intervening interactively in the relationship of family origin individual successes (Papatheodorou & Papanastasiou, 2016; Papanastasiou, 2018) and on studies where social origin provably continues to affect educational success (Atkinson, 2007; Blanden et al., 2005; Chevalier et al., 2003; OECD, 2010; Peter et al., 2010; Woessmann, 2004). More specifically, using primary data drawn from the European Social Survey, the absolute upward mobility index and the relevant Bartholomew (MB) and Prais-Shorrocks (MP-S) mobility indices are calculated for the years 2002–2016, in order to compare and evaluate the effectiveness of the European welfare states in enhancing intergenerational educational mobility. The paper is structured as follows. Section 23.2 provides information on the connection between social mobility, education and social policy. Section 23.3 presents the data used, their limitations and the methodology of the study. Section 23.4 presents the results and Sect. 23.5 draws conclusions and makes further research suggestions.

23.2 Social Mobility, the Importance of Education and Social Policy The key role of education in the study of social mobility is first proven by its direct connection with status and professional perspective. According to Heath (1981), education is considered as very important, especially for people belonging to the lower classes, as it equips themselves with the necessary provisions to achieve social growth. Similar results were obtained from Glass’s (1954) study, in which the likelihood of mobility of working or middle-class individuals increased when (they) completed secondary education, while Blau and Duncan (1976) proved that education has the strongest impact on professional success. The fact that the individual’s social background continues to be a key factor in his/her educational career pathway proves the necessity of intergenerational research. According to Atkinson (2007), education is class-based, as middle-class children are more likely to achieve educational achievement than working-class children. Similarly, Blanden et al. (2005) found that children from low income families are less likely to succeed in education, while Chevalier et al. (2003) concluded that extensive access to higher education took place while the impact of family background on individuals’ educational achievement had increased. The OECD (2010) study, which compared the results of 15-year-old students to their performance in math, literature and science, concluded that the educational level of parents has the greatest impact on educational performance. The same results

23 Assessing the Intergenerational Educational Mobility in European. . .

361

were provided by Woessmann (2004), who compared the effects of the family environment on children’s educational achievement in 17 European countries and in the United States. Finally, it is commonplace among social researchers that children from disadvantaged families tend to have less success prospects than children from families with more advantages. The fact that regardless of society and historical period, family and social origins affect the future success of individuals has been characterized by Erikson and Goldthorpe (1993) as constant flux. The comparison of educational systems and their results has been studied worldwide in the light of their sociological, economic and historical dimension (West & Nikolai, 2013). In her research, Allmendinger (1989), classified educational systems according to their degree of standardization and stratification and found out that institutions with a high degree of stratification are those that have a decisive influence on the professional status of their graduates. Hoffmeyer-Zlotnik and Warner (Warner, 2007), studying the layout of different education systems, came up with the typology of four educational models. In the first type, with a representative in Germany, primary education is short-lived and there is a distinction between lower and upper secondary education. Higher education includes parallel schools (which provide additional vocational education), academic high schools and technical and non-technical universities. In the second type, represented by Luxembourg, primary education is longer, while in the lower secondary education there are a limited number of school types. In higher education there are different types of general and vocational schools and in higher education there are academicvocational institutions and universities. In the third type, represented by Denmark, there is no distinction between primary and secondary education. More specifically, primary and lower secondary education are included in a single school, while in secondary education there is one type of vocational school and several types of general education. In higher education, the distinction between vocational and university education is slight. Finally, in the fourth type, represented by France, a characteristic feature is the large participation in kindergarten and pre-kindergarten. The duration of primary education is longer and there are no subdivisions of lower secondary education. In higher secondary education there is little vertical differentiation, while higher education has several differences. Although education is one of the first state policies adopted in the nineteenth century and is at the heart of social services, it has been studied in a relatively limited way in the light of different social protection systems (Hega & Hokenmaier, 2002). One of the first approaches to the study of educational and social policy is that of Heidenheimer (1981), who, by examining the public policies developed in Europe and America, argues that the development of Western welfare states is characterized by an “exchange” between public investment in secondary education and investment in social security programs. Heclo (1985) came to similar conclusions about the “compensation” between public investment in education and the expansion of social programs. In the light of the well-accepted typology of Esping-Andersen (1990), Hega and Hokenmaier (2002) studied the relationship of “compensation” between education spending and social security spending (in 18 OECD countries for the period 1960–1990). They concluded that welfare states with similar social security

362

M. Symeonaki and P. Tsinaslanidou

policies are grouped in the same way in the field of educational policy. They found that the countries with the highest expenditures in both education and social security were those with a social democratic system of social protection. Countries with conservative-corporate welfare states have higher spending than countries with a liberal social welfare state in their social security programs, while states with a liberal social protection system tend to spend more on their education policy. Another important finding of the research is the increased participation in general education programs at the secondary level of liberal social states, compared to the other welfare states. In relation to educational inequality, Peter et al. (2010), using the typology of Esping-Andersen, studied the “within” and “between” school differences of students compared to their socioeconomic background, and found that socioeconomic origin influenced more educational outcomes of students in conservative-corporatist welfare states, less so in liberal ones while the lowest influence was found in social democracies. To assess the effect that the educational level of father has on children’s educational success, in their study Papatheodorou and Papanastasiou (2011) applied the method of generalized regression in the 14 oldest Member States of the European Union, combining the Esping-Andersen typology and the southEuropean model. According to the results of the research, the education of the father has a significant effect on the education of the individuals of all countries under examination. The countries in which intergenerational educational mobility is higher are those with a social democratic system, while the lowest intergenerational educational mobility is found in the countries of the south-European model. Symeonaki and Stamatopoulou (2014a) studied the patterns of intergenerational educational mobility in Greece and their changes for different birth cohorts, while also investigating the transmission of educational attainments from both father and mother through generations over time, based on data drawn from the European Social Survey. Distance and similarity measures were proposed to complement traditional methodologies. Symeonaki and Stamatopoulou (2011) explored the transition to higher education as an issue of intergenerational educational mobility, while they regarded intergenerational mobility as a distance measure between probability distribution functions (Symeonaki & Stamatopoulou, 2014b). Stamatopoulou et al. (2013) studied the intergenerational transmission of education with evidence from the ESS and Symeonaki et al. (2012) used the European Survey on Income and Living Conditions (EU-SILC) to study intergenerational occupational mobility. Fuzzy Markov Systems and symbolic, heuristic knowledge were used in Symeonaki et al. (2011) to study intergenerational educational mobility in Greece with data drawn from the ESS. To conclude educational intergenerational mobility defined as the trajectories observed from one generation to another and between different social classes has been used in the literature to measure whether and to what extent the socio-economic status of origins (measured in terms of parental education) transmit from parents to children and consequently can be seen as an indicator of equality of opportunities.

