+ All Categories
Transcript

A Multilevel Profile Evaluating Social Inequality

by

Irina Campbell, PhD, MPH

US Dept. of State Fulbright Scholar in Global Health

Kazakhstan, 2006-2007

email: [email protected]

© 2000, 2002, 2007

Irina Campbell, PhD, MPHColumbia University

All Rights Reserved

2

SUMMARY

This report describes a multilevel profile of self-reported physical health in

Moscow, examining individual and contextual determinants. A random citywide sample

of Moscow adults with household telephones (N=2000) was collected Sept. 17-19,

1991, with a completed interview rate of 81.8%.

Central research goals were to estimate 1.) individual level determinants,

defined as life chances, health choices, and measures of social relations, including

social cohesion and capital, 2.) contextual level determinants, defined as average and

relative inequality; and 3.) to specify how both average and relative inequality were two

complementary aspects of contextual level social status which had direct and mediating

effects on individual health.

The components of physical health were influenced by different sets of

individual determinants. Choosing one specific or several outcomes has significant

implications for public health interventions.

Hierarchical effects of social status on health were found in logistic and

hierarchical regressions, independently of health choices. These effects were moderated

by urban inequality. On the whole, living in high inequality urban areas was associated

with a greater risk for poor physical health, independently of the health choices made by

area residents.

Greater social connectivity was expected to have a positive effect on physical

health; and a buffering effect on low status residents living in high inequality areas.

This effect was observed in the reverse direction, suggesting that the construct validity

of measures of social capital defined for market economies may be less applicable in

3

Moscow.

The importance of assessing the individual in a social context was shown in

comparing logistic with multilevel models. The multilevel model permitted identifying

contextual effects of relative inequality as interactions between macro and micro level

determinants.

The health crisis in Russia cannot be reduced to individual health choices such

as drinking or smoking behaviors. This research demonstrated that individual health is

an integral part of a social matrix. The place one lives may foster health, independently

of individual health choices, through the quality of a civic society and social relations,

or else impede it in an environment of social inequality and compromised psychosocial

well-being.

4

TABLE OF CONTENTS

LIST OF FIGURES .............................................................................................................8

LIST OF TABLES .............................................................................................................10

ACKNOWLEDGEMENTS.............................ERROR! BOOKMARK NOT DEFINED.

PREFACE ..........................................................ERROR! BOOKMARK NOT DEFINED.

CHAPTER 1: INTRODUCTION ....................................................................................16

STUDY GOALS .......................................................ERROR! BOOKMARK NOT DEFINED.

GENERAL TERMS:........................................................................................................21

INDIVIDUAL AND MULTILEVEL MODELS............................................................23

PROPERTIES OF SOCIAL STRUCTURE....................................................................27

INEQUALITY AS AN EMERGENT PROPERTY OF SOCIAL STRUCTURE:.........29

INEQUALITY AS MODERATING CIVIC COMMUNITY AND HEALTH ..............34

INEQUALITY AS MODERATING LIFE CHANCES AND HEALTH .......................36

URBAN HEALTH PROFILES AS POLICY .................................................................36

CHAPTER 2: CONSTRUCTS OF HRQOL DETERMINANTS ...............................40

MACRO-MICRO PROBLEM ........................................................................................41

THE CIVIC COMMUNITY .........................................................................................48

LIFE CHANCES ..........................................................................................................52

HEALTH CHOICES ....................................................................................................54

MULTILEVEL MODEL.................................................................................................56

SUMMARY OF RESEARCH HYPOTHESES..............................................................59

CHAPTER 3: CONSTRUCTS OF HRQOL OUTCOMES........................................64

GLOBAL AND DOMAIN SPECIFIC INDICATORS ..................................................69

STRUCTURE OF QOL...................................................................................................71

HAPPINESS AND SATISFACTION ............................Error! Bookmark not defined.

SELF-RATED HEALTH ..............................................................................................71

PHYSICAL HEALTH...................................................................................................72

5

CHAPTER 4: HEALTH POLICY IN RUSSIA............... ERROR! BOOKMARK NOT

DEFINED.

HEALTH POLICY AND INEQUALITY................ERROR! BOOKMARK NOT DEFINED.

CHAPTER 5: HEALTH INEQUALITIES IN SOCIAL CONTEXT ........................75

INTERNATIONAL CONTEXT OF HEALTH INEQUALITY....................................77

NATIONAL CONTEXT - RUSSIA................................................................................83

MORTALITY ................................................................................................................86

POVERTY.....................................................................................................................90

MORBIDITY.................................................................................................................93

INCREASING INEQUALITY ......................................................................................94

URBAN CONTEXT........................................................................................................95

MOSCOW.....................................................................................................................97

AREA VARIATION ....................................................................................................104

CONCLUSION..............................................................................................................115

CHAPTER 6: METHODS..............................................................................................116

DATA COLLECTION ..................................................................................................119

CHARACTERISTICS OF THE SAMPLE POPULATION .......................................119

SURVEY INSTRUMENT............................................................................................125

MEASURES OF HRQOL OUTCOMES ...................................................................125

MEASURES OF HRQOL DETERMINANTS............................................................129

MODEL ASSUMPTIONS ............................................................................................139

LOGISTIC REGRESSIONS.......................................................................................141

HIERARCHICAL LINEAR REGRESSION ...............................................................143

SUMMARY ...................................................................................................................147

CHAPTER 7: RESULTS OF AN INDIVIDUAL MODEL OF HEALTH .............149

SELF-RATED HEALTH ..............................................................................................154

HAPPINESS AND SATISFACTION...........................................................................159

EFFECTS OF INEQUALITY ON QOL ....................................................................164

PHYSICAL HEALTH PROFILE..................................................................................171

6

DIMENSIONS OF PHYSICAL HEALTH .................................................................177

DISABILITY AND CHRONIC CONDITIONS..........................................................177

ACUTE SYMPTOMS .................................................................................................183

ENERGY LEVELS......................................................................................................187

EFFECT OF INEQUALITY ON PHYSICAL HEALTH ...........................................190

DISCUSSION ................................................................................................................191

EFFECTS OF HEALTH CHOICES..........................................................................192

EFFECTS OF CIVIC COMMUNITY........................................................................195

EFFECTS OF LIFE CHANCES................................................................................196

CHAPTER 8: RESULTS OF A MULTILEVEL MODEL OF HEALTH...............199

SPECIFICATION OF THE MULTILEVEL MODEL.................................................204

RANDOM COEFFICIENT MODEL............................................................................205

RESULTS FOR CONTEXTUAL EFFECTS..............................................................207

RESULTS FOR FIXED EFFECTS............................................................................208

SUMMARY.................................................................................................................212

THE INTERCEPT-AND-SLOPES-AS-OUTCOMES MODEL.................................213

RESULTS FOR FIXED AND INTERACTIVE EFFECTS........................................215

RANDOM EFFECTS AND SUMMARY ...................................................................223

CHAPTER 9: CONCLUSIONS.....................................................................................232

SUMMARY OF HYPOTHESES..................................................................................232

MAJOR FINDINGS ......................................................................................................233

SUMMARY OF CONCLUSIONS ...............................................................................235

CONTEXTUAL EFFECTS OF INEQUALITY ON QOL..........................................235

CONTEXTUAL EFFECT OF INEQUALITY ON PHYSICAL HEALTH.................237

MULTILEVEL INTERACTIONS..............................................................................238

IMPLICATIONS ...........................................................................................................244

STUDY LIMITATIONS ...............................................................................................247

FUTURE RESEARCH ..................................................................................................249

APPENDICES ..................................................................................................................250

7

APPENDIX 1: TOTAL PERMANENT MOSCOW POPULATION, 1989, UNWEIGHTED AND

WEIGHTED SAMPLING DISTRIBUTIONS, MOSCOW, 1991, BY SEX AND URBAN AREA .....250

APPENDIX 2: RESIDUAL ANALYSIS OF PHYSICAL HEALTH PROFILE.................. ERROR!

BOOKMARK NOT DEFINED.

APPENDIX 3: AVPLOTS OF PHYSICAL HEALTH PROFILE WITH FULL MODEL OF SOCIAL

DETERMINANTS ..........................................................ERROR! BOOKMARK NOT DEFINED.

APPENDIX 4: ORDERED LOGIT REGRESSION OF PHYSICAL HEALTH PROFILE BY LIFE

CHANCES, HEALTH CHOICES, AND CIVIC COMMUNITY .................................................251

APPENDIX 5: TYPE OF INEQUALITY IN URBAN AREA BY PHYSICAL HEALTH PROFILE

AND QOL: LOW-HIGH DEVELOPMENT OF NEW URBAN RESOURCES BY LOW-HIGH ACCESS

TO URBAN RESOURCES IN RESIDENCE AREAS OF MOSCOW, 1991;(N=1629) PERCENT ...253

APPENDIX 6: KISH TABLES.........................................................................................254

APPENDIX 7: OVERALL DESIGN EFFECT .....................................................................254

INTERVIEWER EFFECT..........................................................................................256

APPENDIX 8: RUSSIAN LANGUAGE QUESTIONNAIRE ..................................................267

APPENDIX 9: RUSSIAN LANGUAGE RESPONSE CODESHEET ...ERROR! BOOKMARK NOT

DEFINED.

APPENDIX 10: ALAMEDA PHYSICAL HEALTH PROFILE..............................................268

APPENDIX 11: MEASURES OF HRQOL (SELF-RATED HEALTH, LIFE SATISFACTION ,

LIFE HAPPINESS) AND LIFE CHOICES, LIFE CHANCES, AND CIVIC COMMUNITY ..........271

APPENDIX 12: CODEBOOK.....................................ERROR! BOOKMARK NOT DEFINED.

REFERENCES.................................................................................................................279

8

LIST OF FIGURES

FIGURE 1: CONCEPTUAL MULTILEVEL MODEL OF URBAN HRQOL ..................................17

FIGURE 2: MACRO AND MICRO PROPOSITIONS ACCOUNTING FOR GEOGRAPHIC VARIATION

IN HEALTH .....................................................................................................................33

FIGURE 3: THE STRUCTURE OF A MICRO, MACRO, AND MULTILEVEL MACRO-MICRO

PROPOSITION .................................................................................................................57

FIGURE 4: A STRUCTURE OF SELECTED MULTILEVEL MACRO – MICRO PROPOSITIONS.....59

FIGURE 5: ADMINISTRATIVE UNITS (89) OF THE RUSSIAN FEDERATION, 1993 ...... ERROR!

BOOKMARK NOT DEFINED.

FIGURE 6: CHANGES IN MORTALITY RATES, 1970-1990/1993, SELECTED WEST AND EAST

EUROPEAN COUNTRIES .................................................................................................76

FIGURE 7: AGE-STANDARDIZED DEATH RATES BY SELECTED CAUSES IN 1993, EUROPEAN

STANDARD POPULATION; ALBANIA 1992 DATA, USA 1991 DATA.................... ERROR!

BOOKMARK NOT DEFINED.

FIGURE 8: TRENDS IN MALE LIFE EXPECTANCY AT BIRTH IN RUSSIA, FORMER SOVIET

REPUBLICS, CENTRAL AND EASTERN EUROPE (CCEE), 1970– 1994 ..........................80

FIGURE 9: MALE LIFE EXPECTANCY IN ESTABLISHED MARKET ECONOMIES (EME) AND

FORMER SOCIALIST ECONOMIES (FSE) BY PER CAPITA GDP OF COUNTRY, 1993.......82

FIGURE 10: FEMALE LIFE EXPECTANCY IN ESTABLISHED MARKET ECONOMIES (EME)

AND FORMER SOCIALIST ECONOMIES (FSE) BY GDP OF COUNTRY, 1993 ..................82

FIGURE 11: NUMBER OF BIRTHS, DEATHS, AND NATURAL CHANGE IN POPULATION SIZE

(PER 1000 PERSONS) BY DISTRICT, MOSCOW, 1995...................................................100

FIGURE 12: INFANT MORTALITY BY URBAN AREA, MOSCOW, 1995................................103

FIGURE 13: EXAMPLES OF MICRO-RAYION PLANNING FOR MOSCOW IN 1950S AND 1960S

....................................................................................................................................105

FIGURE 14: CITY ADMINISTRATIVE DISTRICT (33), MOSCOW, 1989 ................................106

FIGURE 15: LIVING SPACE IN SQUARE METERS PER PERSON BY DISTRICT, MOSCOW, 1989

....................................................................................................................................106

FIGURE 16: SOCIAL AREAS, MOSCOW, 1960 ....................................................................108

9

FIGURE 17: GROUPING OF MOSCOW CITY DISTRICTS BY RESIDENCE PREFERENCES AND

SOCIAL VALUATION FACTORS, 1984. ..........................................................................110

FIGURE 18: PERSONS WITH HIGHER EDUCATION (%) BY DISTRICT, MOSCOW, 1989 .......111

FIGURE 19: PENSIONERS (%) BY DISTRICT, MOSCOW, 1989 ...........................................111

FIGURE 20: CANCER MORTALITY (PER 100,000 PERSONS), MOSCOW, 1989...................114

FIGURE 21: INFECTIOUS AND PARASITIC DISEASES (PER 100,000 PERSONS), MOSCOW,

1992 ............................................................................................................................114

FIGURE 22: STANDARDIZED FACTOR SCORES OF ACCESS TO SOCIAL RESOURCES IN URBAN

AREAS BY NEW DEVELOPMENT IN URBAN AREAS, MOSCOW, SEPTEMBER, 1991 .......135

FIGURE 23: PROBABILITY OF QOL BY PHYSICAL HEALTH ...............................................152

FIGURE 24: MEAN ALCOHOL CONSUMPTION AND MEAN PHYSICAL HEALTH BY

EDUCATIONAL LEVEL, MOSCOW, 1991.......................................................................226

FIGURE 25: MEAN ALCOHOL CONSUMPTION AND MEAN ACCESS TO SOCIAL RESOURCES BY

EDUCATIONAL LEVEL, MOSCOW, 1991.......................................................................228

FIGURE 26: INTERVIEWER WORKLOAD, MOSCOW HRQOL SURVEY, SEPTEMBER, 1991264

10

LIST OF TABLESTABLE 1: SUMMARY OF THE CONCEPTS, INDICATORS, AND MEASURES OF INEQUALITY,

MOSCOW HEALTH PROFILE ..........................................................................................23

TABLE 2: HRQOL COMPONENTS OF THE MOSCOW CITY HEALTH PROFILE.....................70

TABLE 3 SELECTED CAUSES OF DEATH CONTRIBUTING TO CHANGES IN LIFE EXPECTANCY,

RUSSIA, 1987-1994, BY SEX, IN YEARS; ......................................................................87

TABLE 4: PERCENT OF SELF-RATED HEALTH AND LIFE EXPECTANCY, BY COUNTRY, 1990

......................................................................................................................................90

TABLE 5: SOCIOECONOMIC INDICATORS IN RUSSIA, MOSCOW, ST. PETERSBURG, 1990-

1995..............................................................................................................................96

TABLE 6: SOCIAL INDICATORS IN RUSSIA, MOSCOW, ST. PETERSBURG, 1990-1995.......98

TABLE 7: SELECTED CAUSES OF MORBIDITY, RUSSIA AND THE CITY OF MOSCOW, 1988-

1995............................................................................................................................101

TABLE 8: CAUSE-SPECIFIC DEATH RATES IN RUSSIA, MOSCOW, ST. PETERSBURG, 1989-

1997............................................................................................................................102

TABLE 9: COMPARISON OF CITY OF MOSCOW CENSUS, 1989, WITH WEIGHTED SAMPLE,

1991, BY MARITAL STATUS, SEX, AGE, AND EDUCATIONAL LEVEL. ...........................120

TABLE 10: PERMANENT* MOSCOW POPULATION, 1989, UNWEIGHTED AND POST-

STRATIFICATION WEIGHTED** SAMPLE DISTRIBUTIONS, BY AGE AND SEX, MOSCOW,

1991 ............................................................................................................................122

TABLE 11: ESTIMATED NUMBER AND PER CAPITA TELEPHONES BY SELECTED

ADMINISTRATIVE DISTRICTS OF MOSCOW, JAN., 1992 ..............................................124

TABLE 12: ALAMEDA COUNTY PHYSICAL HEALTH PROFILE COMPONENTS IN 7

CATEGORIES BY PERCENT WEIGHT OF CATEGORY IN OVERALL PROFILE ....................127

TABLE 13: FREQUENCIES OF MICRO MEASURES OF HRQOL BY GENDER, MOSCOW, 1991;

WEIGHTED SAMPLE (PERCENT)....................................................................................128

TABLE 14: FREQUENCIES OF MICRO MEASURES OF LIFE CHOICES AND LIFE CHANCES BY

GENDER, MOSCOW, 1991; WEIGHTED SAMPLE (PERCENT) .........................................131

TABLE 15: DISTRIBUTION (PERCENT) OF INDIVIDUAL MEASURES OF CIVIC COMMUNITY BY

GENDER, MOSCOW, 1991; WEIGHTED SAMPLE ...........................................................133

11

TABLE 16: AVERAGE INEQUALITY INDICATORS (STANDARDIZED VARIABLES/1000 AREA

POPULATION); ROTATED FACTOR LOADINGS (VARIMAX) BY 33 ADMIN. DISTRICTS,

MOSCOW, 1989...........................................................................................................136

TABLE 17: DESCRIPTION OF MACRO MEASURES OF LIFE CHANCES, BY 33 ADMINISTRATIVE

AREAS, MOSCOW CENSUS, 1989 .................................................................................137

TABLE 18: ANOVA FOR PHYSICAL HEALTH.......................................................................147

TABLE 19: SPEARMAN CORRELATION OF HRQOL OUTCOMES ......................................150

TABLE 20 : LOGISTIC REGRESSION OF SELF-RATED HEALTH BY QOL, CIVIC COMMUNITY,

AND HEALTH CHOICES, ALL MOSCOW AREAS (N=1629; LOGIT B, ODDS RATIO E^B)..........156

TABLE 21: LOGISTIC REGRESSION OF LIFE HAPPINESS BY QOL, CIVIC COMMUNITY, AND

HEALTH CHOICES, ALL MOSCOW AREAS (N=1629; LOGIT B, ODDS RATIO E^B)..................158

TABLE 22 : LOGISTIC REGRESSION OF LIFE SATISFACTION BY QOL, CIVIC COMMUNITY,

AND HEALTH CHOICES, ALL MOSCOW AREAS (N=1629; LOGIT B, ODDS RATIO E^B)..........161

TABLE 23: LOGISTIC REGRESSION OF JOB SATISFACTION BY QOL, CIVIC COMMUNITY,

AND HEALTH CHOICES, IN ALL MOSCOW AREAS (N=1629; LOGIT B, ODDS RATIO E^B)......163

TABLE 24: LOGISTIC REGRESSION OF HEALTH, HAPPINESS, AND SATISFACTION BY

DEMOGRAPHICS AND TYPE OF AVERAGE INEQUALITY IN URBAN AREA (LOW ACCESS TO

RESOURCES; HIGH DEVELOPMENT OF NEW RESOURCES), ODDS RATIOS (|Z-STATISTIC|)

....................................................................................................................................166

TABLE 25: LOGISTIC REGRESSION OF HEALTH, HAPPINESS, AND SATISFACTION BY CIVIC

COMMUNITY, HEALTH CHOICES AND TYPE OF AVERAGE INEQUALITY IN URBAN AREA ,

(LOW ACCESS TO RESOURCES; HIGH DEVELOPMENT OF NEW RESOURCES), ODDS

RATIOS (|Z-STATISTIC|)...............................................................................................168

TABLE 26: INTERACTIONS IN LOGISTIC REGRESSION OF HEALTH, HAPPINESS, AND

SATISFACTION BY TYPE OF AVERAGE INEQUALITY IN URBAN AREA, (LOW ACCESS TO

RESOURCES; HIGH DEVELOPMENT OF NEW RESOURCES), ODDS RATIOS (|Z-STATISTIC|)

....................................................................................................................................169

TABLE 27: ORDERED LOGIT REGRESSION OF ALAMEDA PHYSICAL HEALTH PROFILE BY

TYPE OF AVERAGE INEQUALITY IN URBAN AREA (LOW ACCESS TO RESOURCES; HIGH

DEVELOPMENT OF NEW RESOURCES) AND LIFE CHANCES, HEALTH CHOICES AND CIVIC

COMMUNITY (LOGIT B, ODDS RATIO E^B) .........................................................................172

12

TABLE 28: ORDERED LOGIT REGRESSION OF DISABILITY BY TYPE OF AVERAGE

INEQUALITY IN URBAN AREA (LOW ACCESS TO RESOURCES; HIGH DEVELOPMENT OF

NEW RESOURCES) AND LIFE CHANCES, HEALTH CHOICES AND CIVIC COMMUNITY (LOGIT

B, ODDS RATIO E^B) .........................................................................................................176

TABLE 29: ORDERED LOGIT REGRESSION OF CHRONIC CONDITIONS BY TYPE OF AVERAGE

INEQUALITY IN URBAN AREA (LOW ACCESS TO RESOURCES; HIGH DEVELOPMENT OF

NEW RESOURCES) AND LIFE CHANCES, HEALTH CHOICES AND CIVIC COMMUNITY (LOGIT

B, ODDS RATIO

E^B)____________________________________________________________________...........180

TABLE 30: ORDERED LOGIT REGRESSION OF ACUTE SYMPTOMS BY TYPE OF AVERAGE

INEQUALITY IN URBAN AREA (LOW ACCESS TO RESOURCES; HIGH DEVELOPMENT OF

NEW RESOURCES) AND LIFE CHANCES, HEALTH CHOICES AND CIVIC COMMUNITY (LOGIT

B, ODDS RATIO E^B) .........................................................................................................184

TABLE 31: ORDERED LOGIT REGRESSION OF LOW ENERGY LEVELS BY TYPE OF AVERAGE

INEQUALITY IN URBAN AREA (LOW ACCESS TO RESOURCES; HIGH DEVELOPMENT OF

NEW RESOURCES) AND LIFE CHANCES, HEALTH CHOICES AND CIVIC COMMUNITY (LOGIT

B, ODDS RATIO E^B) .........................................................................................................186

TABLE 32: ORDERED LOGIT REGRESSION OF HIGH ENERGY LEVELS BY TYPE OF AVERAGE

INEQUALITY IN URBAN AREA (LOW ACCESS TO RESOURCES; HIGH DEVELOPMENT OF

NEW RESOURCES) AND LIFE CHANCES, HEALTH CHOICES AND CIVIC COMMUNITY (LOGIT

B, ODDS RATIO E^B) .........................................................................................................189

TABLE 33: BIVARIATE CORRELATION MATRIX OF URBAN LEVEL VARIABLES, MOSCOW

CENSUS, 1989 (N=33) .................................................................................................203

TABLE 34: RANDOM COEFFICIENT MODELS OF PHYSICAL HEALTH................................211

TABLE 35: RANDOM COEFFICIENT MODELS, MEANS-AND-SLOPES-AS-OUTCOMES, URBAN

AREA INEQUALITY EFFECT ON MEAN PHYSICAL HEALTH.........................................214

TABLE 36: RANDOM COEFFICIENT MODEL, MEANS-AND-SLOPES-AS-OUTCOME, EFFECTS

OF URBAN INEQUALITY AND LIFE CHANCES ON AVERAGE PHYSICAL HEALTH* ......217

TABLE 37 : RANDOM COEFFICIENT MODEL, MEANS-AND-SLOPES-AS-OUTCOME, EFFECTS

OF URBAN INEQUALITY AND HEALTH CHOICES ON AVERAGE PHYSICAL HEALTH* .220

13

TABLE 38: RANDOM COEFFICIENT MODEL, MEANS-AND-SLOPES-AS-OUTCOME, EFFECTS

OF URBAN INEQUALITY AND CIVIC COMMUNITY ON AVERAGE PHYSICAL HEALTH*223

TABLE 39: INTERVIEWER EFFECTS FOR SELECT VARIABLES BY SEX, MOSCOW HRQOL

SURVEY, SEPTEMBER, 1991........................................................................................261

TABLE 40: INTERVIEWER EFFECTS FOR SELECT VARIABLES BY AGE, MOSCOW HRQOL

SURVEY, SEPTEMBER, 1991........................................................................................262

14

ACKNOWLEDGEMENTS

The Ford Foundation provided a grant for the preparation of parts of this

manuscript for publication as a Special Issue of Social Science and Medicine, v. 51,

November, 2000.

Thanks go to all those who took the time to discuss parts of this material which

were presented at:

The World Association for Public Opinion Research Annual Meeting,Copenhagen, 17 September 1993;

The International Conference on Survey Measurement and Process Qualityof the Royal Statistical Association and American Statistical Association, Bristol, UK,1-4 April 1995;

The 3rd International Congress of the International Council for GlobalHealth Progress, UNESCO Paris, France, 18-20 May 1998;

The 7th International Conference on Social Stress Research, Budapest,Hungary, 27-29 May 1998;

The World Health Organization International Healthy Cities Conference,Athens, Greece, 20-23 June 1998;

The 5th International Congress of Behavioral Medicine, Copenhagen,Denmark, 19-22 August 1998;

The 7th Biennial Conference of the European Society of Health and MedicalSociology, Rennes, France, 27-29 August 1998.

Portions were also published in:

Editor, McKeehan, I.V. "A Multilevel City Health Profile of Moscow", SocialScience and Medicine, n. 9, vol. 51 (November, 2000): 1295-1312.

McKeehan. I.V. ”Planning of National Primary Health Care and PreventionPrograms: The First health Insurance Law of Russia, 1991-1993”, in Gallagher, EugeneB. and Subedi, Janardan (Eds.). Global Perspectives on Health Care. Prentice Hall,Englewood Cliffs, New Jersey, 1995: 174-197;

McKeehan, I.V. "Measurement of Interviewer Error in a Moscow TelephoneSurvey", American Statistical Association: 1995 Proceedings of the International

15

Conference on Survey Measurement and Process Quality, Bristol, United Kingdom,April, 1995: 46-51.

McKeehan, I.V. “Health Status of the Moscow Elderly”. SotsiologicheskiyeIssledovaniye, n.3, (March, 1995): 109-114. (in Russian).

McKeehan, I.V. "Meta-Analysis of Soviet Survey Research Methods", AmericanStatistical Association:1993 Proceedings of the Section on Survey Research Methods,Vol. II, January, 1994: 1172-1177.

McKeehan, I. V., Campbell, R., and Tumanov, S.V. “Lifestyles and Habitsinfluencing the health of Muscovites before the implementation of the Health InsuranceLaw of Russia of 1991-1993”. Sotsiologicheskiy Issledovaniye, n.3, (March, 1993): 45-49 (in Russian).

16

CHAPTER 1: INTRODUCTION

After the presidential election of 2000 in Russia, where the mayor of Moscow

was a focal contender, the city has taken on an even more central role in defining the

flavor of democracy in Russia. This monograph reviews an investigation designed

during the dissolution of the USSR. It provides a public health picture of an urban

community for establishing a City Health Profile, four months before the

implementation of macro economic changes in January, 1992.

Central research goals of this multilevel health profile of Moscow (Figure 1)

were to: 1.) provide baseline data on the cross-sectional distribution of micro and

macro determinants of self-reported physical health; 2.) assess the effect, moderating,

additive or interactive, of social relations, in the form of social cohesion, formal and

informal networks on physical health outcomes; 3.) to examine the hypothesis that a

hierarchical effect on health of individual social status is significant, 4.) that the

distribution of contextual inequality is a determinant of individual health, in addition to

individual social status and health choices; 4.) establish the baseline for longitudinal

follow-up comparisons.

A random telephone sample of adults within 33 administrative districts of

Moscow was collected on Sept. 17-19, 1991, with a completed interview rate of 81.8%

(n=1629). The questionnaire replicated items from the California Alameda Study on

Health and Ways of Living, and the U.S. Health Interview Survey in a Russian

translation. Moscow State University sociologists organized instrument translation,

telephone sampling, and data collection.