23 Assessing the Intergenerational Educational Mobility in European. . .

363

23.3 Data and Methodology The data used in the present study are drawn from the European Social Survey (ESS),1 which is a long-term comparative research survey designed to record and document the attitudes, beliefs and behavioural patterns of the European populations and to produce comparable social indicators, able to be used for European social policy. Started in 2002, the ESS is conducted every two years in more than 20 European countries. It involves strict random probability sampling, a minimum target response rate of 70% and rigorous methodological criteria and collects data from nationally representative samples of persons aged 15 years and above, regardless of their nationality, citizenship or legal status. The “homeless” and people living in collective dwellings are excluded from the sample. The main advantage for choosing the ESS data is the fact that it provides the necessary information on parental social status and the surveyed individuals, even if they do not live in the same residence. More specifically, our focus is on the relation of highest educational attainment of both parents and individuals, which were harmonised according to the latest version of International Standard Classification of Education (UNESCO, 2012) and then recoded to reflect low, medium and high educational attainment. Data used were drawn from the years 2002 up to the latest at the time available data, i.e. the year 2016, for all countries that participated for at least 4 rounds. The analysis includes therefore 24 countries, which were categorised according to their welfare state into (Esping-Andersen, 1990; Fenger, 2007; Ferrera, 1996): social democratic (Sweden, Norway, Finland, the Netherlands and Denmark), conservative-corporatist (Belgium, France, Germany and Austria), liberal (Ireland and UK), southernEuropean (Spain, Portugal, Greece and Italy) and post-communist (European subtype: Poland, Czech republic, Hungary, Slovakia and Bulgaria and former USSR sub-type: Estonia, Ukraine, Russia and Lithuania). The design weight (dweight) was applied, in order to correct the different probabilities of selection and to make the sample more representative.

23.4 Results Representative results of the transition matrices estimated for different welfare systems are depicted in Table 23.1. For 2004 in the social democratic, conservativecorporatist and post-communist (former USSR type) welfare states, individuals with low educational parents are more likely to move to the next educational category, while individuals with parents of middle and high educational level are most likely to remain in the same educational categories as those of their parents.

1

http://www.europeansocialsurvey.org

364

M. Symeonaki and P. Tsinaslanidou

Table 23.1 Educational intergenerational transition probabilities, father and mother, ESS, 2004 Ed. level of father/ed. level of respondent Low Medium High

Regime Social democratic Norway Low 0.309a 0.428 Medium 0.133 0.510 High 0.102 0.292 Conservative Germany Low 0.323 0.559 Medium 0.134 0.665 High 0.136 0.408 South-European Portugal Low 0.783 0.154 Medium 0.233 0.302 High 0.232 0.275 Liberal UK Low 0.702 0.153 Medium 0.516 0.226 High 0.300 0.251 Post-socialist – European type Poland Low 0.346 0.606 Medium 0.203 0.665 High 0.140 0.376 Post-socialist – USSR type Estonia Low 0.291 0.496 Medium 0.225 0.438 High 0.155 0.356

Ed. level of mother/ed. level of respondent Low Medium High

0.263 0.357 0.605

0.291 0.104 0.136

0.438 0.462 0.341

0.271 0.434 0.523

0.118 0.201 0.456

0.213 0.123 0.242

0.613 0.589 0.344

0.174 0.288 0.414

0.064 0.465 0.493

0.783 0.189 0.180

0.147 0.453 0.320

0.071 0.358 0.500

0.145 0.258 0.450

0.697 0.447 0.286

0.145 0.224 0.261

0.158 0.329 0.453

0.048 0.132 0.484

0.334 0.216 0.211

0.612 0.625 0.408

0.055 0.159 0.382

0.213 0.337 0.490

0.292 0.205 0.222

0.497 0.471 0.330

0.212 0.324 0.449

a Probability

of the respondent having a low educational level given that his/her father has a low educational level in Norway, 2004. The remainder numbers depicted in the Table reflect respective probabilities

Similar transitions are observed in the European post-communist regime with one difference: the downward transition probability relating to individuals whose mother has a high level of education. In the south-European and liberal welfare states, individuals with highly educated parents are more likely to remain in the same state, while those with low-level parents are more likely to remain in the lower education category. Individuals with medium educated parents are more likely to have a downward shift in liberal regimes, while in the south-European states, in the

23 Assessing the Intergenerational Educational Mobility in European. . .

365

Table 23.2 Educational intergenerational transition probabilities, father and mother, ESS, 2016 Ed. level of father/ed. level of respondent Low Medium High

Regime Social democratic Norway Low 0.252a Medium 0.151 High 0.096 Conservative Germany Low 0.271 Medium 0.092 High 0.109 South-European Portugal Low 0.589 Medium 0.193 High 0.087 Liberal UK Low 0.406 Medium 0.152 High 0.106 Poland Low 0.511 Medium 0.178 High 0.157 Post-socialist – USSR type Estonia Low 0.254 Medium 0.157 High 0.074

Ed. level of mother/ed. level of respondent Low Medium High

0.329 0.402 0.237

0.419 0.446 0.667

0.225 0.132 0.125

0.344 0.341 0.254

0.431 0.527 0.621

0.381 0.482 0.256

0.348 0.426 0.636

0.163 0.086 0.140

0.451 0.416 0.254

0.386 0.498 0.606

0.204 0.398 0.261

0.208 0.410 0.652

0.592 0.236 0.151

0.206 0.389 0.205

0.202 0.375 0.644

0.214 0.255 0.218

0.379 0.593 0.675

0.393 0.086 0.129

0.218 0.278 0.205

0.389 0.636 0.666

0.291 0.357 0.200

0.198 0.465 0.643

0.537 0.170 0.192

0.283 0.360 0.260

0.180 0.470 0.548

0.396 0.401 0.230

0.349 0.441 0.696

0.278 0.143 0.102

0.408 0.421 0.248

0.315 0.436 0.650

a Probability

of the respondent having a low educational level given that his/her father has a low educational level in Norway, 2016. The remainder numbers depicted in the Table reflect respective probabilities

case of the father, they are more likely to move upwards, while in the mother’s, remaining at the same level is more likely. In Table 23.2 the transition probabilities for 2016 show small differences compared to those of 2004. In the social democratic states, intergenerational educational mobility is extremely high, as respondents with parents of all educational levels are more likely to achieve a higher level of education. The transitions observed in the post-communist countries of the former USSR are also on the rise, with people with parents with low and middle level education moving to the next level and individuals

366

M. Symeonaki and P. Tsinaslanidou

with high educational attainments remaining in the same state. In conservativecorporate welfare regime countries, individuals with low-level parents are more likely to move to the middle level, while respondents with high-level parents are more likely to achieve a level of education similar to that of their parents. With regard to the second educational level, the influence of the father is most likely to lead to immobility, while the influence of the mother seems to lead to an upward transition. In liberal and post-communist European social states, people with low and high levels of education are more likely to stay at the same level of education as their parents, while people with middle-level parents are more likely to succeed an upward transition. Intergenerational educational mobility in the south-European model appears to be extremely limited, as the only possible movement observed concerns individuals with a middle-level education of the father. We now focus on estimating the intergenerational mobility of individuals in Europe and its evolution for the years 2002–2016. More specifically, three different mobility indices are calculated in order to reveal the extent of the transitions within generations. The ones used in the present analysis are the well-established mobility indices: The Prais – Shorrocks mobility index (Prais, 1955; Shorrocks., 1978): MP S = 1 n−1 (n − tr (P)), where n is the number of states and tr(P) denotes the trace of the transition matrix P, i.e. the sum of its diagonal elements. n n pij |i − j | The Bartholomew (1982) mobility index defined by MB = n1 i=1 j =1 nij , where nij is the absolute and the upward mobility index defined by u = N1 j >i

number of individuals with the j-th educational level whose father/mother has the i-th educational level and N denotes the total number of respondents. Figures 23.1 and 23.2 show the results of the Prais-Shorrocks index for intergenerational educational mobility from 2002 to 2016. It is clear that, in all social protection systems, intergenerational educational mobility in relation to the mother’s educational level is higher than that of the father. More specifically, regarding the influence of the father, it is observed that until 2006 there is a tendency of convergence between the results of liberal and social democratic social states, while after 2006 the educational mobility in liberal regimes is much lower. Similar results are presented between post-communist and conservative social states for most years under study. Overall, the social democratic social states have consistently one of the highest levels of mobility, while the south-European states have by far the lowest educational mobility. The above trend of higher and lower educational mobility of social democratic and south-European countries is also observed in relation to the educational level of the mother. Regarding that transition, very high mobility is observed in the conservative regimes, while the post-communist and liberal social states are placed in an intermediate category.