The initial chapter introduces the theoretical issues of inequality as they relate to

17

a multilevel health profile and to the ongoing health crisis in Russia. The theoretical

components of the multilevel health profile, as outlined in Figure 1, are presented in

greater detail in chapters 2 and 3. Chapter 4 describes health inequalities in relation to

social context at the international, national, urban, and small area levels, and then tied to

the theoretical model in chapter 2. Chapter 5 describes the empirical methods and

measurement issues of the survey, including design error.

FIGURE 1: CONCEPTUAL MULTILEVEL MODEL OF URBAN HRQOLTYPE OF MEASURE

HEALTH OUTCOME HEALTH CHOICES LIFE CHANCES SOCIAL RELATIONSMICRO

Physical Healthdisabilitychronic conditionsacute symptomsenergy levels

MICRO

Personal healthpracticesbody mass indexsmokingalcohol intakephysical activity

Prevention

Utilization

Economic values

AGGREGATED MACRO

▪Area mean alcohol use

MICRO

EducationOccupationEmploymentHousing typeMobility in work, housing

GLOBAL MACRO

Average Inequalityarea mean factor score ofaccess to material resources

Poverty risksarea ratio Family 5 members

Relative Inequalityarea rate ratiosmall/average apt. sizearea rate ratio lo/hi educarea rate ratio bluecollar manual/ whitecollar nonmanual workers

MICRO

Social cohesionanomie

Social capital:Informal networksFormal groupsDensity ofnongovernmentalorganizations(NGOs)Voter turnout

AGGREGATED MACRO

▪Social CohesionArea mean anomie

Chapters 6 and 7 present the micro and macro level determinants of individual

health-related quality of life outcomes. Chapter 8 concludes this report with a

discussion of the implications of a multilevel profile of urban health and by suggesting

public health priorities which are derived from the Moscow health profile.

The objective of this research is to describe a cross-sectional multilevel health

18

profile of the city of Moscow, which was obtained before implementation of macro

economic changes of January, 1992, in a social epidemiological survey. The

development of a multilevel theory and model of health was undertaken in keeping with

the WHO Healthy City Program and policy for the twenty-first century of Health For

All: “by the year 2000, the actual difference in health status between . . . groups . . .

should be reduced . . . by improving the level of health of disadvantaged . . . groups”

(WHO, 1994). A city health profile provides a description of the social determinants of

health, including the distribution of social disadvantage or vulnerability.

A clearer understanding of the influence of social status differences in individual

level health, within the social context experienced by people, is useful for the

implementation of WHO policy. The city profile enables first, the identification of

health differences between individuals and between social groups in Moscow, secondly,

the recognition of specific structural conditions in the community which affect the

health of people living there, and thirdly, the separation of structural determinants of

health from the effects of individual psychosocial and health behaviors.

Social epidemiology has traditionally been concerned with the distribution of

morbidity or mortality in relation to a causal triad: group or personal characteristics,

geographic, social or community determinants, and change in occurrence over time:

who is exposed or susceptible to what, where, when, and why. These parameters are

included in the design of the health profile, which examines the social origins of health

status: the differential effect of community level social inequality, a characteristic of the

social context that was hypothesized to increase vulnerability for poor health, in

addition to the biopsychosocial risk factors of the individual. Health outcome was

19

defined in the Moscow survey by the Alameda County Physical Health Profile. A

prospective study design was based on the longitudinal Alameda County health studies.

This report describes the baseline data. Follow-up studies have been planned of the

Moscow community.

The social changes, associated with the August Coup of 1991 and subsequent

dissolution of the USSR, which ensued during the implementation of the survey, altered

the conditions of the research plan, delaying the follow-up of the 1991 Moscow cohort

currently underway. The 1991 survey is relevant not only as a necessary baseline for

subsequent longitudinal comparisons, but also as an independent urban sample in

addition to the only other two cross-sectional surveys of that pre-reform period

(Palosuo, Uutela, Zhuravleva & Lakomova, 1998; Lee, 1995).

The city health profile of Moscow gives a historical perspective on the relative

importance of community conditions that moderated the effect of individual risks on

health outcomes during a specific period of sociopolitical reorganization. This survey is

also unique as a sample which collected individual and ecological level data of

Moscow, providing valuable information not available elsewhere. There have not been

other multilevel community health studies done in Russia that can provide an insight

into the relationship between aggregate ecological (macro) factors and individual

psychosocial (micro) factors either recently or during the Soviet period. The impact of

the dynamic changes on the health of the sample population, which have occurred since

the survey was conducted, will be analyzed in a follow-up study.

A context of relative social inequality experienced by people living in different

communities in conjunction with poor health habits or lack of social connectivity can

20

have an additive effect on physical health. The effects of social context on individual

health may be mediated by intervening psychosocial processes that depend on the

characteristics of the social context (Hox, 1998). Some forms of social relations, such

as social cohesion, social support or social networks, can moderate the effects of

inequality on health. Two statistical models are employed to explore these research

questions: logistics regression at the individual level and multilevel or hierarchical

linear regression at both the individual and ecological levels.

Multiple indicators and multiple methodologies are used throughout the profile

to apply a comprehensive approach to urban health assessment. Methodological

pluralism increases certainty in the validity and reliability of conclusions which may be

drawn from a community sample survey.

Public health and personal health are two prongs of community health policy

based to varying degrees upon two fundamental issues: the right to health and the right

to happiness. Neither health nor happiness are principally based on privilege or earned

achievement but on justice and a moral imperative as part of the human condition.

Underlying the individual human rights issue is the general quality of life or material

standard of living in a community and the responsibility and cost to the state for

guaranteeing that sociopolitical conditions do not militate against citizens’ rights to

health and happiness. One part of this debate centers on the question of who is

responsible for personal and who for public health, given the mutual influences between

individuals and the social context in which they live. A second part of this debate

concerns the effect on health of the distribution of inequality in social and material

resources within a community.

21

GENERAL TERMS:Constructs, their operational indicators, and the measurement of the indicators

are distinguished in Table 1. Context may be understood in two principal ways:

statistically, as a measure of a macro level or socioenvironmental/ group effect upon the

micro level or individuals; and sociologically, as a measure of social structure, also a

macro level effect on individuals. In the interest of parsimony and pragmatism

important for the development of efficient health policy, social structure is defined as

the distributions of social groups, social relations, and socioeconomic positions.

Geographical place may be defined at any of various hierarchical levels: street

block, neighborhood, community, regional, national, or even by continent, which has a

resident population with social characteristics and relations. Although the parameters of

the geographical place may be more or less fixed, the demographic parameters are fluid,

depending on the extent of social change, mobility, or immigration and emigration in

the area. Social relations arise in a community from the interactions of residents from

various social groups and socioeconomic positions. Social groups are circumscribed by

gender, race, ethnicity, occupational status, or other homogeneous attributes such as

those related to employment, type of labor, property ownership, ascribed or achieved

status. Socioeconomic position is defined as the stratification of social groups on a

variety of hierarchical dimensions such as poverty, income, education, prestige, or

access to resources. Inequality refers to the position of groups or individuals on these

distributions in two ways: 1.) in relation to the mean of a specific distribution, and 2.)

relative to specific groups on a distribution.

Deprivation in health or happiness due to a relative position in the distribution of

social status or the standard of living in a community is termed here relative inequality.

22

Having marginal access to resources in contrast to mean access, for example, is related

to the position of individuals in the social structure of a community relative to other

groups and to the form which the distribution of resources itself takes. Such relative

deprivation may be due to the inequity of unnecessary, preventable, “unjust”

circumstances which some people face and others do not.

Deprivation may also be due to average inequality where the material standard

of living in a community, like income or other access to resources, refers to the overall

mean availability of resources to all residents. While relative inequality affects specific

groups and individuals differentially in relation to others in a hierarchical distribution of

social status, average inequality measures effect on everyone in a community at the

same level.

Poverty is a lack of resources below an average reference point. Poverty is used

as a proxy for an income covariate in this study. It is not synonymous with average or

relative inequality, but a measure of absolute economic status.

The average and relative distributions of standards of living within and between

communities changes. The relative position of individuals within communities is altered

due to changes in the macro distribution across communities, independently of

individual action. The size of the wealth and health gaps between nations affects the

gaps between individuals relative to each other, and to mean levels within communities.

Where people live and how they fit into their communities, as well as where their

communities stand in the larger global or sociogeographic arena is significant for

individual health. Although health is ultimately experienced at the most personal of

levels, the context of communities and whole societies may be decisive for the quality

23

of that experience. This issue is important when planning effective public health

programs focusing on the community or individual level.

TABLE 1: Summary of the concepts, indicators, and measures of inequality, MoscowHealth Profile

constructs Relative Inequality Average Inequality

theoretical concepts deprivation due to position insocioeconomic distribution disadvantage among specificsocial status groups greater impact of disadvantageon some groups than others lack of individual control overpreventable, avoidable, inequitableconditions

deprivation due to economicand material resources disadvantageous conditionsaffect everyone in group orcommunity at mean level some individual control unavoidable conditions

operational indicators relative position social status comparative difference betweenscores shifting reference point

average position access to material resources absolute score fixed reference point

measurements rate ratio of low/high socialstatus composition in urban area

mean access to materialresources available in urban area

relation betweenconcept and measure

indirect total impact of low social statuson outcome + effect of the range inthe distribution of social statusamong groups of different sizes macro

direct impact of average conditions macro

INDIVIDUAL AND MULTILEVEL MODELSPublic Health has traditionally been interdisciplinary. It combines the interests

of epidemiology, concerned with health assessments of populations, with social policy

or intervention to ensure the health of social groups, and sociology, as the study of

social processes within and between social groups, systems, structure and relations. The

biomedical model has contributed an overwhelming concern with the individual patient

and the curative, healing process. In addition, however, every individual lives in a

24

matrix of values, customs, traditions, language, mores, and norms of behavior, or

subjective and accepted ways of living, as well as a distinctive civic tradition and

political system, which comprise a collective context of lifestyle or culture.

Preventive health programs function within these contexts and, when

sufficiently funded, have the capability of monitoring and taking action to address the

differential gap in quality of life and health status among social groups and individuals.

Personal responsibility for health is often targeted with public educational programs,

while social responsibility is left more frequently to government legislation (for

example: professional certification, Food and Drug Administration labeling oversight,

regulation of chemicals and pharmaceuticals, occupational safety).

Strategies to maximize public health may include modifying those community

environments which have been shown to be effective in changing individual level

behavior, such as advertising for cigarettes or alcohol. The distributions of specific risk

factors which change at the community level need to be addressed within their social

context, insofar as risks have distinctive distributions in distinct communities.

The domain of preventive strategies in public health is the at-risk population as

a whole and that of medicine is the at-risk individual. Improvements in the quality of

life and public health depend in large part upon the development of a multilevel theory

of health which puts the individual back into the community context and provides an

assessment of micro and macro interactions of communicable, chronic and social

diseases. Personal risks to health under control by the individual and social risks to

health not directly controllable by the individual are situated in a larger community

context. To sift through such causal complexity requires maintaining local and national

25

data monitoring systems, which are multilevel, systematic and periodically recurring, to

assess the adequacy, equity, quality, accessibility, and outcome of health programs.

The human rights movement in health has raised the issue of inequality in health

status, which was tied to universal access and comprehensive medical and health

services. Universal access implies the availability of services to all individuals and

groups. Equality in access, measured as an input to the health system, however, does not

categorically imply equality in health status, as an outcome of the health system,

regardless whether measured by sociomedical indicators or mortality and morbidity

rates. Comprehensive treatment of disease also requires preventive services which

promote health through population-based programs such as education, nutrition,

sanitation, vaccinations, and occupational safety.

Multilevel models of health provide for disease patterns among individuals in

groups as a consequence of social relationships between groups and among individuals

within groups. Multilevel health profiles provide pictures of the interaction between

“sick populations” and “sick individuals”, rather than single macro or micro health

status. Multilevel models make a fundamental contribution to the research of disease

etiology and disease prevention conceptually, as well as methodologically.

Mortality data assume a biomedical model of disease causation: a cause, host,

agent, in an outcome chain. Mortality classifications characterize individuals rather than

populations and are thus prone to misattribute the multifactorial epidemiological

research triad (agency, host, environment). Considering fundamental causes of

mortality requires examining social structure, morbidity and psychosocial factors (Rose,

1995; Krieger, 1998), in other words, applying a sociomedical model of health.

26

In seeking to define a model of the complex causes of premature mortality and

increase in specific diseases in Russia, a distinction must be made between proximal

and fundamental factors, and the level of causal attribution should be sought in an

ecosocial context (Krieger, 1998). Social determinants may affect diverse health states,

like mortality or morbidity, directly and indirectly, through various mechanisms of

intervening, mediating, or contributing risk factors (Susser, 1976). Focusing on

immediate, proximal mechanisms may have the unintended consequence of minimizing

fundamental social causes and emphasizing personally controlled actions (compliance

with medical regimens, physical exercise, no drinking or smoking, healthy diet, etc.) to

the detriment of formulating cogent public policy.

The same social cause may operate simultaneously at the community and

individual level but through different paths and with multiple outcomes at each level, as

the following examples illustrate (Blaxter, 1990; Link, 1998; Matteson et al., 1998).

The lack of social support, as a type of relative deprivation, has been linked to increased

CHD (Siegrist, 1995). A loss of control in life and work conditions, or powerlessness,

has been associated with psychosocial symptoms, poor health, and mortality (Bobak,

1996; Seeman, 1995). Rapid social change, like the modernization within communities

which produces increased relative inequality, has been significantly related to heart

disease morbidity and mortality (Lasker et al., 1994). Job insecurity and organizational

change, during the modernization and specialization of a government

telecommunications agency, resulted in worsening self-rated health status, increases in

the body mass index, sleep problems, and clinical measures of hypertension (Ferrie, et

al., 1998). Policy strategies which inadvertently blame the victim or obscure the social

27

conditions which “put people at risk of risk” by emphasizing proximal causes cannot

reduce premature mortality due to increasing inequality within a community or living in

an “anti-modern, stressful” economy like Russia (Rose, #227, 1998).

PROPERTIES OF SOCIAL STRUCTUREMultilevel models of health analyze the emergent properties of social structure,

such as inequality, in conjunction with micro level properties, such as health status,

gender, ethnicity, educational level. Context or emergent properties of structure at each

level refer to those characteristics which exemplify aspects of the whole unit of

analysis and not the separate components of that unit (Blau, 1980). Contextual analysis

can explain the influences which the structure of a unit has within a hierarchy, and

multilevel analysis can focus on multiple hierarchies of units within the same model. It

is a prime objective of the sociological perspective of public health to assess the context

in which macro and micro units change, given the complex nesting of individuals in

groups, within neighborhoods, within cities or rural environments, and within societies

and nations.

Effective public health strategy aims at identifying groups at risk for poor health

and specific risk factors causing poor health. Emergent properties of social structure as

health determinants are often not considered. Health promotion policies involve

developing two intervention strategies by which health risks may be reduced for

vulnerable high risk groups as well as for the general population. High risk intervention

involves reducing rare high risks for a small number of individuals and population risk

intervention involves reducing pervasive low risks for large sectors of a population.

Application of both prevention strategies is necessary to avert the “prevention

paradox”: individual level intervention affects community health minimally, while

28

community level interventions have limited impact upon high risk individuals. The

“prevention paradox” is paralleled by the “risk intervention paradox”: the mass

exposure of a large number of individuals to low levels of negligible risks may produce

a larger number of disease cases than a small number of individuals exposed to high

risks. Treating only very sick individuals leaves scarcely sick populations without

health intervention. Mass prevention reduces negligible risk a little for many and not at

all for some, while most derive at least minimal benefits.

This phenomenon affects the development of public health policies for sick

populations rather than sick individuals (Rose, 1995). Often, barely sick populations are

left without intervention, although many would benefit even from minimal care, while

very sick individuals are vigorously sought out and treated, a cost borne by the

“healthier” community. Public health intervention differs from medical intervention

foremost in its emphasis on the socioenvironmental context of individual health status.

And secondly, public health recognizes that a continuous distribution in health status

characterizes populations, and is not restricted by the clinical designation of individuals

as potential treatment cases with or without disease.

A multilevel evaluation of health therefore gauges negligible and high risk

factors at both the population-level and individual-level. As abundant research has

shown, individual lifestyle behaviors, given specific level of socioeconomic

development, account for a majority of the risk factors for general well-being. A major

issue is to determine which risk factors are amenable to policy intervention and

manipulation. Mass levels of low exposure require mass levels of intervention even if

impact is negligible, since the community will benefit as a whole, and subsequently

29

individual members. The range of risk factors, which affect health, change as a whole

distribution at the community level. Public health policy therefore needs multilevel

research strategies which recognize the sui generis properties of communities, the

emergent properties of social structure, and their attendant risk factors, which cannot be

reduced to a collective aggregate of its individual members. Inequality may be

designated as a structural property of communities rather than individuals, and as such

is prior to individuals in a community context in which they must live.

INEQUALITY AS AN EMERGENT PROPERTY OF SOCIAL STRUCTURE:“Hippocrates, writing in the 5th century BC, advised anyone coming to a new

city to make inquiries in order to assess whether it was likely to be a healthy or an

unhealthy place to live, depending on its geography and water supply (‘soft, hard, or

salty’) and on the behavior of its inhabitants (‘whether they are fond of excessive

drinking and eating, and prone to indolence, or else fond of exercise and hard work’).

This notion, that healthiness is a characteristic of the population as a whole and not

simply of its individual members, lay dormant for a long time. It was revived and

developed by Durkheim, the great French sociologist of the last century…” (Rose,

1995). This spirit of Hippocrates which inquired about the socioenvironmental

conditions that contribute to individual and urban health is as germane today as it was

25 centuries ago.

Inequality produces structural obstacles to maximizing health – a proposition

which has preoccupied the public health research agenda for the past two decades in

Europe, especially after the seminal work of the Black Report established the strength

of this relationship. It was preceded by 140 years of socioeconomic variation in

mortality rates, and the current accumulation of evidence from comparative

30

international studies of almost every country in the world.

The Black Report (Macintyre, 1997) emphasized a typology of competing

theories of health inequality. Since the publication of the Black Report in 1980,

increased evidence about health inequalities has accumulated. Research has often been

organized around the typology of four broad explanations of the population

distributions of health inequalities: 1.) artefact of conceptual definition and

measurement error of health and social inequality; 2.) natural or social selection of the

healthiest by an occupational or social class structure; 3.) variation of material and

structural conditions of occupational positions with wealth and health; 4.) variation of

cultural and behavioral factors such as ideas, values, and individual lifestyle with

health. In each of these explanations, there is an interaction between individual health

and deprivation, mediated by a form of social hierarchy. Each one of these explanations

emphasizes how a different measurement or aspect of the individual – genetic or

biological, social, and personal – is related to social inequality to produce health

inequality.

There has been some debate on the lack of standard definitions and measurement

of health-related inequality as a risk factor or as an outcome, as a micro and macro level

indicator, or as a relative versus average indicator. Absolute standards of living as well as

income distributions have become conventional determinants of public health. Several

dominant definitions have referred to inequality as the relative distribution of an attribute,

such as social status or income, on a continuum from best to worst between individuals,

or in relation to a mean, median or ratio measure of the attribute (Murray et al., 1999;

Sen, 1997; Mackenbach and Kunst, 1997; Illsley and Baker, 1991; Frijters and van Praag,

31

1995). Inequality has been defined by a variety of measures (Pereira, 1990) such as the

distribution of occupation, salary, employment status, prestige, social group membership,

educational level, ownership of assets (housing, cars, etc.). Risk measures have been

distinguished by whether they are simple or complex; whether they can assess “relative”

or “absolute” differences in inequalities; or whether they can distinguish between the

effects of specific “levels” of high or low variation of inequality or the “total impact” of

socioeconomic inequalities upon the population distribution of health (Carr-Hill, 1990;

Mackenbach, 1997).

Health inequalities have been measured by a myriad number of population based

morbidity and mortality rates, prevalence rates, or individual subjective assessments of

self-rated health and quality of life indicators. Health inequalities vary consistently with

social inequalities regardless of type of measure, but the degree of inequality found is

affected by the specific indicator used to define either health or social inequality (Kunst

et al., 1998; Macintyre, 1997; Wilkinson, 1992). Inequality in health has been

successfully related to multiple dimensions of socioeconomic position: occupational

status and prestige, education, and income or access to resources (Siegrist, 1995). Each

dimension of social inequality may not only have a unique distribution in a community,

but be related to different sets of health determinants. The theoretical contribution of the

relative definition of social inequality addresses the structural issue that an individual has

a variety of social relations which are associated with a variety of social positions within

an array of social units (Blau, 1980).

An important consideration for assessing urban level health risks is the use of

area based indicators, for example - individual level (compositional aspect of area

32

where many of unemployed live, area unemployment is due to the individuals in the

area) or context-level (individual lives in an area with few opportunities for

employment thus is exposed to limitations on finding employment due to the area itself

since there are poor transportation conduits or no industries, no job skill retraining

facilities, etc.). Many individual level explanations have skirted the explicit

consideration of how the distribution of population characteristics may be affected by

the community as a whole.

The determinants of health variation between populations are of a different type

than the determinants of health variation between individuals within a population. For

example, life choices, such as exercise, nutrition, smoking or alcohol consumption,

made without hindrance, due either to self-control or to a social context affecting self

direction, need to be evaluated against structural obstacles, such as peer pressure,

advertising, marketing, or lack of adequate health information in a community.

The urban macro indicators of interest in this study are average and relative

inequality which affect individual health status. Analogous to the seminal Black

Report’s explanations of the distributions of health inequalities, there are two basic

propositions which have sought to explain the range, interaction, and direction of effect

between urban environment and individual health (Figure 2; Verheij, 1996): the

compositional hypothesis and the contextual hypothesis. In a compositional hypothesis,

direct and indirect selection can operate in the spatial variation of health-related

quality of life – either healthy people move or sick people stay in specific urban areas

(direct selection), or susceptible people with certain health traits move or stay in

specific areas (indirect selection).

33

FIGURE 2: MACRO AND MICRO PROPOSITIONS ACCOUNTING FOR GEOGRAPHIC VARIATIONIN HEALTH

Macro proposition:geographic variation due toContextual/social causationhypothesis:

Micro proposition:geographic variation due toCompositional/individual selectionhypothesis:

special variation in exposure toenvironmental/structuralfactors:inequality, poverty;pollution, traffic, housing;

quality, crime, recreational resources, sanitation, access to material or social resources

spacial variation in direct selection:sick/healthy people moving/staying in area:

poor people living in rundown areas; downward SES drift/mobility of sick concentration of sick around facilities concentration of healthy around parks, or “younger” areas

spacial variation in exposure tobehavioral factors:

drug/alcohol abuse, stress passive smoking, unsafe driving community group activities religious group membership

spacial variation in indirect selection:susceptible people with certain traitsmoving/staying in area –

large, younger, low-income families blue collar manual workers older persons w/low educational level

If composition of areas is the cause of health outcomes, then the variation in

health between urban areas is minimized when all individual variation in health-related

factors is taken into account, because the urban variation is composed from the specific

individuals in those areas. On the other hand, the contextual hypothesis accounts for the

spatial variation in health through exposure to socioenvironmental or structural factors..

The macro context exerts an independent and direct effect on individual health

outcomes. The contextual proposition is incorporated into the design of the multilevel

health profile.

Multilevel sociomedical theories permit the modeling of causes between social

structure and individual agency, permitting a broader perspective than linear social

selection (poor health causes low achieved social status) or downward drift (poor health

34

causes loss of social status) theories of poor health. The multilevel model takes into

account the interaction between area context, area composition, and interaction between

social context and individual behaviors. It is possible to examine propositions that

relative position on hierarchies of socioeconomic or social status has greater influence

on health than status itself by including macro-micro interactions in the modeling of

health determinants.

INEQUALITY AS MODERATING SOCIAL RELATIONS AND HEALTHStructural problems in public health often remain unsolved because they are

addressed at the individual level. Multilevel factors are not sufficiently specified, resulting

in medicalizing or “blaming the victim”. Promoting a “just say no to drugs” program to

adolescents, for example, but allowing advertising of cigarettes and alcohol which

legitimizes “permissible” drugs, assumes the individual is not embedded in a social

context. Such programs are ineffective and simply protect vested political and economic

interests. At a minimum, policy solutions for structural public health problems need to

outline the distal, mediating, and proximal influences on health at both the community

and individual level. Inequality in the social position of people in a community may act as

both a distal and mediating factor on individual health status. The mediating influence can

exist as both an intervening factor between macro and proximal determinants and a

moderating factor of proximal determinants on health outcomes.

A civic community, often understood broadly as social connectivity, has been

associated with sustaining individual well-being or physical health through social

integration of public and private spheres. It has been hypothesized that interpersonal,

informal-horizontal networks and formal-vertical networks “characterize how individuals

and groups are ‘connected’ to institutional, legal, political and economic structures”,

35

which is necessary for sustaining economic development (Rose, #278, 1998; Kaplan et al.,

1997; Kawachi et al., 1996; Lynch, 2000). The civic community plays an important part in

the pathway between the vertical and horizontal networks, affecting the biopsychosocial

level of health outcome. Despite the lack of consensus on which factors of a civic

community can be distinguished at the macro and micro levels as social cohesion, social

capital, social networks, social support, social integration, or social connectivity, the

literature clearly indicates a causal association with health status (Berkman and Kawachi,

2000).

The lack of formal networks, such as participation in religious and community

groups, the lack of informal networks, such as close friends and family, and the lack of

social cohesion have been associated with greater mortality from cardiovascular diseases

(Bruhn, 1979), declines in life expectancy (House et al., 1988), increases in homicides,

the infant mortality rate (Kawachi et al., 1997), and crime (Wilkinson et al., 1998).

Relative social inequality and surrounding poverty increase social disorganization through

lack of public trust and deterioration of social norms. Civic group memberships have been

related to social integration through the sustenance of social trust and maintenance of

norms, and consequently moderated health-related quality of life (Kaplan et al., 1996;

Kawachi and Kennedy, 1997a, 1997b; Berkman and Kawachi, 2000).

The various indicators of a civic community vary with sociopolitical and

geographic conditions, especially in a large and culturally diverse nation like Russia. The

effect of this diversity on outcome is lost if analyzed as a single sample at the individual

or ecological level. Multilevel models provide greater insight into how civic communities

moderate health given variations in the structure of communities.

36

INEQUALITY AS MODERATING LIFE CHANCES AND HEALTHA context of relative social inequality experienced by people living in different

communities in conjunction with poor habits or social connectivity can have an additive

effect on physical health. The effects of social context on individual health may be

mediated by intervening psychosocial processes which depend on the characteristics of

the social context (Hox, 1998). Social inequality can affect health by moderating some

forms of a civic community, such as lack of social cohesion, social support or social

networks, or some forms of social status.

Educational and occupational hierarchies have been related to health in what has

become known as the gradient effect of social status: higher status ensures better health.

Social hierarchies can define some types of relative inequality as a distribution of position

rather than resources. However, the interaction between education and occupation with

contextual inequality may have a direct effect on health apart from the main effects of

social hierarchies or inequality, whether in position or resources.

URBAN HEALTH PROFILES AS POLICY

The development of a multilevel health profile is a tool: an initial policy step,

which describes individual and community health status; facilitates problem definition;

outlines a strategy for public health programs. The importance for health policy of

addressing inequality not only involves the determination of priorities in problem

definition, program development, and local community action initiatives but also the

issue of ensuring that a social safety net includes health policy and functions properly to

protect basic human rights. Health policy based upon a city health profile which

describes the inequalities in community health status already provides many avenues of

political action for public health promotion and disease prevention.

37

A health profile which looks at inequality in health looks at solutions as well.

There is a need for community health profiles, especially during socioeconomic

transition periods in order to monitor the relationship between health status and social

policy, without “blaming the victim”. The reason for developing a Moscow city health

profile is to ground community health assessments in a policy framework. Moscow, as

the most economically advanced city within Russia, is central in defining the direction

of major policies and social change within Russia as a whole.