23 Assessing the Intergenerational Educational Mobility in European. . .

367

Fig. 23.1 MPS in relation to the educational level of father, Rounds 1–8

Fig. 23.2 MPS in relation to the educational level of mother, Rounds 1–8

Fig. 23.3 MB in relation to the educational level of father, Rounds 1–8

Figures 23.3 and 23.4 depict the Bartholomew mobility index for intergenerational educational mobility from 2002 to 2016. In relation to the effect exerted by the father’s educational level, the highest mobility is observed in the liberal states until 2008 and in the social democratic states for the remaining years. A relative

368

M. Symeonaki and P. Tsinaslanidou

Fig. 23.4 MB in relation to the educational level of mother, Rounds 1–8

Fig. 23.5 Upward mobility index in relation to the educational level of father, Rounds 1–8

convergence of results is seen between post-communist and conservative welfare states, while for the south-European regimes are much lower. Regarding the effect of the mother’s level of education, the highest mobility is observed in the social democratic welfare systems, while the conservative, post-communist and liberal regimes show similar results. The south-European countries consistently show the lowest intergenerational educational mobility. Finally, Figs. 23.5 and 23.6 show the upward mobility index in relation to the educational level of the father and mother for the years 2002–2016. Regarding the influence of the father’s level of education, the highest educational upgrade is in the social democratic, liberal and post-communist countries. Conservative regimes are next, while the south-European countries exhibit the lowest mobility. In relation to the educational level of the mother, the results differ significantly. For all the years under study, the highest educational transitions take place in the social democratic and conservative systems. A relatively high upward mobility is recorded in the postcommunist regimes until 2008, while after 2008 quite high mobility is recorded in the liberal states. In the south-European social protection systems, the lowest upward mobility is again identified.

23 Assessing the Intergenerational Educational Mobility in European. . .

369

Fig. 23.6 Upward mobility index in relation to the educational level of mother, Rounds 1–8

23.5 Conclusions and Further Research The aim of this study was to examine the effectiveness of European social protection systems in enhancing intergenerational educational mobility. Using primary data from Rounds 1–8 of the European Social Survey, the analysis focused on comparing the impact of the immediate family educational environment on the educational outcomes of individuals for social democratic countries (Denmark, Finland, the Netherlands, Sweden and Norway), conservative-corporatist (Austria, Belgium, France and Germany), liberal (United Kingdom and Ireland), south-European (Greece, Italy, Spain and Portugal) and post-socialist countries (Poland, Czech Republic, Hungary, Slovakia, Bulgaria, Ukraine, Russia, Estonia and Lithuania). In order to obtain a first picture of the educational intergenerational fluidity of the systems under study, the Prais-Shorrocks mobility index was calculated. Based on the results of the index, it was observed that all social states had high levels of fluidity compared to the levels of education of both parents. Considering the findings of the educational levels of both mother and father, the most “fair” systems are those with a social democratic regime. In the next positions are the post-communist and conservative social states, presenting quite high levels of mobility, while with a small difference, the liberal systems follow in the ranking. The systems of the south-European model have the lowest performance. Bartholomew’s mobility index was also calculated to consider the distances travelled. The findings prove relatively high efficiency for all systems under study. More specifically, in relation to the educational levels of both parents, the social democratic regimes show the highest efficiency, while in the ranking the regimes of the liberal, conservative and post-communist models follow with similar results. In the south-European countries, the level of education of the family of origin exerts the greatest influence on the distances of travel. Through the calculation of the upward mobility index, the clearest possible impression of the educational upgrades was attempted. For all systems studied, the values of the index show moderate mobility. Higher education upgrades are taking

370

M. Symeonaki and P. Tsinaslanidou

place in the social democratic welfare states, while the results of the conservative, liberal and post-communist regimes are close to those of the countries belonging to the social democratic regime. In the south-European countries, the results of educational upgrades are quite limited, which proves the strong influence of parental educational background on the educational achievements of individuals. It is clear that social democratic welfare states, which are characterized by extensive universal interventions and high levels of redistribution, achieve the highest attenuation of intergenerational educational immobility. Liberal regimes significantly reduce the impact of family backgrounds on educational success, as (although generally characterized by low and selective social benefits) in the field of their educational policy they significantly promote equal opportunities by presenting very high government spending. Slightly lower than those of the liberal states, are the results of the post-communist and conservative social states, which are characterized by extended social policy benefits, while simultaneously have high standardization and stratification levels in their educational models. SouthernEuropean welfare states, where social protection is extremely residual and the burden is significantly borne by the family, show the highest educational stagnation, hindering individuals’ efforts for educational upgrading and creating “dependency paths” between intergenerational successes. It is recognized that the present study does not take into account all the complexity of the factors that may affect the intergenerational educational movements of individuals, nor the complexity of all the effects that the specific social protection system has on them. Although an attempt was made to examine the correlation between intergenerational educational mobility and economic inequality (of all years and countries under study), both with indicators of economic inequality and poverty from EUROSTAT and OECD databases, and with the calculation of poverty indicators from income variables in the ESS, the results did not show the required levels of statistical significance. Further investigation of the phenomenon is indicated. In conclusion, what we consider to be clear from the present analysis and is of utmost importance is the fact that social interventions in the intergenerational cycle can serve as essential defence factors in the promotion of educational equality and mobility, when the policies implemented give the required weight primarily on creating equal conditions and secondarily on creating equal opportunities.

References Allmendinger, J. (1989). Educational systems and labour market outcomes. European Sociological Review, 5(3), 231–250. Atkinson, W. (2007). Beck, individualization and the death of class: A critique. The British Journal of Sociology, 58(3), 349. Bartholomew, D. J. (1982). Stochastic models for social processes. Wiley. Blanden, J., Gregg, P., & Machin, S. (2005). Intergenerational mobility in Europe and North America. Centre for Economic Performance. Blau, P., & Duncan, O. (1976). The American occupational structure. Wiley.