The relationship between health-related quality of life and inequality within a

community has concerned the World Health Organization since its inception in 1948,

continuing into the organization of the Healthy Cities Project. The Healthy Cities

approach has tied community assessment to local policy initiatives, such as specific

targets to achieve “Health for All”. WHO targets the complex urban problems of

inequality in health. The WHO strategy of “Health for All”, a policy adopted by the

European Region in 1984, has monitored progress by periodic evaluations, such as the

global assessment in 1990/1991, and delineation of principles of equity in health. The

WHO Healthy Cities Project has been instrumental in encouraging the organization of

ongoing comparative city-wide health information systems which specifically address

profiling inequity in health for the development of policy measures.

A health profile for Moscow lays the foundation for the development of a

multilevel Healthy City Profile for the World Health Organization Healthy Cities

Program. This focus on municipalitites was adopted in 1980 by the 32 member states of

the WHO European Region (including Russia). The primary objective was to reduce

inequality in health: “by the year 2000, the actual differences in health status between

38

countries and between groups within countries should be reduced by at least 25%, by

improving the level of health of disadvantaged nations and groups”(WHO, 1985).

The WHO-Regional Office for Europe has promoted a set of research guidelines

for the assessment of health, including measures of social inequalities applicable for

studying variation between international urban communities. Health policy can further

or hinder health promotion to the extent that such indicators are addressed. Economic

research of the health-inequality problem has contributed several innovative approaches

to conceptualizing the normative aspects of inequality as equity, the relative inequality

in distributional relationships. Distribution rules which have now become established in

the normative economics perspective on inequality (Pereira, 1982) include entitlement,

the decent minimum, utilitarianism, envy-free allocation, and more recently, “equity as

choice” and “health maximization” (Le Grand, 1987). However, normative economics

has still not been widely used in public health.

Economic measurements of inequality have most commonly steered away from

macro occupational or class definitions (Kunst et al., 1998), and suggested that

individual micro indicators of the distribution of wealth are more relevant than social

group measures. However, several applications of the more widely used macro

measures like the Gini coefficient, the Atkinson, as well as Robin Hood index, have

since been successfully applied to the study of inequality in health (Culyer, 1989;

Lambert, 1989; Mooney, 1987; Musgrove, 1986).

Measurement of inequalities in city-wide conditions which can be directly

addressed by policy are described by the Healthy Cities Project. These included such

urban indicators as types of municipal support for self-help organizations; health

39

education programs; percentage of six-year-old children fully immunized; population to

general practitioner ratio; population to nurse ratio; percentage of population with

health insurance protection; population proportion with access to emergency medical

services within 30 minutes by car; communication of health information.

The distribution of a set of environmental indicators, also dependent on policy,

was considered instrumental by WHO in assessing the ecological influences on urban

health: air and water pollution; household waste collection and treatment quality index;

public accessibility and relative area of green spaces in the city; industrial sites; sport

and leisure facilities; bicycling paths in the city; public transportation; crowding in

living space; emergency services provision. Socioeconomic indicators included square

meter of living space per resident in each city district; substandard dwellings;

unemployment rate; proportion of single parent families; illiteracy rate; crime rate;

percentage of city budget allocated to health and social programs.

Multilevel health profiles may include these macro measures of inequality and

contribute as well to the development of an ecosocial approach to community health

policy. Individually-based concepts of lifestyle as causes of disease may be

reformulated with contextual concepts of inequality as causes of disease due to the

distributions of socioeconomic resources within a community (Wilkinson, 1998).

40

CHAPTER 2: CONSTRUCTS OF HRQOL DETERMINANTS

Explanation of a social science phenomenon consists chiefly in identifying by

definitional specification the necessary and sufficient conditions for an event to occur.

The identification of specific conditions and their interactions within and between

individual and social levels is basic to establishing the theoretical utility of the

multilevel model as an explanatory construct of inequality, and as a predictor of health-

related quality of life. Historically, there have been several main perspectives of

lifestyle, such as the 19th and early 20th century traditions of sociology and

epidemiology, both of which examined individual action in a social context, but

methodologically did not have the appropriate tools, often emphasizing one to the

exclusion of the other.

Early methodological development of the linear statistical model circumscribed

the possibility of adequately assessing the weight of each factor, individual and social,

simultaneously within the same model. Phenomenological, empiricist, and positivist

approaches, with their concomitant methods, provided the dominant paradigms within

the framework of early 20th century social science and public policy, as did advances in

the biological and medical sciences. Theoretical perspectives argued within distinctive

qualitative and quantitative methodologies which formed disciplinary distinctions: the

humanities and cultural sciences; the social and psychological sciences; the natural and

physical sciences; as well as interdisciplinary biomedical and health sciences (Blalock,

1982).

On one hand, qualitative methods circumscribed the populations to which

41

models were applicable; examining individual, social, or historical levels separately;

and thus restricting the generality and universality of understanding individual action to

given specific temporal, cultural, and spatial conditions. On the other hand, the

limitations of a quantitative positivist, linear methodology could not address the

syncretic, phenomenological preservation and integration of the individual experience

as distinct from the social experience, a problem which gave rise to the micro-macro

dichotomy in sociological research and the atomistic-ecological fallacy in statistical

methods (Bottomore, 1975; Lilienfeld, 1976).

MACRO-MICRO PROBLEMBriefly, the micro-macro dichotomy was reflected in Durkheim’s rules of

sociological method which emphasized a realist and wholistic social order, where the

social fact acts as an external constraint, prior to individual action, is internalized by the

moral, socialized individual, and exists (such as a group average) in its own right

independent of individual manifestations (Durkheim, 1966).

Weber’s methodology of the social sciences insisted upon viewing the

individual within a partial social order, as an abstraction rather than a reality. The

verstehende analysis of social structure referred to a.) the subjective meaning of

individual action or b.) average meaning of group action oriented towards others, or c.)

to an ideal type of subjective meaning attributed to a hypothetical individual or group,

and d.) to the opportunities afforded to certain types of individuals to attain specific

positions in the social hierarchy through the operation of various objective and

subjective selective factors, but e.) not to a total social system (Weber, 1949).

Although it would be a gross oversimplification to reduce Durkheim’s position

as one concerned predominantly with the whole, while Weber concentrated on the parts,

42

the issue of how to relate social wholes, social parts or individual actors, as distinct

elements, occupied much of their and others’ sociological discourse of that industrial

age. However, these broad perspectives have exerted an important influence upon the

historical development of the lifestyle construct as currently used in modern social

research. Lifestyle was addressed by Weber directly as a characteristic of status

relationships, combining economic position with cultural expressions of consumption.

Economic classes and social status groups were stratified by Weber according to

production, acquisition, and consumption of goods. Durkheim, in contrast, emphasized

the concept of lifestyle as institutionalized social norms internalized by the individual

(Lynch and Kaplan, 2000).

Weber’s concept of lifestyle was composed of two aspects: life chances or social

opportunities at the macro level; and life choices or individual decisions at the micro

level. Briefly, Weberian lifestyles were collective phenomena associated with status

groups, patterns of consumptions, and individual self-directing choices made within the

constraints of the social opportunity structure. Differential health patterns may be

primarily explained by individual decisions to act in a specific manner which are

influenced by status group norms, the availability of socioeconomic options, and basic

physiological status. Given the historical period, other elements of the social

opportunity structure, such as sociopolitical values and ideology, civic culture, race,

gender, ethnicity, age, were only subsumed and not directly addressed by Weber

(Weber, 1978; Cockerham, 1997, 1993; Abel, 1993).

The important influence of macro level normative regulation of the social

opportunity structure and social cohesion, although not expanded upon by Weber, were

43

explicitly proposed earlier by Durkheim. Durkheim’s theory of normative control of

individual conduct through social institutions emphasized the coherent, stable,

predictable pattern of social norms. Social institutions exerted influence through a

common value system and a body of rules governing means and ends of individual and

group action, as well as an array of social opportunities or life chances. Durkheim

argued for the necessity of social restriction to unlimited individual desire for well-

being, the lack of individual self-control was replaced by external social control,

through the operation of a social conscience, internalized as individual morality. The

moral authority of society, in the name of the common interest, stipulates not only the

aspirations but the rewards offered and expected for every class of individual action

(Bellah, 1973; Durkheim, 1961).

There is a concomitant loss of individual control over achieving personal goals

with prescribed institutional means during a period of social deregulation or

destabilization. In periods of social crises, whether increasing or decreasing inequality,

for example, individuals become dislocated with respect to changing institutional values

and rules regulating the relationship to individual conduct. A loss of control over

individual effort or means, and lack of clarity about ends or attainment of goals, has

been associated with poor health status and to an increase in cardiovascular disease

(Bobak, 1999; Siegrist, 1996; Marmot, 1998).

The breakdown of socially normative regulation of individual action affects life

choices by often placing the individual in a social quandary of not being able act in

traditional and expected ways. New values and norms often arise in situations of social

crisis but the time lag before individuals assimilate the social changes into their

44

personal lives continues as a period of psychosocial disorganization. Command over

resources and attainment of goals are essential aspects of individual happiness which

are disrupted by massive social change or crises, resulting in health crises as well

(Veenhoven, 1996).

Durkheim points to a stratification of well-being, aspirations, and a way of

living considered acceptable or equitable by public opinion, a social mechanism of

regulation which determines the maximum and minimum degree of ease of living

legitimate for each social class or group. Public opinion, the social conscience, is not

immutable but subject to the vagaries of time and space, with relative standards of

luxury and poverty, health and disease. When the continuities of these relative standards

are stable, predictable, and harmonious, according to Durkheim, individuals accept the

equilibrium of social normative regulation of aspirations and satisfactions, characteristic

of healthy societies and healthy individuals, (Traugott, 1978; Cohen, 1959).

Maintaining stability and equilibria between individuals and social groups is

dependent upon social regulation of individual aspirations and expectations; economic

disasters produce rapid declassification, changing the relationship between the

satisfaction of aspirations and the levels of expectations, which has also been observed

to occur in modern high effort-low reward occupational situations (Siegrist,

1997;1996;1990; Theorell, 1992; Converse, 1972; Barnes and Inglehart, 1972).

The sudden acquisition of prosperity, power or wealth also changes social

standards of acceptable lifestyle norms. If one social group suddenly becomes rich, an

alteration occurs in the distribution of social wealth for other social groups, creating an

imbalance in the distribution of social positions, life satisfaction and expectations. This

45

relative inequality in the distribution of social positions is associated with the

disjuncture between aspirations, expectations, and satisfaction, which have been

associated with quality of life measures (Veenhoven, 1995, 1990).

As Durkheim puts it, the scale of the social hierarchy is upset, but a new scale

cannot be immediately improvised to reclassify men and things, thus there is no restrain

on aspirations. Profound disturbances may change social stratification, the principles of

distribution between status and occupational groups, as well as the attendant acceptable

norms of maximum and minimum degree of ease of living. This is a state of social de-

regulation, or anomie. When the stable, coherent pattern of public aspirations,

expectations, and attainments, or way of living, is disturbed at the institutional or social

level because of social crises or rapid transitions, malaise ensues at the individual level

because society has become sick, anomic (Durkheim, 1897).

Although Durkheim’s perspective is historically circumscribed, it has several

modern extrapolations in recent studies of public policy and community cohesion, in

relation to the promotion of health and civic culture (Hoffman-Nowotny, 1972;

Antonovsky, 1993a, 1993b; Stassen, 1988; Andrews, 1976; Bradburn, 1969; Campbell,

1976; Travis, 1993; Srole, 1956; Bell, 1957; Cloward, 1959; Giddens, 1971; Merton,

1964; Willis, 1982).

Another dominant concept of lifestyle relevant for a multilevel model and

utilized frequently in European health research is outlined by Giddens. Giddens tries to

overcome the traditional macro-micro dichotomy, and specifies lifestyle as having two

main aspects: 1.) a self-identity, created and changing in time and space, and 2.)

individual social practices which reflect personal, group, and socioenvironmental

46

characteristics of multiple group memberships. Social class and social milieu are both

important stratification parameters of socioenvironmental factors, as are status groups,

with specific aspiration norms and expected behavior patterns for individual lifestyle

choices. Giddens’ contribution lies in the emphasis on the influence of temporal and

spatial social change upon individual lifestyle patterns, and upon clusters of habits,

orientations, and social identities found in group memberships. These patterns are

constructed as elements of social structure through social interaction by individual

agency, within a cultural and historical context of hierarchical levels of complexity

(Giddens, 1984).

Gidden’s structuration theory emphasizes the reflexive duality of social structure

created by individual actors, who are not abstractions or hypothetical ideal types, or

reflections of internalized social norms, but concrete active agents. Since social

structure is both the means by which action occurs, and the result of individual action,

structuration theory integrates Durkheims’s objective (social) constraint of the

individual with Weber’s subjective (verstehende) and enabling effect upon individual

choice. Whereas Durkheim illustrated the effects of social norms upon individual

health, Giddens points to the social continuity of individual self-identity as a reflexive

element deriving from group membership, which also creates a civic community

through social interaction at various levels of complexity. While Weber was occupied

with the traditional social status hierarchy of socioeconomic position in industrial

society, Giddens addresses multifactorial interaction levels between the individual and a

complex society. Self-identity is expressed in a civic community as a mediating element

between life choices and life chances in the modern and post-modern period for

47

Giddens.

The civic community is an expanded dimensions of lifestyle subsumed by

structuration theory. It has been investigated as a macro effect on health status.

Characteristics of a civic community, such as social capital (civic activity and group

memberships), have been shown to have an important benefit for individual quality of

life (Wilkinson, 2001; Putnam, 1993). Durkheim’s social integration theory has also

been incorporated by research into the effects of a civic community on health primarily

through measures of trust and anomie. The importance of Weber’s measures of social

opportunities, derived from a distribution of socioeconomic positions and status

groups, and individual choices made within the constraints of the social opportunity

structure, have been extensively applied in the WHO distinction between avoidable and

unavoidable conditions of health inequalities. The constraints upon life choices by life

chances has been a challenge for empirical research. The factors resulting in a specific

health status outcome for individuals needs to be identified at the micro level within the

macro social level.

A pragmatic application of the civic community construct may explain

premature mortality and the health crisis in Russia, a natural experiment where

variations in personal and cultural life choices, as well as sociopolitical transformation

of life chances, can be assessed. Research has increased in this area since the collapse of

the Soviet system in 1989-1991. Russia has been described as an hourglass society,

lacking civic culture and especially trust; where an elite, disavowing social

responsibility, skims off the socioeconomic capital from a small middle-class and from

a large number at the bottom of the social hierarchy (Rose, 2000).

48

THE CIVIC COMMUNITYHRQOL has been related to indicators of a civic community, such as social

capital and social cohesion. Part of this research tradition has grown out of the

longitudinal California Alameda County health studies, which began investigating the

impact of anomie, social support and social networks on mortality beginning with a

1965 cohort (Hochstim, 1970). An extensive literature exists now on associations with

morbidity (Berkman and Kawachi, 2000; Berkman, 1985; Doeglas et al., 1996;

O’Reilly, 1988; Lomas, 1998; Berkman and Kawachi, 2000).

Although often used interchangeably, social capital may be conceptually

distinguished from social cohesion. Social cohesion may be a psychosocial subset of

social capital, while the latter may be derived from a political economy framework. One

dominant perspective has defined social capital as the instrumental allocation of goods

and services by individuals through mutual exchange networks of expectations and

obligations (Rose, #318, 1999). Institutional and interpersonal trust in a mutually

beneficial exchange is an underlying social axiom for cementing civic, economic and

political activity. The formation of organizations, small groups or networks as a result

of such normative activity may be one measure of the available stock of social capital in

a culture (Inglehart, 1997). Others have argued that social support and social networks

are only meaningful as interpersonal micro level measures, while both social capital and

social cohesion are intrinsically macro concepts (Berkman and Kawachi, 2000). On the

other hand, social inequality and distrust have been defined as indicators of a lack of

social capital, which is an aspect of social cohesion or social integration (Kawachi,

1999).

Regradless of the nomenclature, or risk of sliding into nominalism, the civic

49

community connotes social integration, in this study, characterized by measures of

horizontal interpersonal relations and vertical formal group relations. Social cohesion

may be understood as the normative social “glue” which links horizontal and vertical

relations.

While civil society implies an independence from state control, a civic

community implies integration with the state through social organizations like trade-

unions, clubs, professional and political participation. The civic sphere is suffused with

social cohesion, derived from values of trust and expected reciprocity, where the public

and private interact with mutual benefit. This normative area is especially conducive for

self-regulated professional activity with autonomy of occupational expertise, which

receives legitimation from the state and provides essential services to the state.

The difference between autonomy and independence from the state is an

important distinction, wherein individual self-interested partnership with the state,

rather than separation from the state, coincides with civic interest. A state with

numerous “independent” social and economic organizations, which are poorly

integrated and lack mutual reciprocity, may result in a corrupt, corporatist, and civil

rather than civic society (Blazer, 1996). Soviet state monopoly of sociopolitical

institutions precipitated demoralisation, separated the activity of citizens from

influencing the state, rather than fostering civic participation. This increased mistrust,

insecurity, loss of mastery over political and life conditions, and other types of personal

disillusionment with public life which have been shown to impact negatively on health

status (Kawachi, 1997).

Whereas social cohesion assumes the mutual social norms and obligations of a

50

civic community, social capital, as the instrumental exchange of goods and services, can

take several forms in either a civil or civic society. Informal relationships of social

capital in the form of friendships and family assumed greater importance in Russia

when social isolation from public participation increased for individuals, and when

formal organization of social capital, such as firms, nongovernmental groups or

voluntary community associations, entered into conflict with Russian state policies and

therefore became unsustainable.

Formal organizations, like professional or trade unions, have been connected to

political parties in post-Soviet Russia, replacing monolithic party control with

politicized groups, further weakening the civic relation between citizens, voluntary

nongovernmental organizations and the state (Kennedy et al., 1998). The lack of social

cohesion in Soviet and post-Soviet formal groups, so characteristic of the civil society

of Russia, was balanced by greater interpersonal reliance between individuals. The

positive association on physical and emotional health of trusting people and having

someone to rely upon when ill was extensively demonstrated in the New Russia

Barometer Social Capital Survey (Rose, #303, 1998).

Primary relationships of social connectivity in the form of friendships and

family assumed greater importance when secondary relationships of community

integration became unsustainable. Research of “healthy societies” has shown how

membership in various social groups is related to the greater prevalence of social trust, a

more effective government, and public well-being. Membership in various social

groups, from sports clubs to kindergarten meetings, has been related to keeping abreast

of political information in newspapers, voting behavior, and a more effective

51

government in Italy (Putnam, 1993). Alameda County studies of mortality have

demonstrated the important protective effects which the number and size of primary and

secondary group networks have upon longevity, even under circumstances of relative

social and economic deprivation (Berkman, 1979).

A sociopolitical model of health and democratization may be balanced by a

psychosocial model for predicting and explaining the increased mortality from

cardiovascular diseases in individuals and countries undergoing rapid social change and

economic transition (Putnam et al., 1993; Siegrist, 1995). He has shown how high effort

and low reward conditions in Germany, among some occupational groups of men, have

resulted in biobehavioural changes and increased cardiovascular disease.

Analogously, when expected social structures and norms are displaced by social

crises and a lack of social coherence is combined with loss of individual life control

(self-regulation), individual self-efficacy is disturbed since a sense of purpose is no

longer connected to social norms and personal success cannot be ensured; self-worth is

disrupted by a change in obtaining social rewards from occupational and economic

status (loss of savings from inflation, loss of wages from unemployment or unpaid

labor, loss of access to material and social resources from increased prices, etc.); self-

integration is disrupted by secondary and primary group disconnections and inability to

obtain social support, indirect economic assistance, or other forms of informal access to

social and material resources.

Certain tension alleviating behaviors, which act as central nervous system

depressors, such as smoking and alcohol consumption, may increase under conditions

of rapid social change. The chronic stress response of the autonomic nervous system

52

then precipitates physiological changes, cardiovascular or other disease, and premature

mortality. The social context of institutional environments and secondary group

relationships (such as occupational groups) is thus linked with mediating psychosocial

behaviors and physical responses at the micro level, as well as to macro level

community health status (Hertzman, 1996).

Cross-cultural research which searches for universal aspects is also essential in

pointing to differences in relationships between social contexts and individual health

(Kohn, 1988). Those relationships found among Western European countries may not

be as uniform for countries of the Former Soviet Union or Eastern Europe. Market

transitions increase the inequalities of the stratification structure within the FSU and

countries of CEE. There are concomitant increases in life uncertainties, individual

controls over the social environment are weakened during massive macro changes in the

political - economic structure. Distress levels may be translated into physiological

processes producing increased disparities in health at the individual, group, and

community levels. Cross-cultural research in public health is clearly important in

systematically clarifying the processes of social change which have resulted in such

structural and cultural variations of health-related quality of life.

LIFE CHANCESLife chances, the social opportunity structure, can be defined as having both

macro and micro aspects, each of which has been related to health status. The size of the

range in the distribution of socioeconomic resources has been demonstrated to be a

predictor of community mortality and morbidity patterns. Socioeconomic position is

understood as a social context which constrains individual opportunity (Kawachi and

Berkman, 2000).

53

Individual social status and socioeconomic position , however, in following

structuration theory, is not predetermined but created interactively with social context.

Individual social status dimensions of education, occupation, and income, as well as

individual health status, are situated within the distribution of production, acquisition,

and consumption of goods (Kaplan, 1997; Wilkinson, 1997; Marmot, 1997). Relative

inequality, such as the degree of steepness in hierarchies of social positions, social

marginality, and normative disorganization, have been more strongly associated with

mortality within countries than have absolute measures like average income or policy

norms (average price of a food basket). Lower deprivation between groups relative to

each other has been related to higher life expectancies, egalitarianism, smaller ranges in

the distributions of national income and access to social resources (Wilkinson, 1997).

The lifestyle construct was broadly based upon Weber’s distinction between

individual life choices constrained within a structural context of life chances, as well as

Durkheim’s theory of normative control of individual conduct through cohesive social

institutions. Social institutions govern individual and group action through a common

value system and a stable, predictable body of rules. Institutions also exert influence

upon the structural array of life chances, the social possibilities and probabilities within

which life’s actions occur. These common values and normative constraints form a

consistent and cohesive context, setting out relative social and cultural expectations and

permissible action.

Life choices or individual conduct occur within this institutional context and

have social consequences. However, individual behaviors are limited by a milieu often

beyond individual control, as in exposure to environmental pollutants, or by social

54

institutions, like professions, the job market or residential neighborhoods, where the

individual must negotiate partial control. Individual and community well-being is

disturbed during periods of sociopolitical crises; conflicting, inconsistent occupational

opportunity structure; or when the cohesion between social values and rules for action,

or social expectations and individual aspirations to acquire social rewards (goods and

services) is frustrated and disrupted. Changes in the range of social stratification due to

sociopolitical disruption of life chances and a civic community transform some

emergent properties of social structure which, in turn, affect public health and quality of

life.

HEALTH CHOICESAlthough chronic disease and morbidity surveys in populations have a long

history from the early part of the 20th century, the Alameda County studies were among

the first longitudinal studies of health and longevity (Syme and Berkman, 1976;

Breslow, 1996). They grew out of the tradition of the United States National Health

Survey, established in 1956, and the U.S. Commission on Chronic Illness, which

conducted county-level surveys on chronic disease in 1957 and 1959. The Human

Population Laboratory in Berkeley, California, was established by the National

Institutes of Health in 1959 to conduct the Alameda County studies, focusing on

assessing the interrelationships between the WHO health dimensions, demographic

characteristics, and lifestyle, including personal habits, socioeconomic, cultural, and

environmental factors (Belloc, 1971, 1972, 1973; Hochstim, 1968, 1971; Wiley and

Camacho, 1980).

This effort engendered nearly half a century of prospective longitudinal research

since the initial Alameda County sampling cohort in 1965. Mortality and health status

55

follow-ups were first conducted in 1974, and almost twenty times since the original

cohort. The Alameda County studies demonstrated what has now become accepted as

common knowledge, that lifestyle and social relations influence longevity and quality of

life, controlling for initial physical health status.

The debate defining lifestyle as an individual level risk factor for specific

populations has become inherent in the construction of health insurance benefit

packages and continues as one of the most controversial and fundamental issues in each

American presidential race, as well as in Russian parliamentary politics surrounding the

implementation of private health insurance legislation (McKeehan, 1995).

In the USA, the Surgeon General's Goals for the year 2000, continued to

emphasize the research results of the Alameda County Studies, identifying three broad

areas which influence population health status and require public policy intervention:

1.) first, health practices or Health Promotion - decreasing risk factors, personal habits

such as smoking, lack of physical exercise, poor diet, alcohol abuse; 2.) second,

ecological factors or Health Protection - decreasing occupational and environmental

toxic exposures, decreasing accidents; and 3.) third, medical care factors or Preventive

Health Services - increasing access to services such as prenatal care, infant programs,

family planning, hypertension control (CDC, 1991). Health promotion and disease

prevention indicators were included in the Alameda County survey instrument.

The long-term research objectives of the Alameda studies were replicated in the

construction of the Moscow HRQOL Survey instrument. Individual preferences and

personal health habits, such as physical exercise, smoking, and alcohol intake were

included as proximal determinants of health which were amenable to individual control

56

and change; initial physical health status was controlled by height.

Height (like age, sex, marital status, educational level) has been demonstrated to

be a direct measure of the average “net nutritional status of the members of a human

population during their childhood and adolescence”. It is an indirect measure of

childhood welfare, which is more comprehensive than income or wealth, having a

lifelong effect which influences mortality (Blaxter, 1990). The body mass index was

included as a choice variable indicating quality of nutrition and diet. Periodic nonacute

checkup visits to dentists and physicians were added as measures of health promotion

and prevention. The willingness to pay for private medical services was an item which

indicated changing economic values in the Russian public, accustomed to free

socialized medicine.

MULTILEVEL MODELA multilevel analysis of infant mortality and community risk factors

demonstrated that community context, like family poverty rates and local welfare

expenditures, influenced mortality as a “fundamental cause” (Matteson, et al., 1998). In

various studies, smoking, drinking, long-term illness, standardized mortality ratios, and

female heart disease death rates were reported to be affected by household or

neighborhood deprivation measures as independent main effects, controlling for a

variety of individual risk factors, like social class, age, ethnicity, car ownership, or pre-

existing health conditions (Duncan et al., 1999; Gould, 1996; Langford, 1996; LeClere

et al., 1998; Rice et al., 1998).

In a study controlling for ecological effects, mean area consumption of alcohol

was strongly correlated with the prevalence of heavy drinking in areas of England,

suggesting that social factors which increase the average consumption of alcohol in the

57

population of an area may increase the prevalence of heavy drinking among individuals

(Colhoun, 1997). Further, it has been shown that not taking into consideration

nonresponse rates can seriously bias the relationships found (or not found) within a

survey. Income inequality was not found to be significantly related to life expectancy in

a study of 13 OECD countries until substantial nonresponse rates were corrected in

subsequent analyses (Kawachi, 1997). However, such models are still left at either the

micro or macro level, precluding an examination of influences between levels on the

social dynamics of HRQOL, which may operate differentially yet simultaneously at the

community and individual level.

The hierarchical linear model has three different analytical propositions based

on different data levels: individual and group or contextual levels (there can be any

number of hierarchical levels); and micro propositions, macro propositions, and macro-

micro interaction propositions (Figure 3; Snidjers and Bosker, 2000).

FIGURE 3: THE STRUCTURE OF A MICRO, MACRO, AND MULTILEVEL MACRO-MICROPROPOSITION

There are various common types of contextual effects which may be modeled

with hierarchical linear regression. A simple macro-micro relationship, as depicted in

58

Figure 3 (Snidjers and Bosker, 2000), refers to the direct contextual effect on

individual members of the “context” or group. For example, ghetto crime rates induce

criminal behavior in ghetto teenagers. Or, national income inequality is associated with

premature mortality. A contingent macro-micro relation control for another micro level

factor, such as ethnicity, income, education or occupation. This modifies the

proposition, for example, so that ghetto crime rates induce criminal behavior in ghetto

teenagers, controlling for ethnicity. Or, income inequality is associated with premature

mortality, controlling for the development of a civic community.