23 Assessing the Intergenerational Educational Mobility in European. . .

371

Breen, R., & Jonsson, J. (2005). Inequality of opportunity in comparative perspective: Recent research on educational attainment and social mobility. Annual Review of Sociology, 31, 223– 243. Chevalier, A., Denny, K., & McMahon, D. (2003). A multi-country study of inter-generational educational mobility. Institute for the Study of Social Change. Erikson, R., & Goldthorpe, J. (1993). The constant flux. A study of class mobility in industrial societies. Clarendon. Esping-Andersen, G. (1990). The three worlds of welfare capitalism. Policy. Fenger, H. J. M. (2007). Welfare regimes in Central and Eastern Europe: Incorporating postcommunist countries in a welfare regime typology. Contemporary Issues and Ideas. Social Sciences, 3(2), 1–30. Ferrera, M. (1996). The southern model of welfare state in social Europe. Journal of European Social Policy, 6(1), 17–37. Glass, D. (1954). Social mobility in Britain. Routledge. Heath, A. (1981). Social mobility. Fontana. Heclo, H. (1985). The welfare state in hard times. APSA. Hega, G., & Hokenmaier, K. (2002). The welfare state and education: A comparison of social and educational policy in advanced industrial societies. German Policy Studies, 2(1), 143–173. Heidenheimer, A. J. (1981). Education and social security entitlements in Europe and America. In P. Flora & A. J. Heidenheimer (Eds.), The development of welfare states in Europe and America. Transaction. Hoffmayer-Zlotnik, J. H. P., & Warner, U. (2007). How to survey education for cross-national comparisons: The Hoffmeyer-Zlotnik/Warner-Matrix of education. Metodoloski Zvezki, 4(2), 117–148. OECD. (2010). A family affair: Intergenerational social mobility across OECD countries. In Economic policy reforms going for growth. OECD. Papanastasiou, S. (2018). Intergenerational social mobility and types of welfare state in Europe. Gutenberg. Papatheodorou, C., & Papanastasiou, S. (2016). Family origin and poverty in EU countries: The role of social protection systems. In M. Petmezidou & T. Kallinikaki (Eds.), Social research paths. Pattern. Papatheodorou, C., & Papanastasiou, S. (2011). Intergenerational mobility in the EU. INE-GSEE. Peter, T., Edgerton, J. D., & Roberts, L. W. (2010). Welfare regimes and educational inequality: A cross-national exploration. International Studies in Sociology of Education, 20(3), 241–264. Prais, S. (1955). Measuring social mobility. Journal of the Royal Statistical Society, Series A, 118, 56–66. Shorrocks. (1978). The measurement of social mobility. Econometrica, 46, 1013–1024. Stamatopoulou, G., Symeonaki, M., & Michalopoulou, C. (2013). Intergenerational transmission of education in Greece: Evidence from the European Social Survey 2002–2010, 15th conference of the Applied Stochastic Models and Data Analysis International Society (ASMDA), Barcelona, Spain, 25–28 June, 2013. Symeonaki, M., Filopoulou, O., & Stamatopoulou, G. (2011). Measuring intergenerational educational mobility in Greece, 14th conference of the Applied Stochastic Models and Data Analysis International Society (ASMDA), Rome, Italy, 7–10 June, 2011. Symeonaki, M., & Stamatopoulou, G. (2011). Exploring intergenerational educational mobility in Greece with data drawn from EU-SILC. In Proceedings of the University of Cyprus conference on social justice and participation: The role of higher education, Cyprus. Symeonaki, M., & Stamatopoulou, G. (2014a). Exploring the transition to Higher Education in Greece: Issues of intergenerational educational mobility. Policy Futures in Education, 12(5), 681–694. Symeonaki, M., & Stamatopoulou, G. (2014b). Intergenerational mobility as a distance measure between probability distribution functions. In C. H. Skiadas (Ed.), Theoretical and applied issues in statistics and demography.

372

M. Symeonaki and P. Tsinaslanidou

Symeonaki, M., Stamatopoulou, G., & Michalopoulou, C. (2012). Intergenerational Occupational Mobility in Greece: Evidence from EU-SILC. Demographic Analysis and Research International Conference, Chania, Greece, 5–8 June 2012. United Nations Educational, Scientific and Cultural Organization Institute for Statistics. International Standard Classification of Education ISCED 2011. UNESCO Institute for Statistics, Canada 2012. West, A., & Nikolai, R. (2013). Welfare regimes and education regimes: Equality of opportunity and expenditure in the EU (and US). Journal of Social Policy, 42(3), 469–493. Cambridge University Press. Woessmann, L. (2004). How equal are educational opportunities? Family Background and Student Achievement in Europe and the United States. CESifo (No. 1162), Munich.

Chapter 24

A Different Approach to Current Developments in the Twenty-First Century – Grouping European Countries in Terms of Mortality Panagiotis Andreopoulos, Fragkiskos G. Bersimis, and Alexandra Tragaki

24.1 Introduction This study attempts a different approach to geographical grouping in terms of the future mortality rate of 22 European countries. Grouping is conducted under each Beta Gompertz Generalized Makeham (BGGM) distribution parameter. Specifically, the values of the 4 parameters of the BGGM distribution are affected by: infant mortality (θ ), aging population – expressed as the number of older people in a country over 70 in the total population (ξ ), the random risk factor depends on the age (κ) and the random risk factor affecting the total (λ) of each country. The analysis period is expressed from 1960 to 2040 but focuses on the presentation at 12 different time points: every decade from 1960 to 2010 and every five years from 2015 to 2040. In particular, the time series are as follows: 1960–70–80–90–00– 10–15–20–25–30–35–2040. The goodness of fit of the Beta Gompertz Generalized Makeham distribution was evaluated in 2019 (Andreopoulos et al., 2019, 2020b) and showed satisfactory results as regards the appropriate statistical criteria in different data sets from different geographical areas of the European Union (Andreopoulos & Tragaki, 2020; Andreopoulos et al., 2019, 2020a). The process of spatial grouping is done through Pareto analysis (Powell & Sammut-Bonnici, 2015) and the Cluster analysis (Hartigan, 1975). Pareto analysis is a statistical technique that helps in making the right decisions. In this case, it concerns the understanding of human mortality and consequently combating

P. Andreopoulos () · A. Tragaki Department of Geography, Harokopio University, Athens, Greece e-mail: [email protected]; [email protected] F. G. Bersimis Department of Informatics and Telematics, Harokopio University, Tavros, Greece e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 C. H. Skiadas, C. Skiadas (eds.), Quantitative Methods in Demography, The Springer Series on Demographic Methods and Population Analysis 52, https://doi.org/10.1007/978-3-030-93005-9_24

373

374

P. Andreopoulos et al.

and (or) reducing the causes that aggravate it. Cluster analysis procedure aims to determine groups of cases, i.e. the clusters, with similar properties based on specific features.

24.2 Methodology In Pareto graph data are placed from left to right in descending order. The first group (A’ Group) consists of the countries to the left of the horizontal axis and display the highest values for each parameter studied, compared to the rest of Europe. The point of choice of the countries that will be the first group, is the point of the “upside down arm” in the non-smoothed data. Therefore, the values to the left of the first (or second) turning point are the countries with the highest values. While the second group (B’ Group), consists of the countries to the right of the horizontal axis and display the lowest values of the parameter under consideration. The data analyzed in this study have been normalized for the best possible “positive” visual effect, in order to avoid the high stochastic noise displayed by the mortality data, but with the result that the phenomenon of the “upside down arm” becomes more indistinguishable. The aforementioned gap is filled by using cluster analysis. Specifically, k-means method provided two groups that were common according each parameter with the corresponding results from Pareto Analysis. The √ transformation was therefore used e value (Campbell & Meyer, 2009) where it is a simple method of smoothing the estimates derived from the data. The smoothing did not affect the result of the order in the Pareto analysis, nor did it affect the grouping in the cluster analysis.

24.3 Results The results from both the Pareto analysis and the Cluster analysis are common in terms of the 21 countries’ binary classification in terms of Beta Gompertz Generalized Makeham distribution’s parameters. Starting with Pareto analysis for each parameter, some first similar conclusions appear between men and women, but mainly between countries, giving an additional feature to the phenomenon of mortality: spatial grouping or spatial correlation

24.3.1 Projection of the Parameter That Expresses Infant Mortality (θ) The A’ Group consists of the countries to the left of the horizontal axis (Fig. 24.1) and show the highest infant mortality rates compared to the rest of Europe.

24 A Different Approach to Current Developments in the Twenty-First. . .

375

Fig. 24.1 Pareto chart for men and women, showing the distribution of countries in descending order of frequency with parameter θ, with a cumulative line on a secondary axis as a percentage of the future certainty of each country, in all 22 countries

For men and women, the groups of countries with the highest infant mortality rates in all 22 geographical areas are Portugal, Ireland, Switzerland, Poland, Latvia, Estonia, Norway, the Netherlands, Iceland and the United Kingdom. The A’ Group in the case of men is completed by Austria and in the case of women by Belgium and the Czech Republic. While, the groups of countries with the lowest infant mortality, men and women are Greece, Finland, Lithuania, Denmark, Italy, France, Sweden, Spain and Germany. The B’ Group of men is completed by the Czech Republic and Belgium. While Austria completes (Fig. 24.1) the second group of women. The course that will follow infant mortality for both sexes in the high value of the A’ Group, is spatially common for 10 countries, with a rate of 45%. While the corresponding percentage for the B’ Group is 41%. In the case of parameter θ , countries’ grouping by using cluster analysis produces the same groups as Pareto analysis. By using distance (d), i.e. the Euclidean distance between each case (country) and its corresponding classification center, it is observed that countries belonging to the same cluster differ relatively as regards the parameter θ expressed by Infant Mortality. Such cases are Greece and Denmark for men, as well as Lithuania and France for women in the case of the low Infant Mortality group (Fig. 24.2). In addition, Latvia and Portugal for men and Belgium and the UK for women differ relatively in the countries’ high Infant Mortality group (Fig. 24.2).

376

P. Andreopoulos et al.

Fig. 24.2 *Distance’s bar chart in ascending order for each country in each cluster separately for men and women, as regards parameter θ *Distance indicates the Euclidean distance between each case (country) and its corresponding classification center

24.3.2 Projection of the Parameter That Expressed as the Aging Rate (ξ ) The values of parameter ξ show a similar picture for both men and women in all 22 countries. The differences between the countries are small and essentially their only categorization results from the extraction of the numerical conclusions. In Fig. 24.3, from the Pareto analysis, the first group includes 14 countries with the highest aging rates for both sexes. It seems that these countries will have high rates of population aging. The second group includes 2 countries with the lowest parameter value. It is worth noting for women that 20 of the 22 countries are included in the A’ Group, reinforcing the international literature (Christensen et al., 2009) which states the comparative advantage of women in longevity over men, with the parallel interest in its development aging population remains equally high (Carinci et al., 2020). Also, the imprint of the parameter ξ , in women appears with a higher average value compared to men, something that was expected according to the data so far. The course that the parameter ξ will follow for both sexes, is spatially common for 14 countries in the high value of the first group. For both sexes, the groups of countries with the highest aging rates in all 22 geographical areas are France, Switzerland, Austria, Finland, Belgium, Portugal, Spain, Sweden, Denmark, Italy, Netherlands, Norway, United Kingdom and Iceland. In women, Czech Republic, Estonia, Ireland, Lithuania and Poland complete the first group (Fig. 24.3). In the second group, the countries with the lowest aging rates for both men and women are Greece and Germany, without this meaning that they have a young population. On the contrary, the relative position of these countries in relation to other countries is marginally better. The B’ Group of men is completed by Ireland, Czech Republic, Poland, Lithuania, Estonia and Latvia (Fig. 24.3). In the case of parameter ξ , countries’ grouping by using cluster analysis produces the same groups as Pareto analysis again. By using distance (d), it is observed that

24 A Different Approach to Current Developments in the Twenty-First. . .

377

Fig. 24.3 Pareto chart for men and women, showing the distribution of countries in descending order of frequency with parameter ξ , with a cumulative line on a secondary axis as a percentage of the future certainty of each country, in all 22 countries

Fig. 24.4 *Distance’s bar chart in ascending order for each country in each cluster separately for men and women, as regards parameter ξ *Distance indicates the Euclidean distance between each case (country) and its corresponding classification center

countries belonging to the same cluster differ relatively as regards the parameter ξ expressed by aging rate for each country and its corresponding classification center for both clusters. Such cases are Greece and Germany for men, as well as Belgium and France for women in the case of the low aging rate group (Fig. 24.4). In addition, Spain and France for men differ relatively in the high aging rate group (Fig. 24.4).

378

P. Andreopoulos et al.

24.3.3 Projection of the Random Risk Factor Depending on Age (Parameter κ) The parameter κ, expresses the random risk factor as a function of age (Powell & Sammut-Bonnici, 2015). In Fig. 24.5, it is clear that the parameter is also influenced by gender, while from the Pareto analysis, in the A’ Group, 4 countries are entered with the highest value of the parameter concerning both sexes together. In contrast, in the second group, only one country is included for men and women with the lowest value of the parameter. Therefore, the groups of countries with the highest age-dependent randomized risk factor for both men and women in all 22 countries are the Czech Republic, Latvia, Estonia and Norway. The first group of women is made up of Switzerland, Poland, Denmark, Ireland, Lithuania, Iceland, Portugal, Belgium, Sweden, Finland, Austria, France, Spain, Italy, Switzerland and Greece (Fig. 24.5). The first group of men is limited to 4 countries. In the second group, only Germany is the group of countries with the lowest age-dependent risk factor for both men and women. The course that the parameter κ will follow in both sexes, concerns geographically 4 of the 22 countries in the high value of the first group.

Fig. 24.5 Pareto chart for men and women, showing the distribution of countries in descending order of frequency with parameter κ, with a cumulative line on a secondary axis as a percentage of the future certainty of each country, in all 22 countries

24 A Different Approach to Current Developments in the Twenty-First. . .

379

Fig. 24.6 *Distance’s bar chart in ascending order for each country in each cluster separately for men and women, as regards parameter κ*Distance indicates the Euclidean distance between each case (country) and its corresponding classification center

In the case of parameter κ, i.e. the random risk factor depending on age, the countries’ grouping by using cluster analysis produces the aforementioned groups produced from Pareto analysis. By using distance (d), it is observed once again that countries belonging to the same cluster differ relatively as regards the parameter κ expressed by the random risk factor depending on age. Such cases are Latvia and Norway for men, as well as Belgium and Greece for women in the case of the high random risk factor depending on age group (Fig. 24.6). In addition, Denmark and Switzerland differ relatively in the low random risk factor depending on age group, in the case of men (Fig. 24.6). Closing the comment on the future trend of parameter κ, a noteworthy element is the similarity of parameter values for women in the European Union. Finally, it seems that women on an individual level show more vulnerable behavior than men. In males, there does not appear to be a strong spatial dependence between populations on this parameter, although the Nordic countries are closer to the first group with the highest value.

24.3.4 Projection of the Random Risk Factor Affecting the Total Population (λ) The parameter λ, expresses the random risk factor, which is independent of the age of the individual and can potentially affect both sexes. However, there is evidence (Fig. 24.7) that in the end a “risk” does not affect men and women with the same intensity, nor does it affect countries between them. In Fig. 24.7, from the Pareto analysis, similar characteristics for both sexes appear in the first group of 9 countries with the highest parameter value (increased number of deaths, independent of age). While, in B’ Group are included 5 countries with the lowest price. The countries with the highest random risk factor, which is independent of the individual age and affects both sexes are Netherlands, Iceland, Estonia, Norway, Italy, Denmark, France, Finland and Switzerland. Group A’ is completed by Sweden