The third common macro-micro relation is the cross-level interaction, where the

relation between micro factors x – y is dependent upon the contextual factor, Z, or in

other words, the relation between context Z and dependent micro variable y is

dependent upon micro variable x. The proposition would change, for example, to

considering the relation between ethnicity and crime is dependent upon ghetto context,

or in other words, the effect of ghetto context on teenager criminal behavior is

dependent upon ethnicity. Or, a civic community and premature mortality are

associated with national income inequality, in other words, national income inequality

and premature mortality are dependent upon the development of a civic community.

There are also reverse micro-macro relationships, where the individual factors

impact upon contexts, for example, poor individual health affects public health policy.

More complex macro-micro-micro-macro relations may also be modeled in a variety of

combinations (Snijders and Bosker, 1999). The proposition of how individual quality of

life affects public health policy may be schematized as follows: preventive policy

promotes implementation of equity which encourages access to social resources and

59

positive health lifestyles, which in turn leads to less morbidity and successful public

health policy (Figure 4; Snidjers and Bosker, 2000).

FIGURE 4: A STRUCTURE OF SELECTED MULTILEVEL MACRO – MICRO PROPOSITIONS

SUMMARY OF RESEARCH HYPOTHESES

This chapter described the social determinants of health, conceptualized as a

multilevel health profile of a civic community, life chances, and life choices.

Three dimensions of social determinants were related to a broad lifestyle

construct as predictors of individual health. The general construct of lifestyle was

derived from a broader sociological conception of individual action (Blaxter, 1990).

Lifestyle was defined as a collective and cultural pattern of ways of living, including

life chances, characteristics of a civic community, and individual life choices, in

keeping with the four Black Report explanations of health inequality. Multiple

dimensions of lifestyle were formulated to minimize measurement artefact due to single

indicators, increasing the reliability and validity of the construct (Sullivan, 1979;

Carmines, 1979).

60

Lifestyle incorporated first, a social selection and structural aspect (life

chances), the clustering of social characteristics among groups which influence

individuals’ options to live in a particular way within those groups; second, a

sociocultural aspect (civic community) which provides the normative content of specific

values and goal-oriented behavior; and third, individual health-related behavior (life

choices). The distribution of life chances served to differentiate between relative

inequality and average inequality among individuals, as well as urban areas.

The three dimensions of lifestyle were operationalized as shown in Figure 1:

1.) Macro indicators of life chances:

average inequality as a factor scale of access to material resources

between urban residential areas; the score was derived from 33 separate indicators;

the construction of this factor is described in the methods section;

poverty risks in urban areas; a proxy measure for economic status

since no direct measures of income at either the macro or micro levels were

available in the survey; poverty risk was assessed at the urban area level as the

ratio of families with five or more members in each area. Poverty was most likely

among large families in Russia, with 3 or more children. Based on the average

family size of 3.1 in 1991, families with five or more members could be

considered to be at risk of being poor. It was assumed that the likelihood of areas

having a greater concentration of poverty or other form of deprivation if they had a

greater proportion of large families was reasonable in lieu of a more direct

measure of average income levels.

relative inequality as multiple indicators of the composition of social

61

status in urban areas, including education, occupation, and employment. At the

community level, ratios of blue-collar manual to white-collar non-manual residents

within areas, ratios of below average to above average living space per person, and

ratios of low to high educational levels were measures of relative deprivation or

inequality between the 33 administrative areas in the city of Moscow. The

composition of areas was defined by the simple proportion of white collar workers,

blue collar workers, educational level, family size, or apartment size. In and out-

migration by social characteristic was not available for urban areas to examine

direct selection of specific persons in urban areas. The rate ratios were contextual

indicators of the relative distribution of social groups to each other within urban

areas.

2.) Micro indicators of a civic community:

social cohesion as a scale of social anomie;

social support as a scale of the quality of marital relations; and

social capital as primary group connectivity with friends and family,

and secondary group community connectivity with group memberships or

activities.

3.) Macro and micro indicators of health choice:

a set of micro level personal health practices, including smoking,

alcohol consumption, physical fitness, body mass index; medical utilization of

preventive physician visits and acute care visits; and economic values such as the

willingness to pay privately for physician services.

an aggregated macro indicator of alcohol intake within each area.

62

A multilevel theory of health explains how multidimensional health outcomes

may be determined by behaviors of individuals within groups, as a consequence of

social relationships between networks or groups, within a larger social context. A

multilevel statistical model explains the variation in physical health by apportioning the

effect directly to characteristics of the individual, to community contexts, or to the

interaction between the individual and community context, without losing information

about the independent effect of each level of determinant on health outcomes.

Multilevel modeling is one parsimonious method, among several more sophisticated

estimation procedures, which can provide a robust analysis of the components in social

hierarchies, networks, or causal paths, which integrates conceptual and statistical

clustering.

Based on the research goals outlined in Chapter 1, the following hypotheses

were formulated to estimate 1.) individual level effects and, 2.) contextual level effects

on health-related quality of life:

1.) a.) the components of HRQOL outcome, life satisfaction, life happiness,

self-rated health status, and physical health, were expected to be influenced by different

sets of individual determinants; the components of HRQOL contribute unique

information for assessing outcome; components can not be eliminated or conceptually

substituted for each other without losing significant information on outcome. Choosing

one specific or several outcomes has significant implications for public health

interventions.

b.) it was posited that there was a hierarchical effect of life chances on HRQOL

outcome and that it was stronger than the effect of poor health choices; people with less

63

education or lower status occupations were at greater risk for poorer HRQOL even if

they made all the healthy choices.

c.) the hierarchical effects of individual life chances on HRQOL outcome are

expected to be moderated by the effects of a civic community, in the form of social

cohesion, social support, formal and informal networks, controlling for health choices.

2.) a.) it was posited that the distribution of inequality was an urban-level

contextual determinant of individual health; characteristics of the social context,

measured as average and relative inequality, were determinants independent of

individual risk factors such as social status and health choices;

b.) because of the effect of cumulative risk factors, poor health choices among

resident in high inequality areas would have a greater impact on physical health than

among residents in low inequality areas who made the same choices;

c.) living in a civic community which connects individuals by social capital,

social support, social cohesion was expected to have a positive effect on physical

health; and a protective or buffering effect on those living in high inequality areas,

controlling for social status and health choices; the buffer effect was expected to be

greater for low status individuals living in high inequality areas;

d.) in sum, it was expected that the risk for poor HRQOL was a cumulative

function of living in high inequality areas, which were less likely to exhibit the

parameters of a civic community, and more likely to have residents with low social

status rather than individual health lifestyle choices.

64

CHAPTER 3: CONSTRUCTS OF HEALTH OUTCOME“Scientific” operational definitions rather than popular, self-reported definitions

tend to look at behaviors in terms of consequence and not internal states or processes,

exemplifying the difficulty of verifying the validity of a concept in various cross-

cultural populations, where meanings may differ due to social context. Different health

status indexes reflect differences in measurement and conceptualization, and thus have

different referent phenomena. Social concepts are especially subject to the "Uncertainty

Principle", which operates beyond the upper limit of measurement accuracy. This raises

the question of how useful it is to measure the degree of HRQOL with indicators which

are fuzzy or imprecise, without treating the degree of fuzziness itself as a separate

variable. Otherwise the fuzziness is referred to as residual error, as the unspecified part

of the conceptual definition which may confound measurement and theory.

The degree of social fuzziness in cross-cultural comparisons may be captured by

self-perceived, self-rated indicators, which describe populations from their own vantage

points rather than those of the expert. In other words, HRQOL is both methodologically

and conceptually a multidimensional, general, and abstract construct rather than one

which is unidimensional, unique, and concrete.

In cross-cultural research, empirical relations such as reliability and validity

between concepts and indicators may be different for different populations. The

generalizations drawn from population-based research to theoretical concepts are not

universal given that the rules of inference are statistical (Coleman, 1986). It is argued

that the validity of indicators must be established in each population. Operational

definitions of indicators are often confused with theoretical constructs and linguistic

terminology. Theory requires a meaningful specification of explanatory, conceptual

65

interrelations, not simple quantification of a phenomenon.

Global, self-perceived indicators of health status and quality of life transcend the

issue of whether linguistic terms have phenomenal referents in society, since concepts

can exist in language and not in society, but still have meaning for the individual. The

confounding of measurement and theory is often due to the reification of concepts:

concepts may have the same meaning in different cultures but operational indicators of

concepts may have different meanings in different cultures. Nonetheless, indicators

should be applied appropriately for a specific culture, so that a study of polygamy in

Great Britain would be difficult to entertain. Analogously, measuring health and

happiness among women in Somalia and Great Britain by the questions, “how would

you rate your health” or “how happy are you”, will assume cultural referents within the

self-perceived indicators. The cultural referents need to be defined explicitly only in a

comparative study which investigates the influence of that specific cultural phenomenon

(Becker et al., 1978).

A central problem in theory construction is the definition of the phenomenon to

be investigated. This introduces language as a central category of theoretical concern in

cross-cultural research. The more abstract or complex the phenomena, the more the

conceptual terms correspond to common-language categories, rather than those of

expert, professional jargon, and have greater relevance for the everyday life of people

who perceive and define them. The meaning of everyday terms has greater variability,

less uniformity and increased subjectivity, therefore multitheoretical constructs

operationalize such everyday concepts, which are not amenable to a single, unified

explanatory theory (Elder, 1976).

66

The search for a common scientific terminology has led to a premature demand

for the cross-cultural identity of meaning as the golden criterion for construct validity.

Equivalent but nonidentical concepts have often been confused with being incongruous

or invalid. Cross-cultural equivalence does not necessarily imply identity, even under

conditions of ceteris paribus (all other things being equal). Sociocultural norms are

tacitly imbedded in health assessments and sociomedical indicators. What is accepted as

healthy in rural Siberian villages, for example, may not be tolerated as healthy to the

same extent in Moscow or on an Arkansas farm. The universalism of scientific validity

has often been assumed to be independent of culture, although Merton (1959) and

researchers such as Kohn (1987) and Elder (1976) have long argued that scientific

activity is situated in a sociocultural community.

Comparative research is thus not only confounded by the social context of

research methods, but also by the social context of the research problem. Some concepts

are intrinsically more comparative than others and are thus better suited for application

to cross-cultural research. Health and quality of life are such comparative categories of

daily living, abstract and non-specific, and therefore more suited for cross-cultural

research.

Common dimensions of QOL, such as objective, macro, community in contrast

to subjective, micro, individual measures do not imply social rather than psychological

dimensions, but rather different levels of abstraction and generality/specificity. QOL

research requires simultaneous work in theoretical and methodological development,

particularly looking at the problem of the relative rather than absolute distribution of

well-being. It is necessary to isolate the endogenous and exogenous variables affecting

67

different dimensions of QOL as a dependent, outcome variable, as well as an

independent, predictor variable. (Siegrist, 1987).

Cross-cultural research in health-related quality of life has been grounded in the

social indicators movement, where often expert and "professional" definitions of

indicator content have been supplemented by lay "self-perceived, subjective" indicators

rather than "official, objective" data. Quality of life and health status have been defined

both at the “subjective” and “objective” level (Katz, 1987). Most researchers in the field

define “quality” as a subjective rating on a continuum, scale, ranking, or ordinal range

from high to low or from better to worse. The term “life” subsumes both objective and

subjective dimensions. Objective dimensions include environmental conditions, work,

community, food, shelter, socio-political, cultural, and ecological. Subjective

dimensions include attitudinal conditions of life, such as global or domain-specific

feelings, affective, cognitive, or value-orientations (Schuessler, 1985; Armer and

Grimshaw, 1974).

Global self-perceived indicators minimize the problems of cross-cultural

translations and maximize functional and semantic equivalence, insofar as the concepts

are abstract enough to leave room for individual meaning without losing the element of

general applicability over various populations. The issue of selecting equivalent

measures for cross-cultural research tautologically assumes the validity of the original

concept as a universal: it is comparable only if equivalent, if equivalent therefore

universal; similarities between cultures are emphasized and not differences. But it is

thus essential to look at concepts with divergent methods and indicators.

It is of questionable validity to emphasize similarities in comparative research,

68

as it necessitates the use of equivalent concepts. There needs to be an increased

emphasis on demonstrating the validity of concepts at various levels of specification:

not only subjective and objective, specific and global, but micro and macro.

Standardized methods and literal translations of instruments are useful only if

conceptual equivalence is assumed to exist a priori, but this is itself a restrictive

research strategy, given that validity is dependent upon independent replication with

multiple methods (Grimshaw, 1969).

The mistaken emphasis upon equivalent measures further assumes the validity

of the measures, and that there exists a single best way of measuring concepts instead of

multiple ways. Various measures of the same concept may also be valid, as in

translations (Deutscher, 1968). It is important to consider whether an indicator measures

the “same concept” in different cultures, i.e. conceptual equivalence, or a culturally

specific concept, i.e. indicator equivalence of measures. For example, intelligence is a

universal phenomenon (conceptual equivalence), but the IQ test may measure exposure

to certain educational systems, cultural ideas, or standardized test taking techniques

(indicator equivalence).

The validity of a concept may be demonstrated in one culture and within another

culture, even with different operational indicators. If a construct is valid in one culture,

and valid within another culture, then it is valid cross-culturally but may still be

nonequivalent. This may be especially true for consensual rather than standardized

translations.

Translations may be seen as nonequivalent measures, especially of difficult or

linguistically unique concepts. Translations into different languages may therefore

69

produce different measures of the same concept from the original language, and in

another language the translated measure is nonequivalent both as concept and indicator.

Standardized “authorized translations” are useful only if conceptual equivalence has

been demonstrated already. Otherwise, the concept of cross-cultural validity itself must

be revised to include the preeminence of intra-cultural validity.

Some concepts are culture-specific, other concepts are inter-cultural, that is,

“universal“ or “transcultural”, rather than culture-bound. Global indicators, such as

general QOL or self-rated health, are such linguistically abstract transcultural concepts.

Any given language is in and of itself cultural, therefore an increased level of linguistic

abstraction is related inversely to cultural specificity (such as mathematics or symbols).

However, the problem of operationalizing a concept reduces abstractness, necessitating

the use of multidimensional constructs such as the global indicators of happiness,

satisfaction and general self-rated health to capture what was a single term, such as

quality of life.

Global, self-perceived indicators, which have the greatest abstractness within

cultures, tap the “universal dimension” or transcultural level with cross-cultural

referents (Werner and Campbell, 1970).

GLOBAL AND DOMAIN SPECIFIC INDICATORSThe following indicators were included in the outcome measures of the Moscow

health profile: general life satisfaction and life happiness as global components; specific

domain satisfaction with job and domain happiness with family; general self-rated

health as a global component and domain specific physical health (Table 2).

70

TABLE 2: HRQOL COMPONENTS OF THE MOSCOW CITY HEALTH PROFILE

HEALTH DIMENSIONSGLOBAL DOMAIN SPECIFIC

Self-reportedphysical health

Profile score disability/impairmentschronic conditionsacute symptomshigh energy levelslow energy levels

The communist ideology attempted to eliminate money as a mechanism of

“private property”. The command economy in Soviet Russia was organized to function

with social and economic planning, rather than a supply-demand based price system.

The post-Soviet economy of Russia has not yet made a successful transition to a

monetized market economy. Major research of income in Russia, after January, 1992,

the moment when price controls were lifted and the transition to a market economy was

officially initiated, examined the association between money income and household

welfare or well-being (Rose, #215, 1993; RLMS, 1992-1999).

These surveys have demonstrated that alternative means to official money

income have developed in Russia in order for households to acquire and maintain their

welfare: barter; reliance on social capital; participation in alegal, nonmonetized social

economies (like household production of food); and working in illegal, second

economies (like having social connections to the access of foreign currency). The

application of various quality of life indicators in Russia as outcome measures of

“objective” and “subjective” well-being have greater utility for public policy than the

use of traditional economic measures of household income.

71

STRUCTURE OF QOL global indicators may be single items or additive sums of specific domains like:

family life; housing, health, job, occupation, environment, etc.;

global indicators vary less for ascribed, stable group characteristics, such as gender

or ethnicity; variation increases among social status characteristics, such as income,

education, and other fluctuating situations in health, family life, social relations, etc.;

SELF-RATED HEALTHThe definition of health as optimum physical, mental, and social well-being was

reconceptualized as a 1.) global, self-rated general health status, and 2.) domain-specific

component of quality of life - self-perceived and self-reported functional limitations,

impairments, chronic conditions, acute symptoms, and energy levels as indicators of

physical status derived from the Alameda County Physical Health Profile.

Health status has been traditionally measured as a multidimensional index plus a

global index because the global index captures "unquantifiable" elements or residual

nonspecific content associated with health. Global self-perceived health status cannot be

reduced only to a residual category of HRQOL. Self-rated health has been associated

with health outcome independently of medical, physiological, psychosocial, or

behavioral risk factors (Idler and Benyamini, 1997).

The global health rating gained popularity after the Alameda Studies reported

the usefulness of lay definitions of health. Wholistic, lay definitions of health status

were also related with exposing the “iceberg” phenomenon of under-reported morbidity

in a simple, direct, and individually-meaningful way (Kaplan and Camacho, 1983;

Verbrugge and Ascione, 1987). An overwhelming number of health surveys in the

1980’s included the self-rated health indicator. It has also performed well in cross-

72

cultural and comparative replications, providing invaluable information as a global

dimension of health status.

Self-rated health was included in the Moscow health profile for the following

reasons. Domain specific dimensions of health status capture information specific to

physical, functional, physiological, mental/ psychological health:

then global health, as a residual category, measures something “more” than

medical evaluation and needs to be included to partial out remaining variation; self-

perceived health status has been shown to be an independent predictor of health

outcome and not only a proxy measure of “objective” health status;

then global health is a “subjective” domain wherein an individual has access to

information, which is important in cross-cultural studies, where the appropriateness and

equivalence of indicators are additional methodological considerations for validity and

reliability;

self-perceived health has been shown to be one of the strongest predictors of

mortality in many studies, independent of age, sex, initial physical health, or medically

evaluated “objective” health status;

self-perceived health can be included not only as an outcome measure but also as a

self-assessed nonspecific risk factor.

PHYSICAL HEALTHIn addition to global health, domain specific dimensions were included in the

Moscow health profile based on the four major domains of the Alameda Physical

Health Profile: disability/impairment, chronic conditions, acute symptoms, and energy

levels. The Alameda Physical Health Profile was developed as a measure of general

physical health for community-based samples. The Alameda Physical Health Profile is a

73

multidimensional, non-disease-specific, self-perceived measure. It has been applied in

cohort studies of mortality, controlling for a variety of psychosocial and demographic

factors in numerous community studies.

The Alameda profile was one of the first operational quantitative approaches to

the WHO definition of health. This concept of health constructed a multidimensional

spectrum with three axes: the physical, mental, and social, where an individual could be

simultaneously evaluated on any one or all three axes. The statistical formulation was

first developed by the Alameda County studies as part of a project to measure the

morbidity in a general community, using sample survey methods (Breslow, 1972).

While other parts of the Alameda instrument measured mental and social health,

morbidity was operationalized by the physical health profile, consisting of specific

chronic diseases, symptoms, disabilities, functional status, and energy levels

The Alameda Physical Health Profile may be viewed as an “objective”

component of health, insofar as the scale is composed of multiple observable items.

Global evaluations of self-rated health may be viewed as the "subjective" component of

health insofar as the rating is non-observable self-perceived assessment. The Physical

Health profile is weighted toward an absence of disease continuum, whereas self-rated

health operationalizes well-being and taps into nonspecific, unlabeled conditions, as

well as a state of being.

The Physical Health Profile has been successfully used as a measure which

compares individuals, social groups, or entire communities over time (Brock et al.,

1988; Gottlieb, 1984; Metzner, 1983; Pope, 1982; Reed, 1983; Wilson and Elinson,

1981), and was applied for the first time in Russia within the Moscow Health Profile.

74

The post-Soviet Russian Federation consists of 21 autonomous republics, 55

regions (including 6 “krais”), 10 autonomous districts and 2 cities which have the legal

status of regions, Moscow and St. Petersburg (Figure 5). Together, they make up 89

constituent administrative units of the Russian Federation. Many of these units are little

known outside Russia, even though several are bigger than many European countries.

The health care crisis in Russia represents one of the most significant challenges to

public health policy today. The primary question for post-Soviet health policy centered

on the extent to which public health functions could be privatized.

75

CHAPTER 5: HEALTH INEQUALITIES IN SOCIAL CONTEXT

This chapter will review the international, national, urban, and small area

variation of health in relation to socioeconomic inequality. Health is an international

phenomenon, situated in the larger context of a global community. A plethora of

comparative studies have sought to account for a growing gap of health inequalities

within and between nations. The East-West health gap is related to the geographic

distribution of nations, as well as to their political, social, cultural, and economic

profiles (Figure 6). The health gradient among nations has been shown to follow the

distribution of egalitarianism; economic; sociopolitical choice; or personal freedom

(Hertzman, 1995).

The mortality differential among nations has been associated with various

theories of the East-West health gap, some of which are dependent on policy:

socioeconomic transformation; environmental pollution; lack of an adequate social

safety net; relative poverty, socioeconomic deprivation; historical, generational effects

of a Soviet heritage; regional disparities; psychosocial stress; and others are dependent

on individual lifestyle – poor health practices and violent behavior (Hertzman, 1995).

Public health is affected not only by the gradient of public and personal wealth

but also by the social stress of political turbulence and uncertainty. The rapid rate of

social change among the eastern European countries and the Newly Independent States

(NIS) in the transition from socialist to market relationships has accelerated

deterioration of health status. There are common factors and profound lifestyle changes

occurring in both eastern Europe and the NIS, accompanied by new values, the

reexamination of acceptable norms and the creation of socioeconomic institutions.

76

However, the complex structure of social factors which might explain the East-West

health gap may differ within and between East European countries, and differ from the

factors salient in Russia.

FIGURE 5: Changes in Mortality Rates, 1970-1990/1993, selected West and East Europeancountries

(source: WHO Health For All database)

77

INTERNATIONAL CONTEXT OF HEALTH INEQUALITYGeneralizations derived from data in one specific country, for example Poland

or Albania, cannot be applied without additional verification for Russia. In addition, the

issue of data quality and mortality/morbidity classification consistency between

countries must be considered when comparing data (Figure 7;Goldstein et al., 1996).

Mortality rates for eastern Europe do not show as severe fluctuations for the same age,

sex, and cause-specific death rates as trends for Russia or other NIS countries,

particularly after the sociopolitical failures of the Perestroika Period . Mortality

increases were substantially larger for Russia than eastern Europe in 1990-1994,

principally due to cardiovascular disease and external causes of death - homicides,

alcohol poisoning, accidents and violence. Self-rated health was worse and

dissatisfaction with income was greater in countries of eastern Europe than in the West

(Carlson, 1998).

It is noteworthy that 1991 is a focal period for health status changes, as well as

sociopolitical and economic changes, for Russia and the NIS but less so for countries of

central and eastern Europe. Another East-West contrast is the integration of East and

West Germany. After 1989 and German unification, the life expectancy of former East

Germany (DDR) seems to have initially fallen, recovered, and then made substantial

progress toward former West German (FRG) levels, as compared to pre-1989 levels and

to other countries in eastern Europe.

At the global level, income and life expectancy are strongly associated within

and between countries. Life expectancy is associated less with average deprivation as

measured by the Gross Domestic Product among economically similar countries (for

78

example, established western market economies (EME)), and more with relative

deprivation, when comparing economically dissimilar countries (for example, former

socialist economies (FSE)) with economically similar countries (Figures 8) .

The relation between average income, whether GDP or per capita, and life

expectancy is attenuated after a certain standard of living is achieved as has been the

case in market economies. Average income is weakly related to mortality within

wealthier countries, like the United States, which has one of the highest standards of

living. The relative distribution of income has been demonstrated to be more strongly

associated with differences in death rates within wealthy countries, given uneven

distribution and concentration of resources (Wilkinson, 1997). Mortality tends to be less

in countries with more egalitarian distributions of income like those of Scandinavia

where relative deprivation is less pronounced.

Health inequality is experienced differentially by gender in the former socialist

economies of Russia and East European versus West European countries (Goldstein et

al., 1996) . This health disadvantage has affected the productive male population

(Figure 9) in lower GDP countries to a greater degree than women (Figure 10), and in

a different manner. Russian male life expectancy has plummeted below all other former

socialist countries, disproportionate to decreases in average inequality or GDP. The

relation between life expectancy and GDP among men in FSE countries is not as clearly

linear as among men in market economies, suggesting factors other than economic may

be more influential in the loss of life expectancy.

The life expectancy of men in former socialist economies appears to have almost

an inverse relation to average income as compared to either men in market economies

79

or women in former socialist economies. The life expectancy of men in the wealthiest

of the former socialist economies is also the lowest.

Although Figure 10 indicates a linear relationship between increasing national

income (per capita Gross Domestic Product) and increasing life expectancy among

women in countries with western economies, that is not the case among women in

countries with former socialist economies. This suggests that women in FSE countries,

like FSE men, are at risk for a lower life expectancy due to factors other than income.

The effect of average income on the life expectancy of men and women in western

economies, however, is consistent with a direct relationship between wealth and health.

80

FIGURE 6: Trends in male life expectancy at birth in Russia, Former Soviet Republics,Central and Eastern Europe (CCEE), 1970– 1994

(source: Goldstein et al., 1996: 14).

81

These data may be accounted for by a differential distribution of inequality and

variation in the factors comprising inequality among women in different countries as

contrasted to men. The differences in life expectancy between FSE and EME countries

may also be accounted for by the relative inequality in sociopolitical institutions

(Wilkinson, 1997). Clearly, structural factors of nations such as development of

egalitarian institutions in addition to relative socioeconomic position must be

considered, for example among gender groups, to explain such striking international

differences in health status. This is more so in the case of Russia, which appears to

have had the greatest changes in male life expectancy in the 1990’s.

Several hypotheses have been offered as explanations for the East-West health

gap phenomenon. Notable are the models outlining the impact on health and well-being

of a civic community, which integrates individual citizens and encourages participation

with the state in a mutually beneficial relationship, in contrast to a society which

segments public and private spheres by a hegemonic state (Blazer, 1995). The structure

of soviet culture separated the individual from involvement with the state and reinforced

personal powerlessness, diminished personal initiative, and divided the private and

public norms of individual values and action (Bobak, 1999), in the context of state

rhetoric which identified personal good with state ideology. There was a lack of

institutional integration of the family or community into the soviet vision with a

concomitant exclusion of the individual from participating in the state monopoly over

public and private life. The Perestroika period prior to the collapse of the Soviet Union

was plagued with an intensification of disparity between personal achievements and

satisfactions possible within a socialist socioeconomic and political structure when

82

FIGURE 7: Male Life Expectancy in Established Market Economies (EME) and FormerSocialist Economies (FSE) by per capita GDP of country, 1993

FIGURE 8: Female Life Expectancy in Established Market Economies (EME) and FormerSocialist Economies (FSE) by GDP of country, 1993

83

personal aspirations escalated and emulated those in the West. The information

explosion which followed the breakup of the Warsaw Pact in 1989 contributed to a

markedly increased awareness of relative deprivation (Hertzman, 1995).

NATIONAL CONTEXT - RUSSIAThe health status of nations deteriorated as social cohesion deteriorated with the

ideological and market transformations of the 1980’s. Britain and eastern European

countries experienced increases in relative poverty, income inequality and premature

mortality among the working-age populations (Wilkinson, 1996). A similar dynamic

operated in Perestroika Russia, when widening income differentials within the country

were due to fiscal policies which cut back the communist welfare state to stimulate

economic growth and privatization (Kaplan, 1998), changing the distribution of social

status (Rimashevskaya, 1997).

Dissolution of the Soviet Union and international competition also weakened the

market position of those with inadequate education and job training, while enabling

others to flourish in an unregulated neoliberal market, which increased social

stratification and income inequality. Working-class male laborers in Russia, with

minimal vocational education, have been at greatest risk for unhealthy lifestyles and,

consequently, poor health (Cockerham, 1999). Life expectancy in Russia has been

associated with increased inequality of income between regions (Walberg, BMJ),

independently of per capita income (Kennedy, 1998).

Deterioration in the economic transition period after 1989 was characterized by

a drop in real wages, increased unemployment, unpaid state employment, and an

overall fragmentation of social services in Russia. There was also an increase in real

poverty and a larger concentration of wealth, resulting in a steeper hierarchy of income

84

distribution within Russia (Eberstadt, 1999). Russia has been undergoing an erratic

transformation from a communist state since 1991, reorganizing its military, political,

economic, and social policy. The issue of a deteriorating social safety net as a

consequence of transition to a market economy has been a crucial factor in decisions of

the IMF and World Bank for providing economic support for post-Soviet Russia.

Russia’s public health crisis is historically unprecedented. No other

industrialized country has ever experienced such a severe and prolonged deterioration

during peacetime. The loss of life from this health crisis in Russia has been a

catastrophe of historic proportions, which has been estimated by the World Health

Organization, for the period 1992-1995, as exceeding 1.8 million deaths, more than died

during the entire first World War (Eberstadt, 1999).

A substantial gender gap of increased mortality among Russian men was evident

in the cause-specific death rates for each year from 1970 to 1995. After an increase in

life-expectancy in 1985-1987 of 3 years in men and 1.5 years in women, male life

expectancy in Russia decreased by 1.1 years between 1987-1991, and by 6 years in

1992-1994. Women’s life expectancy was stable in 1987 and decreased by 3.2 years in

1992-1994. There was a 45% increase in the absolute number of deaths per year

between 1989-1994; 79% of this increase was attributable to premature deaths while

21% were due to the natural aging of the population (Shkolnikov, 1995).

However, life expectancy is composed from infant mortality rates and adult

mortality rates (Bobak, 1996). Within the countries of the Former Soviet Union, the

infant mortality rate, an accepted indicator sensitive to average deprivation, is better

than the adult mortality rate. This suggests that the changes in life expectancy are due

85

not so much to average deprivation in income alone as to other factors which affect

adult mortality, such as changes in political and social structure, levels of social stress,

and concomitant changes in personal lifestyles.

After decades of stagnation, and post-Soviet crony capitalism, Russia’s health

profile in the 1990s no longer resembled that of a developed country but was worse than

many Third World countries. In 1997, overall life expectancy in Russia, below 67 years,

was lower than forty years earlier. Mexico, even after major natural disasters and

political problems in the 1990s, had a life expectancy six years higher than Russia.

(Zdavookhraneniye Rossiyskoy Federatsii, 1999; World Development Report 1998/99;

World Health Statistics Annual, 1993, 1996). Death rates for Russian men in their early

30s were twice as high in 1994 as in 1964. For men in their early 50s, death rates were

almost two and a half times higher in 1994 (Eberstadt, 1999; Vishnevsky, 1995).

Between 1987-1994, death rates increased 1.96 times for working-age women, between

40-44 years (Notzon et al., 1998). However, infants and pensioners were not affected to

the same extent as middle-aged men and women.

In the post-Soviet decade, Russia suffered outbreaks of typhus, typhoid, cholera,

and diphtheria, even in Moscow. Tuberculosis has more than doubled since 1990, and

has been formally designated an epidemic by WHO. The impact of infectious diseases

has been primarily due to laxity in immunizations, dearth of vaccines, and migration of

infected persons through porous borders from Asian regions. Deaths from infectious

and parasitic diseases were relatively small, however, about 2 percent of the overall

age-standardized death rate. The main decrease in life expectancy, between 1987-1994,

was due to mortality from cardiovascular diseases and external causes of violence,

86

trauma, and accidents (Table 3; Vishnevsky, 1995).

MORTALITYThe two principal causes of death, cardiovascular and external factors, account

for two-thirds of the Russian overall mortality rate. The epidemic of heart disease in

Russia, for men and women alike, is higher than the death rate in the U.S. for all causes

combined. In Europe and America, cardiovascular mortality reached a high in the late

1950s and 1960s, and subsequently declined, while in Russia, high rates have continued

to climb. Even Russia’s female CVD mortality rate was roughly twice as high as male

rates from countries like Canada, Italy, and Spain in 1996 (World Health Statistics

Annual, 1996).

Mortality from external causes in post-Soviet Russia is twice what it was during

the Brezhnev and Gorbachev eras, and four times as great as the U.S. death rate

attributed to injuries and poisonings. The risk for a Russian male born in 1995 was

nearly 1 in 4 for dying from external trauma, while in Britain, the risk was about 1 in 30

(World Health Statistics Annual, 1996). Several other countries have a history of critical

drops in life expectancy: Spain in 1936-39, Western Germany in 1943-46, Japan in

1944-45, and South Korea in 1950-53. These mortality crises were transient and a direct

consequence of international or civil wars. In 1944-45, male life expectancy at birth in

Japan was under 25 years, yet three decades later, Japan reported one of the highest

male life expectancies in the world. Death rates from CVD are nearly twice as

87

TABLE 3 SELECTED CAUSES OF DEATH CONTRIBUTING TO CHANGES IN LIFE EXPECTANCY,RUSSIA, 1987-1994, BY SEX, IN YEARS;

Year Cause of deathAll causes Circulatory

diseasesAccidents/traumas

Respiratorydiseases

Othercauses

Men1987-1991 -1.42 -0.11 -1.32 0.13 -0.101992 -1.44 -0.23 -0.84 -0.08 -0.211993 -3.06 -1.09 -1.29 -0.28 -0.161994 -1.49 -0.68 -0.46 -0.09 -0.15

Women1987-1991 -0.03 0.38 -0.31 0.14 -0.27

1992 -0.53 -0.15 -0.25 0.02 -0.111993 -1.81 -0.94 -0.49 -0.13 -0.151994 -0.83 -0.53 -0.17 0.02 -0.07

high for modern Russians as they were for postwar Japanese; death rates for injury and

poisoning, over three times higher (Eberstadt, 1999).

The death rate from homicides in Russia, between 1985 and 1995, increased

directly with alcohol consumption rates (Shkolnikov, 1995). Homicides were fifty

times higher in 1993, following the precipitous drop in living standards, disorganization

of basic public services, and the crime wave which swept Russia after price

deregulation and other economic changes initiated in January, 1992.

The entrenchment of the health care crisis in Russia at the beginning of the new

millenium represents a significant challenge to public policy concerned with the

widening health inequities of the past decade. The changes in demographic, life

expectancy, and mortality patterns within Russia and across Central and Eastern Europe

have been extensively documented (WHO, 1994, 1998). There have been about 1.3 to

88

nearly 2.0 million more deaths observed than expected between 1989-1994 within

Russia, resulting in a disproportionate loss of person years for the male working age

population through year 2025 as well as a critical demographic imbalance in family

structure. The impact of the mortality crisis has been gender-specific. A greater

population ratio of women to men, diminished marriage and fertility rates, greater

divorce rates among women, more children growing up in single-parent families, and an

increasing number of lone elderly women have all been projected to affect the future

human capital potential and economic productive capacity for at least the next two

decades in Russia (Rose, 1998; Bennett, 1998).

A study of 450,000 death certificates in the city of Moscow, for the period 1993-

1995, showed that external causes and alcohol-related deaths were unequally distributed

and greater among certain population groups: the least educated, blue-collar workers,

manual laborers, and the unmarried. Increased mortality was related in the study to

causes which existed prior to 1992: an increase in alcohol abuse and malnourishment,

poverty, or the economic transition. Psychosocial stress was associated to the inability

of individuals to cope with the transition in Russia, leading to unhealthy behaviors

which increased the risk of premature mortality. Recommendations by the study were

made for more specific micro level research on mortality in Russia, which would permit

a change in the medical classification of causes of death to a classification of

psychosocial and economic determinants (Shkolnikov, 1996). However, such

explanations are left at the individual level, precluding an examination of more

fundamental social causes of premature mortality in Russia.

Social factors have been consistently associated with increases in Russia’s death

89

rates (Kennedy, et al., 1998). A drop in male life expectancy, between 1990-1994, was

found to be region specific, associated with the Robin Hood Index, an increase in the

relative income inequality of urban areas such as Moscow and St. Petersburg, increased

rate of labor turnover and crime. These factors are related to modernization,

employment insecurity, organizational and social change, mentioned above. Decreases

in life expectancy was related to death rates from accidents (1.67 years), alcohol related

causes (0.84 years), and cardiovascular disease (0.52 years) in Russia.

The relative deprivation of persons in high income areas which exhibited growth

in crime, impoverishment and stress, related to labor market changes, may explain the

greatest fraction of the fall in life expectancy among Russian males after 1992

(Walberg, 1998). Although individual lifestyle has been implicated in the deterioration

of health indicators, the apportionment of effect size between social factors and

proximal psychosocial factors, such as alcohol abuse, has not yet been made. Recently,

more detailed analyses of interregional patterns of life expectancy and cause-specific

mortality rates between the 49 oblasts of Russia were conducted. The effects of

individual lifestyle, such as level of alcohol consumption, were found to have had less

significant impact on life expectancy in comparison to sociodemographic factors, such

as abortion, marriage, and birth rates (Becker, 1998).

Reliable death certificate validation in Russia has not been routinely conducted

and access to individual level mortality and morbidity statistics is still strictly controlled

by official government agencies. It is therefore difficult to conduct independent

replication of Russian mortality research. The accuracy of recording the cause of death

in the death certificate is affected by the coding and location of death: only about 24%

90

of deaths in Russia occurred in hospitals in 1985, and over 6% were certified by

feldshers or medics (Elliott et al., 1996). In addition, coding procedures in Russia were

followed regionally, not centrally, until the mid-1990s (Shkolnikoff, 1997). Mortality

data should thus not be viewed as any more “objective” than self-reported health or

morbidity data (Table 4; Rose, 1995).

TABLE 4: Percent of self-rated health and life expectancy, by country, 1990

Country % rating health“very good”

Life expectancy(years)

Ireland 48 71.6USA 40 71.6UK 39 72.4Sweden 38 74.2Australia 36 73.2France 19 72.6Japan 9 75.9USSR 3 65.1

Although the causes of mortality differ from those of morbidity, recent

availability of non-governmental survey data has permitted more extensive

investigation into the social determinants of morbidity associated with the health crisis

in Russia, such as income inequality, social cohesion, and social capital (Bobak et al.,

1998; Rose, 1994; Rose, #304, 1998a; Rose, #303, 1998b; Bobak and Marmot, 1996;

Frijters and van Praag, 1995).

POVERTYThe general relationship between poverty and poor health has been well

documented (Klugman, 1997). Before the disintegration of the Soviet Union, and up to

1992, the greatest prevalence of poverty was among families with pensioners and single

91

heads of household (40%), and families with children (50%) (Rimashevskaya, 1997).

Between 1991 and 1994, inflation increased by over 1300%, gross industrial output and

real wages fell over 40%. In October, 1996, the highest overall poverty rates in Russia

were among children (44.5%) and the lowest among the elderly (30.6%); the latter,

however, experienced the greatest number falling into poverty between 1992 (1.2%)

and 1996 (19.0%) (Zohoori et al., 1998). Between the end of 1991 and 1992, the value

of the average pension decreased by 47% and the average household income by 39%

(McAuley, 1994). Every third Russian, about 51.7 million in the first half of 1999, was

reported by the Russian government statistics agency, Goskomstat, to live below the

poverty line, in comparison to 22% in 1998. During this same time period after the

August, 1998, financial crash, there was a 46% drop in imports and 11% drop in exports

(Kruiderink, 1999).

Income from employment in state-controlled enterprises dropped from 41.5% in

1992 to 22.9% in 1996, while cash and noncash income from home production and the

informal sector only rose from 9.4% to 22.4%. Nearly 50% of Russians, in 1993, were

participating in unofficial economic transactions to supplement their official incomes.

Earnings from official economy jobs did not meet the daily needs of over two-fifths of

Russian workers in 1992, rising to over two-thirds in 1994 (Rose, #227, 1994). Private

sector income sources increased only 7% from 1992 to 1996. About two-fifths of

working age Russians were working without being paid in 1994, which increased to

over 50% in 1996. In 1998, the Social Capital survey of Russia found that more than

three in five Russians did not routinely receive wages, pensions or economic

entitlements (Rose, #304, 1999).

92

The World Bank reported that, in 1992, 63% of poor households in Russia were

employed; the earnings of substantial numbers of employees did not rise concomitantly

with inflation and fell below the cost of the subsistence minimum (Milanovic, 1994).

The subsistence minimum was introduced in 1992 as a measure of poverty and was

based on a minimum nutritional food basket, as recommended by WHO and national

Russian agencies (Mikhalev, 1996). The subsistence minimum provided for 68% of

earnings to be allocated for food (80% among pensioners), allowing for only one-third

to one-fifth of personal earnings to be paid for medications and health services, clothes,

shoes, and other consumer goods. Persons with incomes below the subsistence

minimum were considered to be below the poverty line. In 1992-1993, the average

pension was about equal to the subsistence minimum, and average salaries, if paid at all,

in cash and not in kind, were less than three times the poverty line (McAuley, 1994).

A representative sample of Russia in 1994, estimated poverty conditions to

affect 58% of families (Zaslavskaya, 1994), and most of the poor were of working age

with low wages, not pensioners. Three-quarters of large families with three or more

children were designated as receiving earnings below the subsistence minimum in 1994;

and from 22-29% of poor households had adult dependents or disabled family members.

The Russian Longitudinal Monitoring Survey reported that older persons, living with

families, became poorer with socioeconomic changes: in September, 1992, 22.3% were

at the poverty level, in December, 1994 – 27.6% (Popkin, et al., 1996). Privatization of

the medical care system, low wages and short cash supply, inflated prices, marketisation

of housing, electricity, water, telephones and other utilities, increased the cost for the

minimum standard of living after 1992. This economic situation constrained many

93

working families, with children, older dependents, or disabled, to ration necessities like

clothes and medicines.

MORBIDITYRisk factors for poverty are also determinants of chronic conditions and

disability. Unpredictable, rapid transformations in the political economy, social stress

and anomie, depreciating social capital, impoverishment, and an institutional, policy,

and administrative vacuum characterized post-1991 Russian society and have been

implicated in poor health status. The New Russia Barometer found that severe material

deprivation and lack of life control predicted poor physical health and poor self-rated

health. (Rose, 1995; Rose, 1998). The structural factors determining mortality and the

factors determining chronic conditions and disability may be the similar, vary in

severity, social intervention, and by specific disorders. Disability may not be related to

premature mortality at all for some causes (i.e., external causes), it may be the

consequence of better disease treatment protocols, or the precursor of the normal aging

process. However, it is difficult to claim individual-level health habits as causes of

macro-level distributions of chronic conditions and disability without a multilevel

analysis which includes the distribution of social inequality.

Economic changes radically increased the inequality in the distribution of

incomes and occupations. The Gini coefficient increased by a fifth in one year from

0.27 to 0.32 at the end of 1992. This magnitude in the change of income inequality took

ten years to occur in Great Britain, between 1976-1986 (McAuley, 1994). Real per

capita income fell to 1970’s levels by 40% between 1991-1992, and the distribution of

income was less uniform, varying 11-fold between the highest and lowest deciles in

1993 (Shkolnikov, 1996). The minimum wage lost a fifth of its value by 1992, and the

94

minimum pension was worth 48.3% of the subsistence cost of living. The

impoverishment of the Russian population, after 1992, may be attributed to

organizational change such as privatization without adequate job security;

unemployment, low, and unpaid wages; increase in single-parent families; and an

increase in the number of families in which dependents exceeded wage-earners

(Rimashevskaya, 1997).

Impoverishment and economic uncertainty have been accompanied by a radical

shift in the birth/death ratio of the Russian population: increased mortality among adult

working-age males and a stunting in fertility or the desire to have children. The birth

rate influences the infant mortality rate, which together with adult mortality rates are

related to life expectancy, and the construction of generational life tables.

INCREASING INEQUALITYRussia’s health trends today also exemplify what has been termed "negative

momentum". In Spain, Western Germany, Japan, and South Korea, health improved

for several decades before their respective political crisis occurred. Rebounds in health

status were rapid after the end of the crisis period. This has not been the case for Russia,

where health conditions had been stagnating for several decades prior to the political

and socioeconomic policies of post-Soviet reforms.

At the end of Perestroika, the Russian Federation was the world’s fifth largest

nation. In the first two decades of the twenty-first century, according to United Nation

projections, Russia’s population size will drop to ninth place in the world and the GDP

to about twentieth. The political and strategic repercussions of the health crisis in Russia

may have even greater implications for future global security. Disproportionate

increases in mortality among the employed population compromise the integrity of

95

national intellectual, competitive, and creative resources. Life expectancy is strongly

associated with income and for Russia, it is a good predictor of productivity and per

capita GDP. Even with a smaller population in two decades, and an improved life

expectancy, a future Russia could attain a real GDP of only about a trillion U.S. dollars

at the current rate, assuming 2.5 percent average growth per year (World Development

Report, 1998/1999). This could derogate Russia to the sidelines of global decision-

making, after occupying the first ranks since World War II.

URBAN CONTEXTIt is important to consider the regional diversity of the Russian Federation

separately in policy and health outcome debates, given the disparity between Moscow

and other geographic areas (Table 5). Regional and urban environments vary

sufficiently to warrant separate analyses to minimize a “regression to the mean” effect

when generalizing from federal level data. Health profiles at the urban level decrease

the array of intervening individual and fundamental social factors accounting for the

inequality of health in a community (Molinari et al., 1998).

A 1992-1994 analysis of health profiles from 47 cities across Europe, which

were members of the WHO Healthy Cities Project, indicated that cities, as coherent

collective wholes, have population distributions with characteristics distinct from the

sum of its individual members. The international and community-wide implications of

the health disparity between cities has been systematically investigated by WHO with

the aim of reducing health inequalities.

To orient policy priorities to resolve health inequalities, it is important to

distinguish between identifying intervening individual risk factors (like secondary

smoke or drinking and driving) and fundamental social conditions (like the lack of

96

TABLE 5: SOCIOECONOMIC INDICATORS IN RUSSIA, MOSCOW, ST. PETERSBURG, 1990-1995RUSSIA MOSCOW ST.PETERSBURG1990 1995 1990 1995 1990 1995

N telephones/100 families 35.7 46.0 92.0 102.8 76.6 87.1N sq.m. living space/ per capita 16.4 18.1 18.0 20.0 17.7 18.8% privatized apts/all apts whichcould become privatized 0.16 36.0 0.1 41 0.01 29N private autos/1000s 58.6 92.7 70.6 153.5 56.2 119.3% employees with wages delayed,not paid in full** n/a 51 n/a 39+ n/a 37+Ave. monthly wage(1000 rbls in value of that yr)* 303 472 267.6^^ 584 n/a 443Ave. monthly wage+social payments(1000s rbls) 220.4# 531.6 325.5 726.7 212.2 506.8% with per capita income >1000rbls/mos (1000s rbls) n/a 10.5 n/a 40.1 n/a 13.9Ave. per capita income(1000s rbls/month) 215 532.9 691.1 1707.8 223.2 599.5% with total income in top quintile n/a 46.9 n/a 59.2 n/a 44.3Ave. cost of living/capita(1000s rbls/month) 1.9^ 264.1 107.8 328.5 86.9 262.2% with total income belowave. cost of living 22.4## 24.7 13.7 19.1 23.0 20.0% with per capita income 2 belowave. cost of living n/a 37.6 n/a 12.4 n/a 13.2

# data for the year 1993; ## data for 1994; ^ data initially available for the year 1992 in value of ruble forrespective year (Goskomstat Rossii, 1996). ^^ data for 1989;*excludes any social payments in 1995 (Munro, N., Regional Database\maps for web\rusregions.htm,1998);** VCIOM, Russian Nationwide Surveys, 1996, 1997 (Munro, N., 1998);+ data for the Central Region (Moscow, capital); Northwest Region (St. Petersburg, capital).

education, emergency services, or employment). Health policy which seeks to

eliminate only the intervening mechanisms or individual factors in health inequality will

not eliminate the relationship between the disease and fundamental social conditions

(Fiscella, 1997). Fundamental social causes of disease influence access to resources

such as “money, knowledge, power, prestige”, as well as interpersonal resources like

social support, social cohesion and social capital. Differential access to resources which

affect the “risk of being at risk” for health outcomes results in inequality of health status

(Link, 1995). Access to resources, as has been discussed earlier, is related to geographic

97

location.

Cities in eastern Europe and Russia have a younger population structure but a

lower life expectancy than western and northern European cities. Many more eastern

European cities reported substantially greater levels of air and water pollution, less

relative area of green space and less living space for residents than those in the west and

north. St. Petersburg, Russia, had the highest standardized mortality ratio of 14 cities in

the world. Almost two-fifths of Moscow residents complained that there were

insufficient green parks in the city in 1995, three-fifths said that there was not enough

fresh air (www.mos.ru , 1998). St. Petersburg and Minsk were the only two out of 45

cities in the world which reported not having any sport and recreational facilities for

every 1000 inhabitants. Out of 38 cities, St. Petersburg ranked fifth highest for having

single parent families; second in reporting a young median age for a first birth; and the

only city in the world to report an abortion rate over 100% of the birth rate (WHO,

Healthy Cities, 1998).

Although the health status of Moscow is mirrored in the health profile of its

sister city, St. Petersburg, Moscow has distinct characteristics: a higher income, greater

percentage of the population with higher education, lower unemployment and job

insecurity, less crime, smaller divorce rate, and the best working conditions in the

Russian Federation (Table 6).

MOSCOWA health profile of Moscow is specific to its historical and geographical context.

It is the capital of the Russian Federation and chief city in the European region of

Russia. Moscow, which celebrated its 850th anniversary in 1997, has always

98

TABLE 6: Social Indicators in Russia, Moscow, St. Petersburg, 1990-1995

RUSSIA MOSCOW ST.PETERSBURG1990 1995 1990 1995 1990 1995

Pop size (millions) 148.200 147.739 8.967^ 8.625 5.035 4.774Pop. change +2.2 -5.7 +10.2^ -8.9 +9.7^ -8.9Ave. household size 3.2 2.84# 3.1 2.74# n/a 2.79#N divorces/1000 marriages 424 619 505 565 n/a 623% high/prof educ% sec/tech educ%sec/gen educ% general primary educ

16.3^^31.3^^33.9^^15.0^^

18.433.034.312.4

33.222.027.517.3*

39.830.524.54.8

n/a 32.635.125.35.7

N crimes/100,000 1240 1860 669 1066 1144 2104N striking workers, in 1000s 99.5 489.4 0.3 0 0.1 26.5Unemployed, as % employed 8.4^^ 8.7 5.2 5.2 7.4 9.8% Unemployed With high/prof educ With sec/tech educ With sec/gen educ With general educ

13.7**27.2**37.2**19.2**

9.228.643.617.1

n/a 22.726.640.79.3

n/a 21.537.229.511.5

% workers With unhygienic cond. With heavy phys labor With unsafe cond.

32.5^^6.6^^3.3**

21.42.71.0

11.6#1.2#1.1#

11.81.50.3

n/a 17.91.10.4

(source: Goskomstat, 1996; Rossiiskii Statisticheskii Ezhegodnik; Munro, 1998).

# data for the year 1994; ^ data for 1989; pop. change calculated in comparison to1979 census (Moskva v Tzifrakh, 1990); all other pop. change given with reference toprevious year; * includes incomplete sec/gen; ^^ % officially registered unemployed in1991 when data first reported; ** data for 1992

been Russia’s seminal city, the seat of the Tsars and the bureaucratic elite, which

controlled national resources and wealth (Klugman, 1997). Moscow's unique economy,

the centralized location of Commonwealth, Federation and municipal governments,

accounts for the largest percentage of the city's revenues, and does not exemplify any

other metropolitan area in the Former Soviet Union.

The capital had a gross domestic product (GDP) more than 13% above the

Russian Federation mean and wage levels 27% above the mean in 1995 (Munro,

99

Regions web site, 1998). Comparable GDP data for Moscow in 1989-1991 were not

published during the soviet period census. It is likely that the economic status of the city

relative to other regions was not substantially different then. Moscow remains the major

cultural mecca, housing universities, museums, theatres and art institutions, as well as

almost three times the number of students in higher education the Federation.

Advantageous social conditions in the city as compared to the region or

federation, however, have not provided a cushion for residents’ health. Moscow and St.

Petersburg had the highest life expectancies in the Russian Federation in 1978-1979 but

the life expectancy of Moscow residents in 1990 was ten years below what it had been

in 1970 (State of the Environment, 1998; St. Petersburg Profile, 1995; Shkolnikov,

1996).

Goskomstat, the government statistical agency, produced several population

projections for Moscow into the next millenium. The long range prognosis was not

favorable, even with the slight decrease in mortality during 1995-1997. Goskomstat

estimated that the population of the city would fall to 7.577 million people by the year

2010. This is about one and a half million less inhabitants than in 1990, a depopulation

due to the imbalance from the birth:death ratio (Figure 11; Goskomstat, 1996a), rather

than migration. This was an alarming indicator of a compromised quality of life and

loss of productivity when, overall, more inhabitants died in Moscow than were born at

the close of Perestroika (Goskomstat, 1998).

Mortality and morbidity data were officially published in Moscow for only a

few conditions before 1990 (Vishnevsky, 1995; Goskomstat, 1996). Published

morbidity data were based on Soviet era concern with the volume and quantity of health

100

services rather than health status. Incidence of morbidity was defined by Goskomstat as

the number of persons with an initial diagnosis of illness received after a medical visit,

FIGURE 9: Number of births, deaths, and natural change in population size (per 1000persons) by district, Moscow, 1995.

classified according to the ICD-9 standard (Table 7). Officially recorded diagnoses

depended on utilization behavior, thereby underestimating the prevalence of illness in a

general community, if those in need did not seek medical care. Most studies of

morbidity have therefore been based on a limited number of surveys, which have only

been conducted on a wide scale since 1992 (Palosuo, 1995; RLMS, 1992).

101

TABLE 7: Selected causes of morbidity, Russia and the City of Moscow, 1988-1995

(initial diagnosis per 100,000 polyclinic visits)Conditions Russia Moscow

1988 1990 1995 1988 1989 1990 1990# 1997#All causes ofmorbidity

66000 65120 67880 n/a 63351 56743 122088 153900

Infectious diseasesScarlet feverDiphtheriaPertussisMeaslesViral hepatitsHIV-carriers

3420132.3*0.8*28.8*130.5*227.0*n/a

349083.60.816.912.4226.7n/a

473048.324.114.04.5167.50.72

n/a189.8#0.5#128.3#140.7#73.5#n/a

2434199.0#1.1#78.6#18.1#76.3#n/a

2233 3216148.04.462.922.484.70.20

35652223.6^^4.05^^48.4^^6.1^^97.258.5

Circulatory systemIncl.IHD+hypertension

1044

369.9*

1123

423.6

1330

391.2^

n/a 1173

476

1124

447

23829

16457

27260

n/aRespiratory system 36450 33620 29530 n/a 34297 29239 34312 60500Digestive system 2750 2720 3630 n/a 1587 1574 10006 11790Congenital anomalies 58 70 110 n/a 15 15 132 3077Traumas andpoisoning

8210 8520 8800 n/a 10229 10511 9979 n/a

Malignant Neoplasms n/a 265 279 302 355 367 2102 n/aAlcoholismNarcotic addiction

266*2.1*

1524.3

15616.9

1504.4

2303.2

1412.9

164634

n/an/a

Gonorrhea STD**

148* 259 354 n/a 211 173 203 n/a644.7

Syphilis (all forms) 9.8* 10.8 355.3 n/a 14.5 16.3 253.6 197^^Active Tuberculosis 41.2* 34.2 57.9 n/a 28.1 22.9 145.2 32.1^^

(source: Vishnevsky, 1995; Moskva v Tsifrakh, 1990; Goskomstat, 1996; Dept ofPublic health, Moscow City Council, Bureau of Medical Statistics, 21 March 1991.Mean Moscow Indicators of the Medical Dept. of Mosgorispolkom 1989-1990.Moscow, 1991).