380

P. Andreopoulos et al.

Fig. 24.7 Pareto chart for men and women, showing the distribution of countries in descending order of frequency with parameter λ, with a cumulative line on a secondary axis as a percentage of the future certainty of each country, in all 22 countries. The maps show the average value of the parameter λ for men and women of each country, respectively

for men and Austria, Germany, Portugal, Greece, Spain and Ireland for women (Fig. 24.7). In the second group, the countries with the lowest random risk factor, which is independent of the individual age, for both men and women are Lithuania, Belgium, the Czech Republic, Latvia and the United Kingdom. Group B’ is completed by Austria, Portugal, Spain, Greece and Germany for men and Sweden and Poland for women (Fig. 24.7). The course that will follow the parameter λ for both men and women, is spatially common for the 9 countries in the high value of the first group and for 5 countries of the second group. The percentage of spatial grouping in both cases is very low, with estimates being very close to each other. The geographical areas that are most likely to be affected by external factors are Western, Central and Northern Europe (Spain, France, the United Kingdom, Ireland, Italy and Austria). In contrast, females also appear to be affected by future deaths, but less so in absolute numbers, with the whole of Europe being distributed almost evenly (Fig. 24.7). Generalizing about what was analyzed, in the case of both sexes, the countries that showed (Oksuzyan et al., 2008) and will show (Elliott & Wartenberg, 2004) in the near future show low infant mortality (θ ), respectively 70 years in most cases (ξ ). In contrast, individual and collective random risk factors do not go hand in hand in most countries. The subjective factor (κ) and the objective factor (λ) are two independent conditions that differentiate the cause although they cause the same

24 A Different Approach to Current Developments in the Twenty-First. . .

381

Fig. 24.8 Distance’s bar chart in ascending order for each country in each cluster separately for men and women, as regards parameter λ

result, the increase or decrease of deaths in a population. It is emphasized, however, that in the case of the random risk factor affecting the whole population, the spatial correlation between countries/regions cannot be considered negligible. Finally, in the case of parameter λ, i.e. random risk factor affecting the total population, the countries’ grouping by using cluster analysis is the same with the grouping produced from Pareto analysis. Distance (d) is applied for observing that countries belonging to the same cluster differ relatively as regards the parameter λ expressing a random risk factor affecting the total population. Such cases are Italy and Netherlands for men, as well as France and Greece for women in the case of the high random risk factor affecting the total population group (Fig. 24.8). In addition, Portugal and Germany differ relatively in the low random risk factor affecting the total population group, in the case of men, as well as, in the case of women, UK and Belgium differ relatively (Fig. 24.8).

24.4 Conclusions In the present study, the four spatial parameters of mortality related to the European scale (a total of 22 European countries) during the period 1960–2040 were identified and examined. Mortality has been modeled through the BGGM distribution, while the demographic and spatial evolution of each parameter is mapped (Andreopoulos et al., 2019). The years 2010–2017 were the time series of the “verification” of the forecast model with the real data. The approach was more than satisfactory for all countries, giving the corresponding dynamics to the Beta Gompertz Generalized Makeham distribution in terms of the reliability of the prediction of human mortality by 2040. Each of the parameters of the model has a specific demographic interpretation that allows exploration mortality from different perspectives: such as infant mortality (Rosano et al., 2000) (parameter θ ), the ratio of people over 70 years in the total population (parameter ξ ), as well as age-dependent (parameter κ) and nondependent random risk factor age (parameter λ). Two demographic characteristics, age and gender, were considered, as they can indicate the composition of the sexes

382

P. Andreopoulos et al.

and the age distribution of a particular population and thus determine its possible future prospects. Among the sexes, with higher parameter values θ than women, men appear to be more vulnerable to infant mortality. These findings are consistent with the corresponding European studies of 2008 (Drevenstedt et al., 2008). Among to them, «An increasing male disadvantage in infant mortality has occurred in both European and non-European countries since the first half of the twentieth century». This disadvantage is mainly due to the even greater vulnerability of men to prenatal mortality. In addition, according to a 2008 European study, women live longer than men in most European countries. The female advantage in longevity was observed mainly in the middle of the eighteenth century. During the twentieth century, this “gender gap” widened further due to economic growth and improved living conditions for women. This trend in the future seems to be limited in terms of numbers significantly, with a decrease in the corresponding averages. In terms of gender, however, men will likely continue to be more vulnerable than women. The higher longevity of women is confirmed by the higher values of the parameter ξ . Women have and will likely continue to have more years of life, resulting in a higher percentage of older women in European countries in the near future. Specifically, the countries on a European scale that show a strong upward trend in aging rapidly (Fig. 24.3) are Spain, Portugal, Greece, France and Italy, respectively. This demographic aging – common to all western countries – potentially causes many problems medical, social, family, financial, insurance, etc., which will take on explosive proportions in the coming decades if they do not receive the attention they deserve; from the respective government policies. By interpreting the time variations in the four parameters, it appears that the majority of the European countries examined show a trend of low (in the case of women) or decreasing (in the case of men) infant mortality. Demographically, this trend reflects the increase in the future life expectancy of man. The increase is also reinforced by the fact that, in the coming decades, deaths in Europe tend to accumulate at the age of over 75, resulting in an increase in the rate of aging, as already mentioned. Therefore, there are indications that an aging population (including men and women) will form in Europe. Combined with the declining trend in birth rates, this phenomenon could make a significant contribution to transforming the age composition of the European population in the coming years. Regarding the future sensitivity of individuals and the population to exogenous factors, the differences by gender are found in the overall trends over time. For the male population, the majority of countries reveal an overall increase in deaths by around 2030, when there is a growing trend in both types of vulnerability (individual and collective). On the other hand, for the female population, in most countries there is a declining trend in the vulnerability of individuals and a slightly upward trend in the vulnerability of the population. Regarding the spatial patterns identified in the four parameters, it could be said that they are significantly influenced by the geographical (Elliott & Wartenberg, 2004). Geographically defined countries such as the Nordic countries (Sweden,

24 A Different Approach to Current Developments in the Twenty-First. . .

383

Norway, Finland, Denmark and Iceland), the Baltic States (Estonia, Latvia and Lithuania), the Netherlands (Belgium), the countries of Central Europe (France, Austria), the countries of South Europe (Greece, Italy, Spain and Portugal), as well as Poland-Czech Republic and the United Kingdom-Ireland, have shown more or less similar trends in the past. This fact seems to continue in a general context. For example, with some exceptions from Finland, the Nordic countries will generally have a low infant mortality rate, a high aging rate and at the same time a low random individual and collective risk factor reflecting the very good standard of living provided to them. Despite being next to the Nordic countries, the Baltic countries will fluctuate sharply in most cases. Perhaps the sensitivity of the parameters in the small populations of these three countries will play a decisive role. They will show younger populations (lower values for the aging rate) than the rest of the countries, which will be characteristic of these countries. From 2020 onwards, there is an ever-increasing male mortality due to unexpected events for the countries of Central Europe and especially for the countries of the Mediterranean. This increase may be due to the socio-economic effects caused by the economic crisis of this period (2010–2018), but also the phenomenon of the coronavirus pandemic (2019– 2021) that is already affecting these areas. In addition, a generally increasing trend characterizes southern countries in terms of the age-related random factor. The composition of their populations from many different nationalities probably plays an important role in this. In addition, the interpretations of the parameter κ are mainly due to age, but the findings showed that each sex responds differently to risk. For each individual, the κ is expected to be higher during adolescence and early adulthood as well as in ages 50–75 years. A higher percentage of men than women are documented to be associated with deaths caused by accidents, suicides and exposure to many “dangerous externalities” in European countries (Wasserman et al., 2005). This rate was found to be proportional to the rate of male population deaths caused by natural disasters and other harmful socioeconomic conditions (Lenart et al., 2019). In addition, since the estimation of the parameter is based on individual data/personal data (which are not always available from the official statistical authorities), the conclusions drawn from this parameter are essentially simple indications. While as it is understood, the separation of the “positive” and “negative” outcome of a situation for each parameter is marginal. An additional element regarding the vulnerability to exogenous factors, concerns the differences from the “nature” of the vulnerability (individual or collective) that are identified mainly over time with age. The most common explanations for these findings are considered to be biological risks, risks acquired through social roles, lifestyle and behavior, and perhaps different access to treatments and health care (Oksuzyan et al., 2008). Surveys of 2008 highlight an ever-increasing mortality in the male population due to unexpected events for the countries of Central Europe, and especially for the countries of the Mediterranean. This increase is likely to record the effects of the 2008 global financial crisis on mortality. An economic crisis is associated with job displacement and rising unemployment rates, which tend to worsen living conditions. It is also associated with mortality, due to specific causes or affecting