*data for 1985; ^ data for 1994; #prevalence rates (all registered cases, including initialdiagnoses); derived from the City of Moscow Government Report about “The State ofHealth in Moscow, 1997”, Moscow, 1998; ^^cases of initial diagnosis; ** selectedSexually Transmitted Diseases (exclude gonorrhea) include chlamydia, gardnerella,urogenital herpes, uroplasmosis= 229.4 - 419.3/100,000; and trichomonas=225.4/100,000;

Cardiovascular diseases contributed the largest segment to decreasing life

expectancy in Moscow, but deaths from unexplained external and other causes

accounted for a substantial proportion of total mortality (Table 8). Of these, the

102

greatest percentage of deaths was due to the residual category of “other” and injuries of

undetermined cause, largely consisting of people found dead with head injuries where

the cause could not be specified (McKee, 1998).

TABLE 8: Cause-specific death rates in Russia, Moscow, St. Petersburg, 1989-1997

(per 100,000)RUSSIA MOSCOW ST.PETERSBURG

1990 1993 1995 1989 1993-1995*1996 1997 1990 1993Life expectancyMalesFemales

6474

5972

5872

6674

58^^72^^

n/an/a

61.373.6

65.8^74.3

62.673.6

Birth rate/1000 14.6^ 9.4 9.3 11.8 8.0^^ 7.9 7.8 10.0 7.0^^Death rate/1000 10.7^ 14.5 15. 12.4 16.9^^ 15.0 14.5 12.2^ 15.9^^Infant mortality/1000 live births 17.4 19.9 18.1 19.2 16.5 13.1 14.4 16.2^^^ 16.2All causes of death/100,000 1116.7 14446.4 1496.4 n/a 1584.4 1515.4 1444.5 1222.1 1754.8Cardiovascular/100,000 617.4 768.9 790.1 n/a 518.8** 864.4 851.3 709.1 1000.5Neoplasms/cancer100,000 194.0 206.9 202.8 n/a 256.8 260.2 258.8 268.5 288.2Accidents, trauma,poisonings/100,000 133.7 227.9 236.6 n/a 193.8 191.9 162.3 113.6 258.6Respiratory/100,000 59.3 74.5 73.9 n/a 50.2 n/a n/a 30.1 61.5Digestive/100,000 28.7 38.3 46.1 n/a n/a n/a n/a 36.4 49.2Infectious/parasitic/100,000 12.1 17.3 20.7 n/a 17.1 n/a n/a 10.1 28.3

(sources: St. Petersburg Health Profile; Goskomstat, 1996; Moscow Gorkomstat 1989-1998; Munro, 1998; Environment of Moscow, 1998; M.McKee, 1998).

^ data for 1989; ^^ data for 1995; ^^^ data for 1991; * data is a three year average, January, 1993-December, 1995, divided by three yearaverage Moscow population (8700.0 - Jan.,1994; 8625.4 - Jan.,1995; 8572.4 -Jan.,1996), mortality data for 1993-1995 was kindly shared by Prof. M. McKee, LondonSchool of Hygiene and Public Health; population size and mortality, 1989, 1996, 1997,was reported in Goskomstat Rossii, Moscow City Committee of Statistics,Administrative Raions of Moscow in 1997, 1998.**data is only available for Ischemic heart disease, which is only a subset ofcardiovascular diseases.

103

The infant mortality rate, a traditionally sensitive indicator of deprivation, was

higher in Moscow than in other regions after macro economic reforms began in 1992;

birth and death ratios were substantially greater than in the Russian Federation;

congenital anomalies at birth were greater; chronic morbidity was higher than the mean,

especially for cardiovascular diseases. More than a decade after Perestroika collapsed,

post-Soviet Moscow was the STD/HIV capital of Russia.

Moscow, as the primary urban center in the Russian Federation, clearly exhibits

the health risks of a variety of factors which accompany any megalopolis of over 8

million inhabitants. In sum, Moscow appears to have an overall health profile distinct

from the Russian Federation mean. Although the wealthiest city, it is by no means the

healthiest, especially in several urban areas (Figure 12; Goskomstat, 1996a).

FIGURE 10: Infant Mortality by urban area, Moscow, 1995

104

AREA VARIATIONBefore the Revolution of 1917, Moscow’s population was one of the fastest

growing in Europe, having increased to about 1.6 million by 1917. (Smith, 1994).

Moscow was developing factories and an indigenous nascent proletariat, as well as

migrant peasant populations searching for employment (Smith, 1989). By 1989, the

built up areas of metropolitan Moscow, including those outside the ring road, had about

13 million inhabitants.

Like London and Paris, other centers of the West which served as loci for the

rise of industrial capitalism, Moscow developed greater geographical intra-city

inequalities in living standards, social and spatial stratification at the turn of the century.

The Soviet leadership tried to eliminate the inequalities of a market distribution of

goods and services by abolishing private property. After the 1917 Revolution, the state

confiscated all housing, nationalized land, and took ownership of economic production.

Inequality and poverty in Moscow were regulated to some extent by urban planning of

housing, transportation, services, and economic infrastructure. The soviet state

developed 5 year plans for economic growth and redistribution of goods and services, a

central allocation system for such resources as housing, and central planning for

recreational, educational, and health care facilities (French, 1995).

A housing construction boom was inititated in Moscow after World War II

during the Khrushchev era. Moscow was designed as a “model communist city” with

the development in the 1950’s of the urban micro-raiyon, homogeneous and “liberated”

from problems like segregation, ethnic living quarters, concentrations of poverty and

wealth (Figure 13; Smith, 1989).

Micro-raiyons were nested within larger administrative districts. The planning of

105

the micro-raiyon included self-contained neighborhoods of about 15,000 people with

educational, cultural, health, and retail services. Within raiyons, per capita norms were

devised for services and staff, such as square meters of shopping space, educational,

health facilities, and numbers of teachers and medical personnel.

FIGURE 11: EXAMPLES OF MICRO-RAYION PLANNING FOR MOSCOW IN 1950S AND 1960S

1950s 1960s

By 1989-1991, there were 33 administrative districts for Moscow (Figure 14;

Hamilton, 1993). The contrasts of the quality of life between and within the districts

were amplified with time and a dominant mechanism of controlling perks, the allocation

of services and housing through employing institutions. Housing was also important in

the state planning of egalitarian living standards with the specification of per capita

space entitlements. Housing was a major employment benefit before 1992 and an

important measure of personal wealth with the spread of privatization and purchase of

apartments after 1992.

106

FIGURE 12: City administrative district (33), Moscow, 1989

FIGURE 13: LIVING SPACE IN SQUARE METERS PER PERSON BY DISTRICT, MOSCOW, 1989

;

107

The 1922 standard for Moscow was 9 square meters per person. This was

achieved only on the average by the 1970s after the Khrushchev constructionboom. The

implemented distribution of per capita space however in the 1980s and by 1991 was not

as egalitarian as initially planned, ranging from a high of about 20 sq.m./person to a low

of nearly 7 sq.m./person (Figure 15; Hamilton, 1993). In addition, housing was divided

into public and cooperative ownership.

Allocation and planning were the purview of governmental, industrial, and other

employing agencies of the state. Rather than ensuring equity of access to resources, the

allocation process produced substantial inequality and geographic clustering of high

quality goods and services with high social status within housing areas (Smith, 1998).

In 1989-1991, Moscow public housing, about nine-tenths of total stock, was

owned and allocated to their employees by the municipal government (75%), by local

industries (16%), and by other ministries. A significant source of social inequality

stemmed from cooperative apartments, about one-tenth of Moscow housing. Coops

were constructed for groups of people within a workplace, who had collective

ownership. A substantial cash deposit was required for coop membership, with monthly

payments over twice as large as that for public housing. Families often pooled resources

to buy cooperative apartments for their children, which could be transferred through

inheritance, unlike public housing. The economic resources necessary to make such a

purchase could only be made by the more advantaged, adding to the stratification of

housing quality (French and Hamilton, 1979; French, 1987).

Some districts developed into socially distinct areas of prestige as more or less

desirable places to live (Figure 16, French, 1995). This roughly paralleled the

108

development of rich and poor, middle-class, or suburban neighborhoods in western

cities, created largely by a market economy.

FIGURE 14: SOCIAL AREAS, MOSCOW, 1960

During the transition in 1991 from a centrally planned state to a market

economy, the contrasts of lifestyle among Moscow districts were even more amplified

primarily through the privatization of housing. Muscovites themselves described where

and why they preferred living in specific city areas in a 1984 survey (Vasil’yev and

Privalova, 1984). City districts were combined into seven larger areas on the basis of a

factor analysis of census data (Figure 17; Smith, 1989). The five factors of urban

differentiation were: 1.) proximity to functional-spatial factors (roads, transport access,

shops, cultural events, etc.); 2.) development of urban construction (characteristics of

109

streets, buildings, traditional buildings, historical establishments, etc.); 3.) availability

of employment establishments (concentration of factories, industries, economic activity,

etc.); 4.) transportations conduits and interconnections (convenience in transferring

between highways and city thoroughfares, subways, buses, suburban trains, etc.); 5.)

ecological quality (air pollution, distribution of forests, parks, playgrounds, recreational

areas). Areas were rated as better (1) or worse (0) on each factor. The survey examined

15,000 applicants for exchanging apartments. Relative preference ratings for living in

areas were calculated based on the ratio per 1000 families who wanted to move out of

an area and into another area. The central area I was preferred the most by apartment

hunters and scored the highest on the urban desirability factors. The central district is

also were the soviet and post-soviet elite work as part of the administrative, managerial,

political, military, academic, and cultural echelons. Northern and southern peripheral

areas VII were the least preferred and least desirable (Smith, 1989).

A Moscow longitudinal survey on housing and economic characteristics,

conducted annually by the Urban Institute between 1992 and 1995, indicated a growth

in relative income inequality between Moscow residents: the largest increase in income

(38%) occurred in the highest income group as opposed to only 8.7% in the lowest

income group. The Urban Institute survey estimated that the richest one-fifth of

Moscow residents owned about half the wealth, including most real property (Lee,

1996). The highest income groups were also most likely to exhibit the greatest mobility

in acquiring housing units after 1992, and to be least likely to experience overcrowding

(less than 10%). In contrast, nearly 50% of the lowest income households in Moscow

were overcrowded.

110

FIGURE 15: GROUPING OF MOSCOW CITY DISTRICTS BY RESIDENCE PREFERENCES ANDSOCIAL VALUATION FACTORS, 1984.

Almost 30% of municipal housing units and over 20% of state employer-owned,

departmental housing was privatized between 1992 and 1995. Three-quarters of all

privatized apartments by 1995 were owned by the non-pensioner, white-collar,

university educated , who were also in the highest income groups in Moscow (Lee,

1996). A 20% sample of permanent male residents of the 1916-1935 birth cohort from

Moscow showed that individual survival was best among the highest educated in 1990-

1994, as compared to the lowest educated, who also exhibited the poorest survival rates

(Shkolnikov, 1996).

Municipal maps of Moscow indicate the extent to which the highly educated

(Figure 18; Smith, 1989) were both younger and living in the more desirable areas of

111

FIGURE 16: Persons with higher education (%) by district, Moscow, 1989

FIGURE 17: Pensioners (%) by district, Moscow, 1989

112

Moscow (Figure 19; Smith, 1989) in1989. These social groups were already living in

distinct districts, adding to the spatial stratification of Moscow after 1989.

In 1992, Moscow was rezoned into 10 districts, combining the 33 administrative

districts into larger coterminous sectors. The reorganization makes it difficult to

compare Soviet with post-Soviet Moscow with any precision since the 10 districts do

not coincide with the original 33 districts. The Moscow districts, both before and after

rezoning, have had specific characteristics reported by the city census.

With the transfer of government property into private ownership, there was a

concomitant increase in the breach of the city sanitation code by industries and firms.

Most code violations involved food handling enterprises, garbage disposal, toilet

amenities, antiquated technology and machines, inadequate ventilation and crowded

working space and facilities. Moscow was the site of several large automobile factories,

heavy industrial plants, construction companies, chemical and oil refineries.

Automobiles and industrial production within the city were cited as primary causes of

pollution (Fomichev, www.mos.ru, 1999). Related to such employment sites, there was

an increase in the number of persons working under unhygienic, unsafe, and physically

difficult conditions, of which between 34%-40% were women. Moscow residents

ranked poor street sanitation, air and drinking water pollution as major ecological

problems in the city. Public surveillance programs somewhat reduced the

occupationally-related illnesses in Moscow between 1993 and 1997 (Government

Report, 1998).

In fact, several Mayoral annual reports have provided geographical mapping of

quality of life, cancer mortality (Figure 20; www.mos.ru ) and infectious disease

113

morbidity (Figure 21), indicating that there was substantial variation not only of

education, age, and housing, but also health by urban area. In all maps of Moscow, the

Kremlin is in the very center of the inner circle, from which all other areas fan out as

from a focal point.

The uneven variation of risk factors and health status was noted by a Mayor’s

Report, pointing to the existence of specific neighborhood pockets of poor health

(www.mos.ru, 1997). Most of the industrial plants in Moscow are located in the eastern,

southeastern, and southern districts, where artesian water has been polluted the most.

These are also the districts with the greatest risk of radioactive contamination, and

necrosis of vegetation, since 1982.

Morbidity rates were influenced by the location of specific health facilities

which report utilization data as illness rates. The NW district is located near one of the

most traditionally prestigious areas in Moscow, which was a high income, high status

neighborhood before and after 1991. Many specialty hospitals were concentrated in the

NW and central areas. The central district of Moscow, adjacent to the NW district, was

also the shopping nexus, attracting many transients, foreign tourists, and migrants from

other parts of the Russian Federation. Many people came to Moscow specifically to

seek specialty health care. However, the social characteristics for areas are based on

census data by residency and were not significantly affected by transient migration.

Although district health status may have been influenced by location of medical

resources, urban areas still varied by important health risks. Although the NW district

recorded the lowest cancer mortality, it had an incidence of scarlet fever which was 1.6

times higher than other areas in 1989; epidemics of paratyphus were registered in NW

114

FIGURE 18: CANCER MORTALITY (PER 100,000 PERSONS), MOSCOW, 1989

FIGURE 19: INFECTIOUS AND PARASITIC DISEASES (PER 100,000 PERSONS), MOSCOW, 1992

115

and southern district schools; drug-related HIV infections and viral hepatits B were also

recorded at above average levels in these city districts. The NE district of Moscow had

lower immunization levels among children than the city average and higher measles

incidence; viral hepatitis C infections and cancer mortality. In 1992, the SW district had

the highest overall prevalence of infectious and parasitic diseases.

CONCLUSIONIn summary, mapping of poor ecological conditions, mortality, apartment size,

housing desirability, education, age and living preferences illustrated substantial

geographic variation. The central districts were the most desirable for housing and

employment. The furthest peripheral districts were also the least desirable. Such

geographic distribution of social and environmental conditions among urban areas of

Moscow made it likely that inequality in health status of residents would also follow a

similar gradient. The inequality in the distribution of social characteristics among the

urban areas, in addition to the characteristics of individual residents, are the contextual

and compositional determinants of urban health. The following section describes the

theoretical constructs of the inequality, health and socioeconomic dimensions of the

multilevel model

116

CHAPTER 6: METHODS

The multilevel model of the city health profile, as described in Chapters 1 and 2

of this report, provides the statistical methods for examining the effects of both the

individual and social contexts on health outcomes. Although this is, to the author’s

knowledge, the first application of a multilevel model to develop a health profile, it has

become an established model of choice when data have a variety of thedoretical or

statistical hierarchically nested levels. The methodology has already been extensively

described (Bryk and Raudenbush, 1992; Bryk et al., 1996; Duncan et. al., 1996; Duncan

et. al., 1998; Kreft and de Leeuw, 1998; Goldstein, 1995).

The use of multilevel models is not limited by nested data if the question of

interest is also the multiple levels of determining effects on outcome (Diez-Roux,

1998). Exploratory analysis of the effects of social context upon individuals requires the

application of the statistical and theoretical multilevel model. If an apriori multilevel

data design is not available, a grouping of individuals into conceptually relevant levels

is appropriate in existing data if theoretical parameters justify grouping data.

Correlation between individual observations in clustered or stratified data requires the

multilevel model on statistical grounds to accurately estimate the sampling variance.

Individuals within groups can be treated as random samples from a larger distribution,

both theoretically and statistically.

For increased accuracy in estimating the standard error in multilevel data, the

number of groups is more relevant in multilevel models than the size of the groups. The

sample size at the individual level is relevant only for the individual level estimates.

The relationship between group sizes and number of groups to standard error estimates

117

havae been examined in several simulation studies (Snijders and Bosker, 2000; Hox,

2000; Kreft and de Leeuw, 1998). It has been demonstrated that at least 30 groups are

required for an adequate maximum likelihood estimate, which still underestimates the

confidence interval by about 9% as compared to 50 groups. The size of each group

should have at least 5 observations for the t-test/Wald-test, used in multilevel

regression, to be robust under conditions of small samples. The “30-30 rule” has been

suggested as a rule of thumb for multilevel designs: 30 groups with 30 observations per

group (Hox, 2000). For stable cross-level interactions to be detected, about 50 groups

with at least 20 observations per groups should be considered as adequate.

In terms of surveys which are often designed as clustered, stratified, or complex

multistage samples, not adjusting for the design effect produces misleading significance

tests. This can lead to finding spurious relationships, such as type I errors (erroneously

rejecting the null hypothesis). This is especially true of sample surveys in Russia, which

routinely use multistage sampling, such as the All-Russian Center for Public Opinion

(VCIOM) surveys, on which the New Russian Barometer is based (University of

Strathclyde, Glasgow – Russia project) and the Russian longitudinal Monitoring Survey

(RLMS), a joint US-Russia project.

Standard errors and the large nonresponse rates typically found in Russian

surveys are not systematically reported in the published literature. Exact sampling

information is rarely provided by sampling points or strata. Missing values are usually

never reported separately, although often response rates are as low as 50% of the

original multistage sampling frame.

Multilevel models have not been applied, for example, in the New Russian

118

Barometer Surveys I – V, which have different multistage sampling frames. Survey I of

1992 was restricted to a random sampling of 34 cities and towns, with stratification of

smaller areas within urban areas, and selection of individuals within households. Survey

II was stratified into 15 Russian regions, then into urban and rural areas within region,

urban populations were stratified proportionate to population in oblasts and kray

capitals, resulting in 153 primary sampling units; within sampling units, households

were selected and individuals within households. Subsequent surveys have had

similarly complex multistage sampling structures (Rose, #256, 1995).

The Russian Longitudinal Monitoring Survey, on the other hand, was

implemented as a time-series with the assistance of the U.S. Agency for International

Development and the World Bank, in conjunction with Goskomstat of Russia,

beginning in 1992. It is a nationally representative sample of 6500 households, not

individuals, clustered within 21 primary sampling units. The RLMS involved a three-

stage sample of addresses, derived from the stratification of 2335 raiyons by 10 quality

of life indicators and percentage urbanicity. This resulted in 21 primary survey sites,

within which 10 districts were selected proportionate to size, resulting in 210 secondary

sampling units. Within the latter, household addresses were selected at random and all

members of the household were interviewed.

Both the New Barometer Survey of Russia and the Russian Longitudinal

Monitoring Survey claim to compare adequately with the last Russian 1989 census. Not

one of the reports from these major surveys, to the author’s, have published any

analyses of the design effect correction of the standard error or how the hierarchical

structure of the multistage sample might have impacted on the statistical results. This

119

health profile of Moscow suggests that exploring the theoretical and statistical

application of multilevel models can be more efficient for making contributions to

social epidemiology and evidence-based health policy.

DATA COLLECTIONA random sample of Muscovites with telephones was collected, September 15-

17, 1991. Only adults 18 years and older were interviewed. The total sample size of

nearly 2000 telephone numbers (n=1991) had a completed interview rate of 81.8%

(n=1629). There was a two-stage sample selection of respondents. The first stage was a

random sample of telephone numbers proportionate to the size of the 33 Moscow

administrative districts, provided by the City of Moscow Telephone Network; the

second stage was the random selection of one respondent in each telephone household

using Kish probability tables (see Appendix). The sample is thus proportionate to the

size of the population which owns household telephones within urban districts.

All findings are restricted to the Moscow population with telephones. Personal

telephone ownership was estimated at 92 telephones per 100 Moscow families or about

4.6 telephones for every five families (Goskomstat, 1996). Only 83% of all families

lived in separate apartments in 1990, without sharing their living quarters with others.

The average family size in Moscow was 3.1 persons in 1989, which did not include an

extended family member such as a grandparent. The size of apartment households, thus,

was larger, with a greater number of adults, than family size. The number of adults

having access to telephones within apartment telephone households was also therefore

greater than 92%.

CHARACTERISTICS OF THE SAMPLE POPULATIONThere were 5.4% (n=88) of the sample who reported residing in communal

apartments. Of those in communal apartments, 11.5% were the elderly, 65 years or

120

older - only 4.2% of all the elderly sampled. A higher percentage of elderly, from 9.8%

- 18.2%, have been reported in the census as living in communal apartments in 1991,

indicating that there was an undersampling of this group. This may have been affected

by a lack of telephones among the elderly generally, and in communal apartment

dwellers specifically, where telephones may have served several families.

TABLE 9: Comparison of City of Moscow census, 1989, with weighted

sample, 1991, by marital status, sex, age, and educational level.

Moscow census Moscow samplePop. 15 yrs.old N=7,213,120 (100%) N=1991* (100%)Sex Men Women Men Women

44.9 55.1 44.2 55.8Marital statusMarriedDivorcedwidowed

71.56.82.9

54.912.118.0

72.86.92.6

59.513.015.9

Age (years)**20-2930-3940-4950-5960-6970

**16.117.513.112.5 7.8 5.0

**14.016.012.113.111.910.8

19.425.018.015.113.6 8.9

18.323.917.014.814.312.0

Educational levelHigher completeHigher incompleteSecondary/technicalSecondary/generalSecondary incomplete

26.43.619.924.815.6

22.83.220.721.314.8

37.16.323.720.012.9

33.65.226.120.714.4

* 18 years; ** % of total population of all ages=8,875,579 (see Table 10; Appendix 1)

121

The number of government and cooperative apartment dwellers for the sample,

however, was 85.8%, consistent with government statistics on the number of Moscow

families occupying separate apartments (Moskomstat, 1990). The weighted telephone

sample, when compared by age, sex, educational level (Table 9, Table 10), and urban

area, approximates the Moscow population distribution (see Appendix 1; Table 11).

DESIGN EFFECT

Several measures were taken to examine survey errors. Adjustment of selection

probability within telephone households with poststratification weights was used to

check the effect of sampling bias (Table 10). It is clear from the skew of the

unweighted sample in comparison with the census population that noncoverage error

was not random. A 10% quality control reinterview of the sample was conducted, which

reduced the effective interview rate to 81%. Accurate sampling probabilities could not

be determined because the Moscow Telephone Ministry did not provide any

information about the derivation of the sampling frame and selection probabilities.

Probability weighting could therefore not be applied in any estimation procedures.

Although the resultant sampling distribution is not a random sample of the general

Moscow population, it is a sample of a fixed universe, the city of Moscow. Fixed

models of inference may be applied.

In addition to sample frame noncoverage, interviewers over sampled women,

30-49 years old, as compared to the general Moscow population enumerated by the

census. Three age groups of men were under sampled. It is likely that procedures for

applying Kish random selection tables within telephone households, as well as varying

time of call and follow-up of "not at homes", were not followed correctly.

122

For countries with developing research infrastructures, such as Russia in 1991, it

was important that survey design and interviewer training address data quality by

decreasing specific components of error through random allocation of workloads;

increasing training and supervision of specific interviewers; and increase training of

interviewers in the random selection of respondents. Subsequent large scale surveys,

such as the RLMS, have incorporated extensive interviewer training into their

fieldwork, although no studies have been published testing data quality to my

knowledge. Further experimental research is also needed in Russia of interviewer

attitude and ideology, as well as item interaction, with respondent and interviewer traits

on responses obtained (see Appendix 3 for details of the effect of interviewer error).

TABLE 10: PERMANENT* MOSCOW POPULATION, 1989, UNWEIGHTED AND POST-STRATIFICATION WEIGHTED** SAMPLE DISTRIBUTIONS, BY AGE AND SEX, MOSCOW, 1991

_____________________________________________________________________________________

CENSUS* UNWEIGHTED SAMPLE WEIGHTED SAMPLE**____________________________________________________________________AGE Total % M% F% Total % M% F% TOTAL % M% F%______________________________________________________________________________________MOSCOW 8875579 100 44.9 55.1 1629 100 29.9 70.1 1629 100 44.1 55.9 0-14 1662459 18.7 9.6 9.1 0 015-19 539136 6.1 3.0 3.1 71 4.4 1.8 2.6 121 7.4 3.7 3.720-24 572508 6.5 3.1 3.4 99 6.1 2.3 3.8 131 8.0 3.9 4.125-29 751852 8.5 4.1 4.4 140 8.7 3.0 5.6 173 10.6 5.1 5.530-34 775131 8.7 4.2 4.5 188 11.6 4.1 7.5 175 10.7 5.1 5.635-39 700160 7.9 3.7 4.2 215 13.3 3.7 9.6 160 9.9 4.6 5.240-44 600006 6.8 3.2 3.6 177 11.0 3.7 7.3 135 8.3 3.9 4.445-49 512091 5.8 2.7 3.1 114 7.1 2.2 4.8 116 7.1 3.3 3.850-54 610956 6.9 3.1 3.8 147 9.1 2.8 6.3 137 8.4 3.8 4.655-59 528703 6.0 2.5 3.5 99 6.1 1.8 4.3 119 7.3 3.1 4.260-64 552706 6.2 2.4 3.8 121 7.5 1.9 5.6 126 7.7 3.0 4.765-69 339707 3.8 1.1 2.7 100 6.2 1.0 5.2 76 4.7 1.4 3.370-74 267809 3.1 .8 2.3 61 3.8 .8 3.0 60 3.7 1.1 2.675-79 257178 2.9 .8 2.1 39 2.4 .2 2.2 66 4.0 1.5 2.580-84 135918 1.5 .4 1.1 34 2.1 .5 1.6 30 1.8 0.4 1.4>=85 65939 .7 .2 .5 10 .6 0 .6 4 0.3 .1 0.2_____________________________________________________________________________________

*The distribution of the permanent 1989 Moscow population of all ages (Goskomstat, 1990);

** the sampling distribution was based only on adults aged 18 years and older, the referentMoscow population for weighting was calculated from all those age 15 years(N=7213121) andwas thus more than the permanent working age Moscow population; age-sex proportions werederived by dividing the number in age-sex groups by the base referent population 15 years;

123

Weights=the proportion referent population in age-sex cells divided by proportion of sampledistribution in age-sex cells; poststratification weights were used because the number of menand women in each household was available only for completed interviews (n=1629), and notfor all selected telephone numbers (n=1991).

Interviewer error varies by mode of data collection and was compounded by

several factors: a telephone sampling frame not representative of the Moscow

population, violation in the random selection of respondents within telephone

households, and inadequate supervision of field work. Interviewers, often employed as

contract piece-workers for telephone polling by local firms in Russia, were

predominantly female, college-students, who studied or worked full-time elsewhere.

Interviewers performed their jobs customarily at home without supervision, using their

own or a friend’s private telephone. In September 1991, when the Moscow Health

Profile survey was conducted, there was a dearth of centralized, monitored telephone

facilities with full-time shift workers manning individual telephones. The quality and

validity of Moscow survey information was dependent upon the subjective probity,

training, and experience of each individual interviewer.

There was scarce information on hand in Moscow about the exact distribution of

telephones, after the August Coup of 1991, a time when still no complete telephone

books were publicly available. The technology to implement Random Digit Dialing or

Computer Assisted Telephone Interviewing was neither available nor economically

feasible. Accurate and verifiable enumeration of telephones within each of 33 Moscow

administrative districts was not accessible, insofar as the government telephone network

refused to divulge details on the exact distribution of telephones.

124

TABLE 11: ESTIMATED NUMBER AND PER CAPITA TELEPHONES BY SELECTEDADMINISTRATIVE DISTRICTS OF MOSCOW, JAN., 1992

____________________________________________________________ AREA MOSCOW CENSUS HOUSEHOLD TELEPHONES # Total N % POP N=2865806 percapita

____________________________________________________________5- 143528 1.6 221068 1.527- 146822 1.6 208654 1.410- 134560 1.5 134372 1.018- 218456 2.5 189352 0.929- 260501 2.6 224849 0.912- 344241 3.9 266593 0.816- 311749 3.5 237542 0.814- 320526 3.6 185661 0.613- 384643 4.3 177116 0.59- 582524 6.6 283334 0.524- 452568 5.1 204243 0.511- 655941 7.4 259062 0.425- 97437 1.1 43015 0.432- 503699 5.7 222022 0.433- 158815 1.8 47883 0.330- 161267 1.8 8923 0.1

____________________________________________________________

Table 11 (Goskomstat, 1992) illustrates, for selected Moscow administrative

districts, that any telephone survey in 1991 sampled only a portion of the population,

biasing population estimates and claims to generalizability. The urban areas closest to

the administrative nerve center of Moscow had the largest number of telephones per

capita.

TRANSLATION

A standard international approach to translation was undertaken. Translations

and back-translations were done independently by bilingual speakers into their native

languages. This minimized problems of arriving at conceptual and technical

equivalence of the instrument in Russian. Equivalence of meaning was arrived at

through a decentering and qualitative procedure with a conference to reach consensus

on the final version of the translation. Two blind back-translations were done on the

final version, compared, and revised in the consensus conference (Werner and

125

Campbell, 1970). The author translated the initial instrument and discussed the nuances

of meanings in all consensus procedures.

Four major types of systematic bias occurred in this sample survey: mode effects

and frame noncoverage from the use of the telephone method of interview; unit

nonresponse within the sampling frame, resulting in an unequal probability of selecting

a respondent within telephone households; and item nonresponse error was related to

missing values, response effect, instrument effect and interviewer effect. Responses,

such as "don't know", "refuse to answer", blanks were often miscoding by Moscow

interviewers, who worked at home unsupervised. Instrument effect was also related to

the translation of an American questionnaire into the Russian language. Missing values

were corrected by regression imputation, using the STATA6 program.

SURVEY INSTRUMENTThe Moscow Health Profile questionnaire was based on items drawn directly

from existing sources: the California Alameda County Study on Health and Ways of

Living (Berkman and Breslow, 1983), and the U.S. NCHS Health Interview Survey

(Adams and Benson, 1991). The questionnaire of 128 items was grouped into four

outcome categories of health-related quality of life outcomes, and four social

determinants of HRQOL: life choices, life chances, civic community and demographic

indicators (see Figure 1).

MEASURES OF HRQOL OUTCOMESThe dependent variables measured HRQOL as life satisfaction, life happiness,

self-rated health, and the Alameda County Physical Health Profile.

Life satisfaction was measured by the question, “How often are you completely

satisfied with your life?”; the available responses were rarely or never, sometimes, and

often. The responses were dichotomized into often and sometimes/never. A second

126

measure of satisfaction (jobsat) was domain specific, measured by the question, “How

satisfied are you with your present job?”; responses were dichotomized as described

above. A third measure of satisfaction (unsat) was an ordinal count of all 5 satisfaction

items. Cronbach’s alpha was .42, indicating that this scale was not a reliable measure of

satisfaction.

Life happiness was measured by the question, “All in all, how happy are you

these days?”. There were three available responses which were dichotomized: not so

happy/ pretty happy and very happy. A second measure of happiness (unhap) was an

ordinal count of all 5 happiness items. Cronbach’s alpha was .46, indicating that this

scale was not a reliable measure of happiness.

Self-rated health was measured by the standard ordinal indicator, “How would

you describe your overall general health?”. The responses were also dichotomized into

excellent/good and fair/poor.

PHYSICAL HEALTH SCALE

The physical profile consisted of specific chronic diseases, symptoms,

disabilities, functional status, and energy levels. The Physical Health Profile (Table 12;

Belloc et al., 1971) was constructed from a series of questions concerning impairments,

disability, 13 specific chronic conditions, 11 specific symptoms, and three energy

levels. These dimensions were combined into a mutually exclusive seven-point ordinal

count spectrum, based on frequency of conditions within the past 12 months: from

optimum health of having 1.) high energy; to 2.) low/medium energy levels; 3.) one or

more symptoms; 4.) one chronic condition or impairment; 5.) two or more chronic

conditions or impairments; 6.) restricting activities, type or hours of work for 6 months

or longer; and 7.) severe disability, reported as difficulty with feeding, dressing,

127

mobility, or inability to work for 6 months or longer.

TABLE 12: ALAMEDA COUNTY PHYSICAL HEALTH PROFILE COMPONENTS IN 7CATEGORIES BY PERCENT WEIGHT OF CATEGORY IN OVERALL PROFILE

% inprofile

N of itemsin category

Category Components of category

7 4 disability - severe reported trouble with feeding, dressing,climbing stairs, getting outdoors, orinability to work for 6 months or longer

8 2 disability – less did not report above, but reportedchanging hours or type of work or cuttingdown on other activities for 6 months orlonger

9 15 chronic conditions did not report any disability, but reportedtwo or more impairments or chronicconditions (list of 12 chronic conditionsincluding heart trouble, asthma, hernia,and 6 impairments given) in the past 12months

19 15 did not report any disability, but reportedone chronic condition or impairment in thepast 2 months

28 11 symptomatic did not report any disability, impairmentor chronic condition but reported one ormore symptoms (list of symptomsincluding tightness in chest, pains in backgiven)

23 4 without complaints low to medium energy level – fewer than3 “high energy” answers to questions

6 4 high energy level – at least 3 “highenergy” answers

The seven categories were not weighted by the severity of loss of optimum

health. Each category was assumed to have a weight equal to all others, simplifying the

comparison between various other health measures. Although the Alameda County

studies employed a ridit scaling technique (Relative to an Identified Distribution), the

physical health profile as applied in the Moscow survey was calculated as either a

128

nominal scale or as a simple ordinal count, without making any assumptions about

equal widths between categories (Belloc et al., 1971).

To investigate a possible differential effect of social determinants on the

components of physical health, four separate ordinal count variables were constructed

from the 13 items of chronic diseases (Cchronic, alpha=.65), the 11 items of acute

symptoms (Cacute, alpha=.72), and the 4 items of energy (Hienergy, Loenergy,

alpha=.62). In addition, the physical health profile was collapsed into 4 ordinal/nominal

categories: impairments/disabilities; chronic diseases; acute symptoms; and energy

levels (Table 13). Dummy variables were constructed from each category.

TABLE 13: FREQUENCIES OF MICRO MEASURES OF HRQOL BY GENDER, MOSCOW, 1991;WEIGHTED SAMPLE (PERCENT)

Total Men WomenHEALTH-RELATED QUALITY OF LIFE OUTCOME % (n=1629) % (n=717) % (n=908)Self-rated Health – how would you describeyour overall general health?ExcellentGoodFairPoor

3.144.642.88.8

4.957.731.95.5

1.533.553.911.2

Life Satisfaction – How often are youcompletely satisfied with your life?RarelySometimesOftenJob Satisfaction – How satisfied are you withyour present job?Not satisfiedSomewhat satisfiedVery satisfied

29.748.019.8

13.250.519.7

26.251.422.5

14.058.327.6

34.647.517.9

13.656.929.4

Life Happiness - All in all, how happy are youthese days?

Not so happypretty happyvery happy

30.359.0 7.8

28.364.4 7.3

32.858.6 8.6

Alameda Physical Health ProfileDisabilityFunctional limitations2 Chronic conditions or impairments1 chronic condition, no impairments1 acute symptomslow energyhigh energy

5.1 6.733.824.118.7 9.9 1.8

3.9 5.321.126.221.518.7 3.3

5.5 7.339.223.217.5 6.1 1.1

(percentages do not sum to 100% due to missing values)

129

MEASURES OF HRQOL DETERMINANTSMICRO INDICATORS

LIFE CHOICE

Life choices were defined as patterned ways of living, including a set of

personal health practices: smoking, physical activities, alcohol intake, seeking medical

care privately (Table 14). Body mass index (BMI) was included as a choice variable,

due to the body weight component, although height has been argued to be a direct

measure of average “net nutritional status of the members of a human population during

their childhood and adolescence”, and an indirect measure of childhood welfare which

has been shown to influence mortality (Blaxter, 1990). Height was included in the

model without the BMI index. All health practice variables were dichotomized into

none or any. In addition, two control variables were investigated in the model: number

of illness days in the past month (daysill) and number of work days lost due to illness in

the past month (dayoffjob).

LIFE CHANCES

Life chances were defined at the micro level as education and occupation (Table

14). Dummy variables were created from all educational categories: primary,

incomplete secondary, general secondary, technical secondary, incomplete higher, and

higher education.

Occupation was measured by 16 response categories, specific to the division of

labor categories in Russia, and used extensively by the sociologists at Moscow

University. The responses were collapsed into 5 dummy variable categories:

▪ professional/engineers (research worker, college teacher, industrial manger,

engineer, service professional-physician, lawyers; creative professional-writer,

artist, journalist);

130

▪ white collar/service (government worker, social worker, trade and retail

worker, coop or joint-venture worker, service worker-registered nurse,

accountant, laboratory technician);

▪ manual (workers, army);

▪ not employed (students, college students, housework, temporary

unemployment);

▪ pensioners

Interactions between the 6 education and 5 occupational categories were

constructed with dummy variables.

In addition, two mobility variables were included in the model: number of times

a respondent changed jobs in the past 10 years (Nwork) and number of times a

respondent changed place of residence in the past 5 years (Naddress).

131

TABLE 14: FREQUENCIES OF MICRO MEASURES OF LIFE CHOICES AND LIFE CHANCES BYGENDER, MOSCOW, 1991; WEIGHTED SAMPLE (PERCENT)

Total Men Women(n=1629) (n=717) (n=908)

LIFE CHOICES % % %Body Mass Index (BMI=WT in kg./HT*2 in m.)

underweight =<18.6average weight= 18.6-25.0overweight =>25.1-30.0obese =>30.1

2.345.138.114.6

1.251.241.3 6.3

2.742.436.718.2

Smoking* – If you smoke regularly, how much doyou smoke in one day?

Any cigarettes 28.4 28.9 28.2Physical fitness – How often do you do any ofthese things: exercises,sports, work on dacha?

N activities – almost everyday/sometimes:

0123

35.436.717.310.6

30.733.320.515.5

37.538.215.8 8.5

Alcoholn* - How many drinks of alcohol did youhave in the past month (in gms.)?

0.25 liter.50 liter

.50-1.0 liter>1 liter

43.931.911.2 5.7 7.3

40.234.6 9.8 6.7 8.7

45.530.811.8 5.3 6.7

Alcoholx* - How many times did you drink wine,beer, or liquor in the past month?

01/week2/week3/week

44.045.4 5.9 4.7

39.648.2 6.3 6.0

45.944.2 5.7 4.1

Pay MD – If you could see any doctor you choseto see, would you pay todo so if necessary?

Yes 83.6 87.9 81.8Free MD – Do you think the government shouldprovide all medical servicefree of charge?

Yes 72.1 66.6 74.5LIFE CHANCESEducational level

PrimaryIncomplete secondary

General secondaryTechnical secondary

Incomplete higherHigher

4.0 8.619.922.4 7.836.9

2.2 7.118.318.1 8.945.3

4.8 9.720.726.1 5.233.6

Occupational levelProfessional/engineer

White collar/service workersManual workers

PensionersNot employed/Students/Housework

MobilityNaddress – During the last 5 years, how manyaddresses have you lived at?

1>1

Nwork – During the last 10 years, how manytimes did you change your place of work?

1>1

34.620.613.420.311.1

68.831.2

56.943.1

38.814.226.011.6 9.4

63.636.4

46.953.1

32.823.4 7.924.011.9

71.128.9

61.238.8

132

SOCIAL COHESION

The Alameda County instrument also included measures on social cohesion.

There were 7 questions which addressed various aspects termed social anomie. Anomie,

discussed in Chapter 2, as social deregulation and loss of integration was argued by

Durkheim to be a state of social disequilibrium and disorder, the disintegration of social

normative controls over individual conduct. The perceived social cohesion scale

(Canomie, alpha=.71) is an ordinal count of the 7 anomie items assessing perceptions

about social disorder, nostalgia, lack of faith, hopelessness, uncertainty, and confusion

about social rules. The scale (Table 15) was examined as two dimensions: a count of

normative items (Cnorm, alpha=.60) and a count of meaning items (Cmean,

alpha=.55). Because of the lower reliabilities of the separate dimensions, the complete

scale was used.

A distress variable was included as a covariate of anomie. The distress scale was

an ordinal count composed from two dimensions: mood (Cmood, alpha=.66), a count of

7 items relating to depressed feelings; and malaise (Cmalaise, alpha=.81), a count of 9

items relating to depressed physical activity. Although anomie and mood/malaise were

significantly correlated at r=.2840/.2359, this was not substantial. The intercorrelation

between mood and malaise was r=.6204, suggesting that the two components should be

combined into one distress scale when entered into a regression model.

SOCIAL SUPPORT

Social support (Cnegspo, alpha=.93) was measured by an ordinal count of 7

items concerned with the quality of the marital and family relationship (Table 15). A

second measure was the number of divorces, which was dichotomized into none and

any.

133

TABLE 15: Distribution (percent) of individual measures of civic community by gender,Moscow, 1991; weighted sample__________________________________________________________________________________________

total men women%(n=1629) %(n=717) %(n=908

CIVIC COMMUNITY________________________________________________________________________SOCIAL COHESIONSocial Anomie - meaningfulness

“old kind of friendship that lasted for a lifetime are lacking today”71.8 64.9 74.8

“things our parents stood for are going to ruin before our eyes”86.2 80.1 88.9

“everything so uncertain these days, anything could happen”91.4 85.4 94.1

“today most people really don't believe in anything”80.2 69.3 84.9

Social Anomie - norms“in a state of disorder, don’t know where you stand from one day to the next”

67.5 59.8 70.8“better off in the old days when everyone knew how he was expected to act”

60.4 50.1 64.9“changes so quickly now, trouble deciding which are right rules to follow”

61.3 48.0 67.1SOCIAL SUPPORTScale of marital quality-6 items relating to emotional support, affection,understanding, expectations in relationship; number of negative responses:

0 25.8 22.4 27.41 4.6 6.1 4.02 19.7 23.6 18.03 20.2 22.6 19.24 30.0 25.4 31.5

Number of divorces0 74.7 75.2 4.51 22.3 22.0 22.42 3.0 2.8 3.0

SOCIAL CAPITALInformal networksNumber of close friends (people that you feel at ease with, can talk to about privatematters, and can call on for help)?none 3.4 12.0 12.7Number of relatives that you feel close to?none 8.9 10.8 7.3

Formal networksGo to religious services (church, synagogue, muslim)

46.2 29.1 52.8Member of any group1-labor union 69.0 9.9 31.12-commercial or trade group 3.4 8.2 1.43-professional association? 3.2 6.7 1.74.-group concerned with children? 2.4 2.9 2.25.-social or recreational group? 2.3 3.4 1.8

______________________________________________________________________

134

SOCIAL CAPITAL

There were two dimensions of social capital: informal networks – measured by

the number of close friends (npal), number of close relatives (nfam), frequency of

contact; formal networks – measured by group membership in religious organizations

(religrp), labor unions (union), commercial or trade groups (protrade), social or

recreational groups, groups concerned with children (kids). The informal networks were

frequencies. The contact variable was dropped from the model due to multicollinearity.

The formal network measures were dichotomized into none or any. Social and

children’s groups were combined into one variable.

MACRO INDICATORS

There are two main types of macro variables which can be used in multilevel models:

global and aggregate. Global variables are contextual and measured from group level

characteristics; aggregate variables are averages of the individual observations within the

group level, and measure the mean characteristic. Several variables were not available as

contextual measures and aggregate variables were created from the sample for mean area

alcohol consumption (Xetohn) and mean area anomie levels (Xanomie).

The social composition of the areas was defined by the census simple proportion of

white collar workers, blue collar workers, educational level, family sizes, or apartment sizes.

Relative ratios were constructed of high to low characteristics within each area. In and out-

migration by social characteristic was not available for urban areas to examine direct selection

of specific persons in urban areas. However, a demographic profile of relative ratios of

persons living in high access versus low access areas indicated that some form of selection

operated to make it more likely that persons with certain social characteristics lived in

135

particular areas. All global variables were derived from the 1989 census for each

administrative area, standardized by population size of area.

AVERAGE INEQUALITY

Two factors were extracted of existing access to resources and the development

of new resources in urban areas (Table 16). The two factors had an inverse relationship

and varied between the central areas with existing access to resources and little

development and the peripheral areas with less access but greater new development in

areas (Figure 22).

FIGURE 20: Standardized factor scores of access to social resources in urbanareas by new development in urban areas, Moscow, September, 1991

(c=central areas surrounding Kremlin; p=peripheral areas surrounding central areas; seeFigure 15 for map)

Access by New Development of social resources in urban areas

std.

sco

re f

or a

cces

s

std. score for new development-1.03 0.00 1.14

-0.58

0.00

3.19

pNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNE

cEcEcEcEcEcEcEcEcEcEcEcEcEcEcEcEcE

pSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpW

cNEcNEcNEcNEcNEcNEcNEcNEcNEcNEcNEcNEcNEcNEcNEcNEcNEcNEcNEcNEcNEcNE

pNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpN

cEcEcEcEcEcEcEcEcEcEcEcEcEcEcEcEcEcE

cWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcW

pNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNEpNE

cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2

pSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpWpW

pNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNW

cWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcWcW

pSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSEpSE

cScScScScScScScScScScScScScScScScScSWcSWcSWcSWcSWcSWcSWcSWcSWcSWcSWcSWcSWcSWcSWcSWcSWcSWcSWcSWcSWcSWcSWcSWcSWcSWcSWcSWcSWcSWcSWcSWcSWcSWcSW

pEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpEpE

cScScScScScScScScScScScScScScScScScScScScScScScScScScScScScScScScScScScScScScScS

cN1cN1cN1cN1cN1cN1cN1cN1cN1cN1cN1cN1cN1cN1cN1cN1cN1cN1cN1

pSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSW pSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSz

cNEcNEcNEcNEcNEcNEcNEcNEcNE

pSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSW

cSEcSEcSEcSEcSEcSEcSEcSEcSEcSEcSEcSEcSEcSEcSEcSEcSEcSEcSEcSEcSEcSEcSEcSEcSE

pNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNW

cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2cN2

pNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpNWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSWpSW pSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSzpSz

136

The variables included area specific distributions of educational establishments,

industries, restaurants and cafeterias, retail trade shops; construction of new housing,

moving into housing of better quality; higher, secondary, elementary and preschool

facilities; cultural and recreational facilities; polyclinics, hospitals and physicians

(Table 16). It is implicit in the new development factor that there were inadequate

resources existing in the peripheral areas, leaving room for additional, new

construction. Two mean factor scores were derived for each of the 33 administrative

districts of Moscow and linked to the respondent’s district of residence.

TABLE 16: Average inequality indicators (standardized variables/1000 areapopulation); rotated factor loadings (varimax) by 33 admin. districts, Moscow, 1989

______________________________________________________________________ACCESS NEW DEVELOPMENTTO RESOURCES OF RESOURCES

--------------------------------------------------------- N govt cafe | 0.95756 N public cafe | 0.86182 N edhigh students | 0.73689 N social clubs | 0.69193 N libraries | 0.95965 N hosp beds | 0.70667 N doctors | 0.93737 N clinic visits | 0.91473 N new housing moves | 0.85689 N food stores | 0.85029 N nonfood stores | 0.81610 N edgen students | 0.70319 N preschool students| 0.81809 N new trade built | 0.75509 N new houses built | 0.73744 ______________________________________________________________________________________ eigenvalues 8.64 2.51

The 3 global macro dimensions of inequality in life chances are described in Table

17. Average and relative inequality were distinguished from area composition. The social

context of individual HRQOL was indicated by the administrative district of residence in

Moscow. Life chances are affected by individual and neighborhood social status. Urban

residential area has been used as an indirect component or gradient of social class differences

(Blaxter, 1990). Access to resources within each of the 33 administrative areas was measured

137

by standardized factor scores, obtained from 15 urban area level indicators from the 1989

City of Moscow census. Two factors of average inequality were obtained.

TABLE 17: Description of macro measures of life chances, by 33 administrative areas,Moscow census, 1989

________________________________________________________________________________________

AVERAGE INEQUALITY:

RISK FOR POVERTYFamily size - ratio of % families 5 members in urban areas to all families

living in urban areaDivorce - % divorced/1000 population in area

INEQUALITY IN URBAN DISTRIBUTION OF MATERIAL RESOURCES/1000 persons in area

N ed high - number of students in higher educational institutions N ed gen - number of students in general basic education, grades 1-10 N food - square meter space in food trade stores N nonfood - square meter space in nonfood trade stores N govt cafes - number of places in government cafeterias/ restaurants N public cafes - number of places in public cafeterias/ restaurants N hospital - number of hospital beds/10000 persons in each district N clinic visit - number of ambulatory polyclinic visits/shift/10000 persons N doctors - number of medical physicians all specialties N clubs - number of cultural clubs N libraries - number of public libraries N new moves - number of residence changes for better housing N new trade - square meter space of trade stores newly constructed N new houses - square meter space of residential houses newly constructed N preschools - number of students in kindergartens and nursery schools

RELATIVE INEQUALITY:

INEQUALITY IN URBAN AREA SOCIAL STATUS COMPOSITIONBlue collar – ratio of %blue-collar to %white-collar workers living in areaApt size–ratio of below mean(<10 sq.m./person)apts. to below the mean(10

sq.m./ person) apts. in area Education level – ratio of %loed to %hied residents in area (secondary general, incomplete

secondary, and lower vs. higher, incomplete higher, and secondary technical)______________________________________________________________________

RELATIVE INEQUALITY

Relative inequality was defined as the relative distribution of the composition of

138

social status in urban areas, which affected quality of life, and which was not under

personal control of individuals. Relative inequality was measured by three rate ratios.

The importance of housing in Soviet and post-Soviet Moscow, described in chapter 2,

was related to social status and the distribution of income as a non-monetary perk by

employing government agencies or factories. Areas in which larger apartments

predominated were higher social status areas. Education has been shown in numerous

comparative research studies to be related to health status and mortality (Kunst and

Mackenbach, 1994a; Kunst and Mackenbach, 1994b). Although basic education was

universally available in Russia, higher education was accessible through competition or

the perk system. Chapter 5 demonstrated how areas varied by the proportion of highly

educated to lower educated residents, and other social status measures such as blue or

white collar occupations (Figures 15, 16, 17).

Finally, social class has often been measured by occupational class, and

operationalized in available international classification systems as manual versus non-

manual categories (Kunst, et al., 1998). A survey of 11 European countries found that in

all countries men in manual classes had higher mortality rates than in non-manual

classes, across all age groups. This socioeconomic distinction was available from the

city of Moscow census by area. The proportion of blue-collar or manual workers to the

proportion of white-collar or non-manual workers, who resided in an area, was a third

measure of area social status. Relative inequality was thus measured by three basic

dimensions of social status: prestige in occupational position, educational skills and

level, and access to housing or material goods. The ratio indicators measured the

relative distribution of social status between areas. The effect of the distribution of

139

inequality in the social status of areas was included in the hierarchical linear regression

on individual physical health status.

MODEL ASSUMPTIONSThe use of ordinary least squares (OLS)models requires that basic assumptions

of OLS are met: linearity, normality of residuals, independent and uncorrelated

distribution of residuals, constant variance of residuals, model specification. If the

observations are not sampled independently from each other, the variances are not

homogeneous, and the residual errors are not independent, the significance tests derived

from the gaussian distribution will be inaccurate. If individual observations are

correlated due to the effect of social context within urban areas, the assumptions of the

OLS model are not met. The multilevel model includes a complex error term which

accounts for the dependent observations between micro-level individuals and contexts

or sampling units from various macro levels. This permits examining several macro-

micro relationships not possible with the basic OLS model.

To examine whether the OLS assumptions were violated by the data, the

distribution of all the variables was inspected using STATA6 software. Several

continuous variables were normalized with log (access and new development factors;

malaise and mood) and square root transformations (npal, nfam, age, daysill). The

ordinal dependent variable with 7 categories, Physical Health Profile (Physindx), was

nearly normally distributed and a boxcox normality transformation was not significant.

An ANCOVA model was estimated for Physical health. Interactions between education

and occupation, gender and age, gender and education, gender and health practices, age

and occupation were included. The full model was significant at the p<.0000 level and

had an adjusted R2 of .39. The factor score for lack of resources in urban areas, new

140

development, was significant, indicating that physical health varied by area of

residence.

In order to proceed with model diagnostics, a regression was estimated. The

STATA command, ovtest, examines model fit by testing whether any higher order

effects exist among the predictors which may have been omitted from the model. The

test was not significant and the null hypothesis that the model had no omitted variables

could not be rejected. Multicollinearity was tested with the variance inflated factor (vif).

Age was found to be strongly correlated. After deleting the untransformed age variable

from the model, as well as all the interaction terms, all vif values were below ten and

acceptable, indicating that the fitted model was not collinear.

The studentized residuals had a skewness of .203, and a kurtosis of 3.38,

indicating that the distribution was nearly normal, which was illustrated by a kernel

density plot of the distribution of residuals as compared to the normal curve (see

Appendix Residual Analysis). A normal probability plot indicated a straight line

without noticeable deviation from normality. Partial regression plots of continuous

variables in the model did not indicate any great departure from linearity.

The Cook-Weisberg test for heteroscedasticity using fitted values of the Physical

Health Profile was highly significant. A linktest for adequately specifying the dependent

variable and predictor model was alsosignificant. A graph of physical health with the

access and new development factors indicated that there were three areas (pE, pSW, and

pNW) which were significant outliers. An examination of residuals by fitted values also

indicated that the largest residuals were located in the same three areas, while the

greatest leverage was also exerted by observations in two of the same areas (pE, pNW,

141

cW, cN2). Partial regression plots of the two area factors indicate substantial

heteroscedasticity, although the regression line is almost straight about zero. A Cooks

D statistic indicated that it was not individual observations but all those in area pE and

pNW which were significant outliers with substantial leverage. Figure 22 illustrates that

these two areas have the greatest average inequality scores: low access to material

resources and lack of resources in new development. Ordinarily, one might consider

deleting the outliers and reestimating the model. Because of the conceptual importance

of the urban areas, this indicates that a multilevel model is warranted.

The basic assumptions of the OLS model were met for the “continuous”

physical health profile, with the exception of constant variance and dependent variable

specification (see Appendices). A linktest indicated that the Poisson distribution was not

appropriate but that an ordered logit distribution of physical health (Physindx) was a

better model of the data.

LOGISTIC REGRESSIONSIndividual level stepwise ordered logistic regressions were conducted for all

health-related quality of life outcome variables by using the ordinal count variables in a

full model, removing nonsignificant variables at the (p.< .10) level to reduce the model.

The full model included all dummy variables, two-level and three-level interaction

terms, covariates, and main effects. Linktests were done after each regression to check

the model specification for the dependent variable. The reduced models were on the

whole more accurate (the predictors were significant and the hatsquared were not) than

the full models (often both predictors and hatsquared were not significant).

The individual level logistic regressions were run with STATA6 software,

(STATA, 1998). The appropriateness of the proportional odds assumptions of the

142

STATA6 ordered logit regressions were tested with the –omodel- command, which

examines whether the OR is constant between ordinal categories of the dependent

variable. If the proportional odds assumptions are not met, a multinomial logit model or

ordered logit without the proportional odds assumption (gologit) are more appropriate.