384

P. Andreopoulos et al.

specific groups of the population (Ballester et al., 2019). For example, a 2016 study (Laliotis et al., 2016) reported that suicide deaths increased in Mediterranean countries during the 2008 financial crisis, with mortality rates being the strongest among working-age men (15–64 years old). Also, the change in the composition of their populations of different nationalities probably played an important role in it. The Netherlands, with a focus on Greece and Italy, has been proportionally absorbing the largest percentages of immigrants in Europe for decades (Lafaut et al., 2019). Making the distinction between men and women, the forecast findings for the female population in the years 2020 to 2040 can be traced back more safely and with a higher approximation to reality than to men mainly due to the low stochastic noise they display. In fact, gender differences in mortality or health are considered complex, as they depend on the biological, social and economic context of each population at the same time. Especially when inequalities in employment, access to health care and similar family responsibilities have reduced the racial “gap” between men and women in recent decades. Finally, like this study, research efforts investigating the dynamics of mortality throughout human life and its evolution over time can be of significant use. In particular, understanding the time fluctuations of mortality in different geographical environments (Zafeiris, 2019) can significantly contribute to identifying the mechanisms that determine aging and mortality, as well as to the design and implementation of socio-economic strategies aimed at improving the quality of life, life expectancy and reduction of premature mortality. In addition, predictions about population demographics are made possible. Based on these forecasts, governments around the world can better manage their retirement commitments, properly allocate their health-focused budgets, and ensure effective social policies.

References Andreopoulos, P., & Tragaki, A. (2020). On demographic approach of the BGGM distribution parameters on Italy and Sweden. In Demography of population health, aging and health expenditures (pp. 169–185). Springer. Andreopoulos, P., Bersimis, G. F., Tragaki, A., & Rovolis, A. (2019). Mortality modeling using probability distributions. Application in Greek mortality data. Communications in StatisticsTheory and Methods, 48(1), 127–140. Andreopoulos, P., Polykretis, C., & Tragaki, A. (2020a). Assessment and mapping of spatiotemporal variations in human mortality-related parameters at European scale. ISPRS International Journal of Geo-Information, 9(9), 547. Andreopoulos, P., Tragaki, A., Antonopoulos, G., & Bersimis, F. G. (2020b). Properties and dynamics of the Beta Gompertz Generalized Makeham distribution. In Demography of population health, aging and health expenditures (pp. 275–288). Springer. Ballester, J., Robine, J. M., Herrmann, F. R., & Rodó, X. (2019). Effect of the great recession on regional mortality trends in Europe. Nature Communications, 10(1), 1–9. Campbell, S. L., & Meyer, C. D. (2009). Generalized inverses of linear transformations. Society for industrial and applied Mathematics.

24 A Different Approach to Current Developments in the Twenty-First. . .

385

Carinci, F., Uccioli, L., Benedetti, M. M., & Klazinga, N. S. (2020). An in-depth assessment of diabetes-related lower extremity amputation rates 2000–2013 delivered by twenty-one countries for the data collection 2015 of the Organization for Economic Cooperation and Development (OECD). Acta diabetologica, 57(3), 347–357. Christensen, K., Doblhammer, G., Rau, R., & Vaupel, J. W. (2009). Ageing populations: The challenges ahead. The Lancet, 374(9696), 1196–1208. Drevenstedt, G. L., Crimmins, E. M., Vasunilashorn, S., & Finch, C. E. (2008). The rise and fall of excess male infant mortality. Proceedings of the National Academy of Sciences, 105(13), 5016–5021. Elliott, P., & Wartenberg, D. (2004). Spatial epidemiology: Current approaches and future challenges. Environmental Health Perspectives, 112(9), 998–1006. Hartigan, J. A. (1975). Clustering algorithms. Wiley. Lafaut, D., Vandenheede, H., Surkyn, J., & Coene, G. (2019). Counting the non-existing: Causes of death of undocumented migrants in Brussels-Capital Region (Belgium), 2005–2010. Archives of Public Health, 77(1), 42. Laliotis, I., Ioannidis, J. P., & Stavropoulou, C. (2016). Total and cause-specific mortality before and after the onset of the Greek economic crisis: an interrupted time-series analysis. The Lancet Public Health, 1(2), e56–e65. Lenart, P., Kuruczova, D., Joshi, P. K., & Bienertová-Vašk˚u, J. (2019). Male mortality rates mirror mortality rates of older females. Scientific Reports, 9(1), 1–9. Oksuzyan, A., Juel, K., Vaupel, J. W., & Christensen, K. (2008). Men: good health and high mortality. Sex differences in health and aging. Aging Clinical and Experimental Research, 20(2), 91–102. Powell, T., & Sammut-Bonnici, T. (2015). Pareto analysis. Wiley Encyclopedia of Management, 1–2. Rosano, A., Botto, L. D., Botting, B., & Mastroiacovo, P. (2000). Infant mortality and congenital anomalies from 1950 to 1994: An international perspective. Journal of Epidemiology & Community Health, 54(9), 660–666. Wasserman, D., Cheng, Q. I., & Jiang, G. X. (2005). Global suicide rates among young people aged 15–19. World psychiatry, 4(2), 114. Zafeiris, K. N. (2019). Mortality differentials among the euro-zone countries: An analysis based on the most recent available data. Communications in Statistics: Case Studies, Data Analysis and Applications, 5(1), 59–73.

Part V

Various Applications

Chapter 25

Examining Items’ Suitability as the Marker Indicator in Testing Measurement Invariance Anastasia Charalampi, Catherine Michalopoulou, and Clive Richardson

25.1 Introduction A prerequisite for meaningful comparisons of constructs across different demographic and social groups, within and across nations is the establishment of their measurement invariance or equivalence (Davidov, 2008; Missine et al., 2014; Davidov et al., 2014; Raudenská, 2020). Measurement invariance ensures that a measurement instrument, e.g. an attitude scale, “measures the same concept in the same way across various subgroups of respondents” (Davidov et al., 2014: 9; see also, Davidov 2008) or across repeated measurements, time points and social categories (Putnick & Bornstein, 2016; Davidov et al., 2014). Without measurement invariance, conclusions based on comparisons of different groups or across time would be ambiguously interpreted (Cheung & Rensvold, 1999; Steenkamp & Baumgartner, 1998; Xu & Tracey, 2017). Although the importance of testing for measurement invariance was pointed out in the literature more than 50 years ago (Putnick & Bornstein, 2016), appropriate statistical methods were developed under a structural equation modeling framework relatively recently (e.g. Cheung & Rensvold, 1999; Steenkamp & Baumgartner, 1998; Vandenberg, 2002; Vandenberg & Lance, 2000). One of the most popular methods for testing measurement invariance is multiple-groups Confirmatory factor analysis (MGCFA; Brown 2015; Cheung & Rensvold, 1999; Han et al., 2019; Jung

A. Charalampi () · C. Michalopoulou Department of Social Policy, Panteion University of Social and Political Sciences, Athens, Greece e-mail: [email protected]; [email protected] C. Richardson Department of Economic and Regional Development, Panteion University of Social and Political Sciences, Athens, Greece e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 C. H. Skiadas, C. Skiadas (eds.), Quantitative Methods in Demography, The Springer Series on Demographic Methods and Population Analysis 52, https://doi.org/10.1007/978-3-030-93005-9_25

389

390

A. Charalampi et al.

& Yoon, 2017). As is the case in any structural equation model, researchers need to “assign a unit of measurement to the latent factor so that the latent factors will have a scale and so that the model is identified” (Han et al., 2019: 1488; see also, Cheung & Rensvold, 1999; Charalampi et al., 2020c). The most commonly used approach is to select an item to serve as a marker (or referent) indicator (reference variable) for each factor and fix its factor loading at one (Brown, 2015; Cheung & Lau, 2012; Han et al., 2019; Millsap & Yun-Tein, 2004). In the literature, little consideration is given to this decision (Jung & Yoon, 2017), which is but rarely reported, or an item is chosen automatically by software defaults (Brown, 2015; Bowen & Masa, 2015). However, in many cases, the marker indicator may influence the interpretation of the model (Jung & Yoon, 2017). Choosing the marker indicator is considered to be complicated especially when there are many items (Cheung & Rensvold, 1999; Cheung & Lau, 2012; Han et al., 2019; Jung & Yoon, 2017). Steenkamp and Baumgartner (1998) suggested that the item selected to serve as marker indicator should demonstrate metric invariance. If a noninvariant item is selected, then the parameter estimates could be distorted leading to inaccurate conclusions (Vandenberg, 2002; French & Finch, 2008). However, without testing for measurement invariance there is no other way of knowing beforehand which items are invariant (Cheung & Lau, 2012). In this respect, Liu et al. (2017: 18 proposed that a “marker variable should have a meaningful metric, or be an indicator of the latent common factor with a high factor loading. For evaluating longitudinal measurement invariance, it is crucial to choose a marker variable whose loading is invariant at all occasions. The model identification strategy requires that two of the thresholds for the marker variable be constrained to be invariant across measurement occasions. Therefore, the marker variable should not only have an invariant factor loading across all measurement occasions, but also have at least two invariant thresholds.” In the cases where there is no obvious item with a meaningful metric, Brown (2015) proposed performing MGCFAs using different items as marker indicator (see also, Han et al., 2019). In this paper, we empirically explore the suitability of items to serve as marker indicator by performing MGCFAs using a different item each time as suggested by Brown (2015). The investigation is based on an eleven-item unidimensional scale measuring emotional wellbeing from the European Social Survey of 2006 and 2012. Measurement invariance is tested for gender and employment status groups of a combined sample of eight European countries.

25.1.1 Emotional Wellbeing Scale As mentioned in our previous work (Charalampi, 2018; Charalampi et al., 2019, 2020a, b), the personal and social wellbeing module was first included in the questionnaire of Round 3 (2006) of the ESS and was repeated with certain changes in Round 6 (2012) of the survey (European Social Survey, 2015; Jeffrey et al., 2015). Combining theoretical models and evidence from statistical analyses, six

25 Examining Items’ Suitability as the Marker Indicator in Testing. . .

391

key dimensions were defined for the 2012 ESS measurement of personal and social wellbeing as follows (European Social Survey, 2015); Jeffrey et al., 2015): evaluative wellbeing, emotional wellbeing, functioning, vitality, community and supportive relationships. Four variables from emotional wellbeing, three items from vitality and one from the supportive relationships dimensions comprise the eight-item version of the Center for Epidemiologic Studies Depression scale (CES-D 8), which was proposed by Van de Velde et al. (2009, 2010). CES-D 8 is a shortened version of the original CES-D that was developed by Radloff (1977), designed to measure symptomatology of depression in the general population and non-clinical settings (Andresen et al., 1994; Carleton et al., 2013; Cole et al., 2000; Karim et al., 2015). In addition to the CES-D 8 scale, the ESS contains three additional items that can be considered as another set of similar questions based on personal feelings (Raudenská, 2020). In sum, there are eleven common items included in both Rounds of the ESS (Raudenská, 2020). This eleven-item scale measures emotional wellbeing as it consists of seven items measuring the existence of negative emotions and four measuring the existence of positive emotions.

25.2 Method 25.2.1 Participants The analysis was based on the European Social Survey Round 3 Data (2006) and the European Social Survey Round 6 Data (2012) for the following eight countries: Belgium, France, Germany, Netherlands, Poland, Portugal, Russian Federation and Spain. These countries were selected from the 29 participants in Round 6 because they had also participated in Round 3 (2006), when the wellbeing module including the items measuring emotional wellbeing was first introduced into the questionnaire. The ESS implements all the strict methodological prerequisites for comparability over time and cross-nationally (Kish, 1994; Carey, 2000) by applying probability sampling, minimum effective achieved sample sizes in all participating countries and a maximum target non-response rate of 30% (The ESS Sampling Expert Panel, 2016). Face-to-face interviewing is used for data collection. The survey population is defined as all persons aged 15 and over residing within private households in each country, regardless of their nationality, citizenship or language; this definition applies to all rounds of the survey. The samples of the eight countries included more women than men with the exception of Germany in Round 6. The mean age ranged from 44 to 52 years in every country. More than 42.9% of the participants were married, the majority had completed at most secondary education with the exception of Russian Federation in Round 6 and at least 39.1% were in paid work.

392

A. Charalampi et al.

25.2.2 Measures The eleven items used for the emotional wellbeing scale were worded as follows (European Social Survey, 2015): felt depressed: how often during the past week (E1); felt that everything I did was an effort: how often during the past week (E2); Sleep was restless: how often during the past week (E3); was happy: how often during the past week (E4); felt lonely: how often during the past week (E5); enjoyed life: how often during the past week (E6); felt sad: how often during the past week (E7); could not get going: how often during the past week (E8); had a lot of energy: how often during the past week (E9); felt anxious: how often during the past week (E10); felt calm and peaceful: how often during the past week (E11). The scale is comprised of seven negatively (E1–E3, E5, E7–E8 and E10) and four positively (E4, E6, E9 and E11) worded items. The scoring of negatively worded items was reversed before the analysis in order to achieve correspondence between the ordering of the response categories following the original grouping of items into the wellbeing dimensions based on Jeffrey et al. (2015). The response categories range from 1 to 4 and are defined as follows: none or almost none of the time (1); some of the time (2); most of the time (3); and all or almost all of the time (4). Therefore, the items’ level of measurement is ordinal, i.e. categorical.

25.2.3 Statistical Analysis MGCFA was carried out using Mplus Version 8.4 for testing measurement invariance of gender (men and women) and employment status (employed and unemployed) groups: configural, metric and scalar (Brown, 2015; Millsap & Yun-Tein, 2004). In performing MGCFA, the following sequence of decisions was adopted as presented in our previous work (Charalampi et al., 2020c): 1. Initially, separate Confirmatory Factor Analyses (CFAs) were performed for both demographic and social groups (Brown, 2015) following the sequence of decisions presented in our previous work (Charalampi, 2018; Charalampi et al. 2019; 2020a, b). As Brown (2015: 246) pointed out, “if markedly disparate measurement models are obtained between groups, this outcome will contraindicate further invariance evaluation”. Model fit was considered adequate if χ 2 /df < 3, CFI and TLI values were greater than or close to .95 and RMSEA ≤ .06 with the 90% CI upper limit ≤ .06, or acceptable if χ 2 /df < 3, the CFI and TLI values were >.90 and RMSEA