Self-rated health was examined as an ordinal variable, but was rejected by the omodel

test. The distribution of self-rated health indicated that it is best dichotomized, given the

small numbers for excellent health and poor health. Life happiness, life satisfaction, and

job satisfaction were all indicated by questions which could be dichotomized.

The Alameda Physical Health Profile was examined as an ordinal variable, as

were the composite dimensions of the profile. Because of the complexity of the profile

construction, the ordinal assumptions were met for the composite dimensions but not

for the profile as a whole. The ordered logit model for the physical profile was not

rejected by the linktest, but a constant odds ratio between ordinal levels could not be

assumed. However, the separate ordinal count variables of Impairment/Disability,

Chronic Conditions, Acute Symptoms, Low Energy and High Energy, which comprised

the physical profile, were not rejected by the linktest for the ordered logit model, or by

the omodel test for meeting the proportional odds assumptions. The physical profile and

its subcomponents were therefore examined with an ordered logit model.

Basic stepwise logistic and ologit regressions were performed for all outcome

variables. Relationships between all health profile dimensions were analyzed in terms of

odds ratios (OR) and percent change in the odds ratio for one unit change in the

predictor variables. Models are reported for each outcome variable, including the

separate dimensions of physical health. Models are also reported separately for low

143

access and high new development urban areas. Only significant odds ratios are reported.

Each HRQOL model was associated with different sets of significant variables. A

multivariate regression was explored with the count variables of happiness and

satisfaction, but the results were not interpretable. This was due to the low reliability of

the scales constructed from several questions, as discussed earlier, and the rejection of

the ordinal distribution, supporting the logistic modeling of these QOL dimensions.

HIERARCHICAL LINEAR REGRESSIONPhysical health was examined in a multilevel model, including age, sex,

education, health habits, anomie and civic community variables at the individual level,

and the macro dimensions of poverty risks, average and relative inequality at the urban

level. A separate multilevel file of macro and micro dimensions was created with HLM

software (Bryk et al., 1996) to separate out the fixed and random variations at the

individual and urban area levels in a hierarchical linear regression.

The multilevel propositions explored by the hierarchical linear regression were

whether living in areas which were characterized by a greater prevalence of poverty and

inequality effects on individual HRQOL outcomes, independently of personal

demographic and psychosocial factors, including indicators of a civic community. The

forms of the proposition were schematized in Figure 3. The contextual variables

included for each of the 33 administrative districts of Moscow were described in Table

17 (family size; access to resources; ratio of lower educated residents to higher educated

residents in areas, ratio of blue collar men to white collar men among area residents,

and ratio of above the mean to below the mean apartment size within areas).

Although individual areas were not homogeneous, a strong correlation between

inequality and poverty risks differentiated between urban areas. Historically high status,

144

high access urban areas were geographically centrally located and lower status areas,

lacking resources were growing in the periphery of Moscow (see Chapter 2/3 maps).

The main multilevel outcome variable was the Alameda Physical Health Profile,

a numeric count scale. As described earlier, the physical health measure was close to

normal, only slightly kurtotic due to underlying ordinal scoring. The hierarchical linear

model is robust enough to handle such a mild kurtosis and because the Boxcox lambda

transformation for normality was not significant, the dependent variable was not

transformed. In the interests of model parsimony and interpretability, only significant

variables were selected based on the stepwise ordinal logistic regressions: age, sex,

physical fitness, social cohesion, social support, and components of social capital.

Interactions were included to investigate possible cross-level interactions between

individual and urban area characteristics.

The initial multilevel equation for the Alameda Physical Health Profile as

outcome was: Yij was: Yij = 00 + p0Xpij + 0qZqj + pqZqjXpij + u1jXpij + u0j + eij

where:

p is the number of explanatory variables X at level L1 (individuals),

q is the number of explanatory variables Z at level L2 (urban areas), and

ij is individual level L1 observation i in level L2 (urban areas) j ;

combining terms produces the following general hierarchical linear equation which

separates the fixed and random elements:

Yij =[ 00 + p0Xpij + 0qZqj + pqZqjXpij ]+[ u1jXpij + u0j + eij ]

Fixed part of equation - Random part of equation - invariate between areas residual variance between areas after fixed variables controlledand where:

145

Zqj is the cross-level interaction = value of Y-X slope at level L1

(individuals) with Z at level L2 (urban areas);

eij is the between individuals, random residual, mutually independent,

mean=0, homoscedastic, normally distributed, constant across macro units, random

effect = unexplained variability of dependent variable at micro level;

u0j is a between macro unit random residual, mutually independent,

mean=0, homoscedastic, normally distributed, random effect of intercept = unexplained

(by micro level intercept) variability of dependent variable at macro level;

u1jXpij is the random interaction between macro unit and X; u1j is a

between macro unit and micro unit random residual, independent from the individual

level residuals but correlated to the macro level residuals, random effect of slopes =

unexplained (by micro level slopes) variability of dependent variable at macro level.

The basic difference between the ordinary least squares regression model (OLS)

and the hierarchical linear model is the complex random residual term, [ u1jXpij + u0j +

eij ]. The contextual effects or unexplained variance of the outcome due to macro units

as estimated by the random residuals, u0j and u1j , are assumed to be independent

between macro units but correlated within macro units; independent of the micro level

residuals; with population mean = 0, a multivariate normal distribution, and constant

covariance (Snidjers, 2000; Bryk et al., 1996;.Kreft and De Leeuw, 1998).

Whereas an OLS multiple regression would have to produce a separate

regression for each macro unit to adequately represent the variability of the micro units

within each macro unit, and a separate regression only at the macro level to estimate

146

variance between macro units, the hierarchical linear regression can summarize this

variability in the complex residual term.

The fixed portion of the general hierarchical linear equation relates to the micro

relationships and does not vary randomly between macro units; it is the modeled

variance or explained variability in the outcome due to the micro factors and non-

randomly varying macro factors included in the equation.

The proportion of variance which is accounted for by the urban area level in the

multilevel equation may be estimated by the intraclass correlation coefficient, I ,

derived from a random effects analysis of variance (ANOVA) model: Yij = + Uj + Rij

The total variance of Yij is the sum of the between-group variance and the

within-group variance. The intraclass correlation is defined as the proportion of the

population variance between-macro units from the total variance: I = 2 / (2 + 2 ) ;

this correlation coefficient denotes the correlation between two randomly selected micro

units within the same randomly selected macro unit or the proportion of total variance

due to variability between macro units. If the macro units do not vary, as in a simple

random sample where 2 (population between-group variance) = 0, the intraclass

correlation is equal to 0, which indicates that there is no macro effect on micro unit

variance. The residual variance which is attributable to the micro level is estimated by

2 / (2 + 2 ).

The intraclass correlation, estimated in a null model without predictor variables,

provides an indication of macro unit differences and the necessity for using a multilevel

rather than a unilevel model. The intraclass correlation coefficient was calculated with

the Stata6 loneway command: a random effects way of variance (Table 18).

147

_______________________________________________________________________

TABLE 18: ANOVA FOR PHYSICAL HEALTH

One Way Analysis of Variance for Nphyslow: Nphyslow=poor physical health

Number of obs = 1629 R-squared = 0.0274

Source SS df MS F Prob > F----------------------------------------------------------------------------Between moscow areas 7889.32 32 246.54 1.41 0.0664Within moscow areas 279985.82 1596 175.43----------------------------------------------------------------------------Total 287875.14 1628 176.82748

Intraclass Asy. correlation S.E. [95% Conf. Interval]

------------------------------------------------ 0.00822 0.00723 0.00000 0.02239

Estimated SD of moscow effect 1.206172 Estimated SD within moscow 13.24499 Est. reliability of a moscow mean .2884365

________________________________________________________________________________________

Although the greatest variation in physical health may be accounted for at the

individual level, the small ICC is still a significant indication of a social context effect

on physical health due to the urban area of residence (Snijders and Boskers, 1999; Kreft

and De Leeuw, 1998).

SUMMARYThe health profile of Moscow was estimated in several stages. First the

individual level logistic model was used to describe health status in the Moscow

population with STATA6 software. Secondly, a multilevel regression was used to

describe the effect of the social context of inequality on individual physical health.

Initially, a null model was estimated to obtain the intraclass correlation coefficient with

HLM5 software for the HLM modeling, followed by a random intercept model to test

for the significance of covariates and their random or fixed variance; third, each

individual micro level variable was examined separately for significant random

variation between urban areas, controlling for gender, age, and education as covariates.

148

Nonsignificant variables were dropped from the model at each stage. Fourth, the macro

urban area variables were added to the model and finally, cross-level interactions

between micro and macro variables were fitted.

149

CHAPTER 7: RESULTS OF AN INDIVIDUAL MODEL OF HEALTH

Self-rated health and the physical health profile were modeled with stepwise

ordered logit regression (p<.10 to retain). Life satisfaction , job satisfaction, life

happiness, and self-rated health were modeled with stepwise logistic regression (p<.10

to retain). Interactions with gender, education, occupation, and civil community

dimensions were included with all main effects and covariates. In addition, interactions

between gender, educational levels and life choices were entered. Only significant

variables are reported.

A Pearson 2 for model goodness-of-fit was performed for each estimated model

using STATA6 lfit for logistic models and linktest for all dependent variable

specification. A residual analysis was performed for the physical health profile

(Appendix 2 and Appendix 3), which indicated that an ordinal model of physical

health could be more appropriate but not essential; that residuals were distributed

normally, and heterogeneity of variance was present due to urban area.

Predictive power of each logistic model was checked with a lroc graph, which

tests the number of predicted successes as a graph of specificity against 1-sensitivity

(Stata6, vol.2, 1999). Approximate likelihood-ratio test of proportionality of odds across

response categories was performed for ordinal outcomes: if Ho was rejected; the logit

coefficient b was not constant across categories, and a general ordered logit model was

more appropriate.

All dependent variables were coded to have a negative direction in outcome:

poor self-rated health, unhappiness, dissatisfaction. Ordinal variables were also counted

150

for negative outcome: the physical health profile changed from high energy levels to

impairments; increasing counts of chronic, acute and low energy levels. Self-rated

health changed from excellent, good, fair, to poor. The only positively coded outcome

was the increasing count of high energy items. All dummy variables were coded 0 for

absence and 1 for the presence of the factor. Due to small numbers in either tail of self-

rated health, the ordinal variable was dichotomized into poor/fair and good/excellent

health.

Spearman pairwise correlations between all dependent variables indicated only

slight association, except for two components of the physical health profile, chronic and

acute conditions, which were moderately intercorrelated (Table 19). The five HRQOL

outcomes had only a trivial, although significant, amount of covariation, and could

therefore be included as covariates in the regression models.

TABLE 19: SPEARMAN CORRELATION OF HRQOL OUTCOMES

PoorPhysicalHealthProfile

Numberchronicconditions

Numberacutesymptoms

NumberlowEnergyitems

NumberhighEnergyitems

Poorselfratedhealth

LifeNotsatisfied

JobNotsatisfied

LifeNothappy

Poor Phys.Health

1.00 .7954 .6106 .3772 -.3942 .4845 .1413 -.0632 .1437

N chronicconditions

(.0000) 1.00 .6353 .3778 -.3931 .4755 .1616 -.0871 .1508

N acutesymptoms

(.0000) (.0000) 1.00 .3923 -.4026 .5070 .1555 -.0195 .1211

N lowEnergy

(.0000) (.0000) (.0000) 1.00 -.9695 .4075 .1198 .0524 .1202

N highEnergy

(.0000) (.0000) (.0000) (.0000) 1.00 -.4144 -.1301 -.0467 -.1283

Poor ratedhealth

(.0000) (.0000) (.0000) (.0000) (.0000) 1.00 .1817 -.0171 .1674

Life notsatisfied

(.0000) (.0000) (.0000) (.0000) (.0000) (.0000) 1.00 .0447 .3190

Job notsatisfied

(.0108) (.0004) (.4308) (.0343) (.0598) (.4901) (.0447) 1.00 .1591

Life nothappy

(.0000) (.0000) (.0000) (.0000) (.0000) (.0000) (.0000) (.0000) 1.00

(n=1629; pairwise correlations above diagonal, significance in parentheses below diagonal)

151

This indicated that the outcomes should be considered independent and could

illustrate the differential effect patterns of average and relative inequality, psychosocial

factors, and material conditions.

Health-related quality of life, in contrast to disease-specific mortality and

morbidity rates, includes the general parameters of death, disease, disability,

discomfort, dissatisfaction (Elinson, 1980). The latter three dimensions represent the

hidden portion of the iceberg phenomena of illness, addressed by self-reported health

measures, which are independent of an exclusively clinical, functional, or professional

evaluation. Prior research has pointed out that “one disorder is not a proxy for all

disorders”, prompting the examination of several outcomes in this study (Thoits, 1995).

It is clear from comparing outcomes that the same set of factors does not affect either

psychological, social, or physical health to a similar extent.

This study demonstrates that including several dimensions of HRQOL better

isolates the set of determining factors which might be common for multiple outcomes.

The effect of different types of inequality on life happiness, satisfaction, self-rated and

physical health may be mediated differently by psychosocial determinants: life chances,

health choices, or the prevalence of a civic community .

It is often difficult to obtain conceptually and statistically intelligible models

given complex social relations. For example, the covariation of education, income, and

occupation are persistently vexing problems in specifying the effect of social status or

hierarchical relations on health. Studies which argue for the predominant effect of a

single determinant, such as income inequality, rather than joint effects on one outcome

to the exclusion of others, often suffer from misattribution by commingling variation

152

from absent or proxy factors. The macro effects of cultural, geographic and historical

factors further compound model complexity in social epidemiology.

For example, in addition to its face value, job satisfaction may be considered a

proxy variable for life control or for the contextual effect of the workplace, acting as an

outcome or determinant in specific models (Figure 23), with different conceptual

interpretations.

FIGURE 21: PROBABILITY OF QOL BY PHYSICAL HEALTH

Probability of QOL by Physical Health Profile

Pro

ba

bility

se

lf-ra

ted

he

alth

1=

po

or;2

=fa

ir;3

=g

oo

d;4

=e

xce

l

Physical health Profile

Pr(healhilo==1) Pr(healhilo==2) Pr(healhilo==3) Pr(healhilo==4)

hienergloenergacute chron=1chron>=2jobdysdisable0

.2

.4

.6

.8P

ro

ba

bility

life

ha

pp

ine

ss 1

=n

ot h

ap

py;2

=so

me

;3=

ha

pp

y;4

=m

iss

Physical health Profile

Pr(Msatisf8==1) Pr(Msatisf8==2) Pr(Msatisf8==3) Pr(Msatisf8==4)

hienergloenergacutechron=1chron>=2jobdysdisable0

.2

.4

.6

Pro

ba

bility

life

sa

tisfa

ctio

n 1

=n

ot sa

t;2

=so

me

;3=

sa

tisfie

d;4

=m

iss

Physical health Profile

Pr(Msatisf3==1) Pr(Msatisf3==2) Pr(Msatisf3==3) Pr(Msatisf3==4)

hienergloenergacutechron=1chron>=2jobdysdisable0

.2

.4

.6

Pro

ba

bility

job

sa

tisfa

ctio

n 1

=n

ot

sa

t;2

=so

me

;3=

sa

t;4

=m

iss

Physical health Profile

Pr(Msatisf9==1) Pr(Msatisf9==2) Pr(Msatisf9==3) Pr(Msatisf9==4)

hienergloenerg acute chron=1chron>=2jobdysdisable0

.2

.4

.6

153

Types and levels of inequality may affect the extent of life control and social

integration which could lead to decreased opportunities for life happiness or

satisfaction, but not poor physical health at specific threshold levels (Marmot, 2002).

Negative and positive outcomes, although often dichotomized from ordinal variables,

should not be assumed to be indicators of inverse models.

Not only are multidimensional and better measures of social status, social

differentiation and social integration needed, but so are measures of multiple outcomes

at multiple levels. Control of various pathways to multiple outcomes at multiple levels

increases the certainty germaine for policy intervention.

Figure 23 illustrates the interrelated distributions of the ordinal HRQOL

variables for each response category. The odds for a negative response, being

dissatisfied or unhappy with life, were nearly the same, whereas the likelihood for

positive responses varied: satisfaction was substantially more probable than happiness,

controlling for the same level of physical health. It appears that dissatisfaction and

unhappiness, as negative outcomes, are about as likely at the positive as well as

negative endpoints of the physical health scale. Satisfaction and happiness, as positive

outcomes, varied substantially only at the well-being or energy endpoint of physical

health, whereas there was minimal variation of QOL with disabling physical conditions.

Life satisfaction was three times as probable as happiness given better physical health

but nearly as likely as happiness when worse physical conditions prevailed.

One can be in excellent/good health, have high energy and be satisfied with life,

yet also three times less likely to be happy. On the other hand, if one has more than two

impairments or chronic conditions or is disabled, the odds for dissatisfaction or

154

unhappiness with life are nearly equal as are the odds for satisfaction or happiness.

This suggests that satisfaction and happiness are orthogonal and provide additive

information about QOL, while dissatisfaction and unhappiness may provide similar

information. Further, the negative responses are not simple inverse relationships to the

positive responses, given an ordinal distribution. The likelihood of life happiness should

be modeled separately from life unhappiness to examine the links to inequality, life

chances, or social integration, holding physical health constant. It further suggests that

the social determinants of the negative response, life unhappiness, may be similar to life

dissatisfaction. Figure 23 illustrates how job satisfaction was more likely with worse

physical health and job dissatisfaction was more likely with better physical health. This

suggests an interaction with education, occupation, or other main effect.

SELF-RATED HEALTHSelf-rated health was bimodally distributed, and highly predicted by the physical

health profile. Having one impairment or chronic condition was the sample mode, thus

the likelihood for rating oneself with either excellent/good or fair/poor health was

similar [~pr(.4)]. As a whole, over 6 out of 10 had fair self-rated health given at least

two or more chronic conditions or worse physical status and about 8 out of 10 had good

self-rated health given better physical health than reporting experiencing any acute

symptoms (Figure 23). Self-rated health, as a global indicator, did not vary as much

between different sets of social determinants of health as did specific dimensions of the

physical profile. The usefulness of global indicators are not only dependent on the scale

of the research question but also on the specificity of the model (Table 20).

The 6th New Russia Barometer survey found that lack of optimism, collapse of

informal networks, material deprivation, and loss of life control influenced poor self-

155

reported health in a national sample of Russia (Rose, 1998). This suggests that self-

rated health is a global dimension of QOL which includes main effects with residual

variance measured by stress, optimism, social dysfunction, and self-assessed status.

Self-related health was associated directly with age, gender, and educational

level. Women were 157% more likely than men to be in poor health. Older women

were at less risk for poor health than younger women, as were older residents of urban

areas undergoing new development. Job satisfaction was not significantly linked to self-

rated health, but life unhappiness and life dissatisfaction were risks for poor health.

There were several interactions between educational level and occupational

status. Pensioners were included as an occupational category by the Moscow University

question format, which could be confounded with age. Controlling for age, interactions

between gender and age, age and education, age and occupation, however, still

indicated an inverse gradient between educational level and poor health among

pensioners. Pensioners with a completed university or higher education had the greatest

odds (5.34) for poor health, whereas those with an incomplete secondary education had

the least odds (3.49). Professionals were also 41% at greater risk for poor health than

other occupational groups. The exception was women with a higher education, working

in white collar jobs, who were 33% at greater risk for poor health than others. This was

related to age, since significantly older individuals worked in white collar positions.

Union participation had a protective effect on health for the lowest educational

levels but other factors had an impact on health with increased education. The odds for

poor health among individuals with an incomplete or general secondary school

education who participated in union activities were about 55-60% less than among

156

TABLE 20 : LOGISTIC REGRESSION OF SELF-RATED HEALTH BY QOL, CIVIC COMMUNITY,AND HEALTH CHOICES, ALL MOSCOW AREAS (N=1629; LOGIT b, ODDS RATIO e^b)

__________________________________________________________________________________________ Odds of poor/fair health vs. good/excellent health: | b z P>|z| e^b % change in OR__________________________________________________________________________________________HRQOLPhys. Health | 0.60865 10.080 0.000 1.8379 83.8Life unhappy | 0.35507 2.382 0.017 1.4263 42.6Life dissatisfy | 0.24970 1.697 0.090 1.2836 28.4DEMOGRAPHICSAge | 0.03117 3.510 0.000 1.0317 3.2Female | 0.94308 1.838 0.066 2.5679 156.8Female*age | -0.02805 -2.831 0.005 0.9723 -2.8Age*urban NewDev | -0.00338 -2.300 0.021 0.9966 -0.3Educational LevelHigh ed6, complete | -0.45421 -2.396 0.017 0.6350 -36.5High ed5, incompl |Tech second ed4 |Gen. Second ed3 |Incomplete sec ed2 |Female * educationalFemale*gen sec ed3 | 0.54421 1.843 0.065 1.7232 72.3Occupational StatusProfessional | 0.34281 1.647 0.100 1.4089 40.9White collar |Manual |Pensioner |Age * OccupationAge*white collar | 0.02110 2.769 0.006 1.0213 2.1Educational level * occupational statusHigh ed6*pension | 1.67565 3.771 0.000 5.3423 434.2Tech Sec ed4*pension| 1.43819 2.482 0.013 4.2131 321.3Gen sec ed3*pension| 1.32539 2.584 0.010 3.7637 276.4Inc Sec ed2*pension| 1.24851 2.811 0.005 3.4851 248.5Female * education * occupationF*highC ed*white col| 0.84581 2.077 0.038 2.3299 133.0HEALTH CHOICESPhysical Activity | -0.85990 -3.430 0.001 0.4232 -57.7Any smoking |Any alcohol |Overweight |F * sport | 0.59302 1.995 0.046 1.8094 80.9CIVIC COMMUNITYInformal NetworksN friends contact |N family contact |Formal Networks/ group membershipsProfess/trade grp |Child/social grp |Religious grp |Union grp | 0.79572 3.899 0.000 2.2160 121.6Education * Group membershipGen sec ed3*union | -0.81195 -2.724 0.006 0.4440 -55.6Inc sec ed2*union | -0.90075 -2.632 0.008 0.4063 -59.4Occupation * Group membershipWhite collar*union | -0.76963 -2.189 0.029 0.4632 -53.7White*profess/tr grp| -0.82762 -1.675 0.094 0.4371 -56.3Manual*relig grp | 0.64497 2.047 0.041 1.9059 90.6Social Cohesiondistress | 0.13442 2.658 0.008 1.1439 14.4F * distress | 0.09468 1.654 0.098 1.0993 9.9__________________________________________________________________________________________logistic model goodness of fit test 2 =1516.5 p>.9450; area under ROC curve = 0.8463;dependent variable link test [hat coeff= p<.000][hatsq coeff= p<.229];pseudo R2 = 0.3055; N obs = 1627;--------------------------------------------------------------------------------------------------------------------------------------------------------------b = raw coefficient; z = z-score for test of b=0; P>|z| = p-value for z-test; e^b = exp(b) = odds ratio for unit increase in X; % = percent change in odds for unit increase in X;

157

other educational groups. This was also the case for white collar workers participating

in union or professional activities. No other formal or informal network participation

was significantly related to self-rated health, except religious involvement. People in

manual occupations who participating in religious activities were more likely to have

poor health than other groups, which could be due to an interaction with age.

A threshold effect of education on health which could be linked to social

participation may exist similar to that reported for income (Marmot, 2002). Group

memberships may have stronger effects at certain educational levels. In lieu of income

measures, this could be expected to obtain for occupational groups as well. It is unclear

why professionals, on the whole, were more likely than other occupations to rate

themselves in poor health. This was not due to a gender or age effect.

It is possible that the chaotic political situation during the Coup and direct

involvement of professionals as government functionaries was responsible for increased

stress. A stressful workplace setting has been repeatedly associated with poor health,

independently of social status (Siegrist, 1996). There was an indication for such an

effect in that those who felt distressed were 14% at greater risk for poor health than

those who did not experience distress. Distressed women were only 10% more likely to

report poor health than unstressed women. This would suggest that men may have borne

the brunt of the health effects linked to the political context during the survey.

Health choices, apart from any sports or physical activities, were not associated

with self-rated health. Engaging in sports was associated with 60% less odds to be in

poor health than those who did not participate in any physical activities. This

relationship held for men since physically active women had nearly twice the odds for

158

TABLE 21: LOGISTIC REGRESSION OF LIFE HAPPINESS BY QOL, CIVIC COMMUNITY, ANDHEALTH CHOICES, ALL MOSCOW AREAS (N=1629; LOGIT b, ODDS RATIO e^b)

__________________________________________________________________________________________ Odds of unhappy life vs happy life:

| b z P>|z| e^b % change__________________________________________________________________________________________HRQOL

Physical Health | 0.09733 1.624 0.104 1.1022 10.2Poor rated health | 0.39752 2.565 0.010 1.4881 48.8Life unsatisfy | 1.18167 9.030 0.000 3.2598 226.0Job unsatisfy | 0.76724 5.066 0.000 2.1538 115.4DEMOGRAPHICSurban new dev | 0.18032 2.756 0.006 1.1976 19.8Educational LevelHigh ed6, complete |High ed5, incomplete |Tech sec ed4 | 1.27283 2.708 0.007 3.5709 257.1Gen sec ed3 | 2.55253 3.015 0.003 12.8395 1184.0Incomplete sec ed2 |Female * educationalF*gen sec ed3 | -1.99007 -2.084 0.037 0.1367 -86.3Occupational StatusProfessional |White collar |Manual |Pension | 2.36032 2.150 0.032 10.5944 959.4Age * educationAge*tech sec ed4 | -0.02031 -2.173 0.030 0.9799 -2.0Age*gen sec ed3 | -0.05732 -2.522 0.012 0.9443 -5.6age * occupationAge*pension | -0.03536 -2.198 0.028 0.9653 -3.5Female * age * educationF*age*gen sec ed2 | 0.04581 1.872 0.061 1.0469 4.7Education * occupationHigh ed6 compl*pension| -2.64537 -2.634 0.008 0.0710 -92.9Female * education * occupationF*high ed6 * pension | 2.97856 2.858 0.004 19.6594 1865.9HEALTH CHOICESSport |Any smoking |Any alcohol |Overweight |CIVIC COMMUNITYInformal NetworksN friends contact |N family contact | -0.01692 -2.228 0.026 0.9832 -1.7Formal Networks/ group membershipsProfess/trade grp |Child/social grp | 0.80964 2.479 0.013 2.2471 124.7Union grp |Religious grp |Occupation * group membershipProfession*child grp | -1.54367 -2.649 0.008 0.2136 -78.6Education * group membershipTech sec ed4*union | -0.49934 -1.738 0.082 0.6069 -39.3Social CohesionF*anomie | 0.08631 2.306 0.021 1.0901 9.0distress | 0.19874 4.469 0.000 1.2199 22.0F*distress | -0.12650 -2.704 0.007 0.8812 -11.9Social SupportPoor marriage | 0.37487 8.407 0.000 1.4548 45.5F*poor marriage | -0.14594 -2.995 0.003 0.8642 -13.6__________________________________________________________________________________________

Logistic model goodness-of-fit test 2 =1568.4 p>.7772; area under ROC curve = .8163; dependentvariable linktest [hat coeff=.949… p<.000][hatsq coeff=-.0499 p<.167]; pseudo R2 = 0.2435; N obs = 1610;b = raw coefficient; z = z-score for test of b=0; P>|z| = p-value for z-test; e^b = exp(b) = odds ratio for unit increase in X; % = percent change in odds for unit increase in X

159

poor health as physically inactive women. One possible explanation might be that

physically active women could be older or engaged in strenuous occupations, while for

men sports was a leisure activity.

HAPPINESS AND SATISFACTIONAge increased the odds for happiness, while marriage was a greater risk for

dissatisfaction with life. There was an interaction between age, gender, educational

level, and occupational status. Several three-way interaction levels between gender,

education, and occupation were not significant and several cells were too small for

estimation: lower educated women in manual, white collar, or professional

occupations, and higher educated women in manual jobs (Table 21).

Happiness was related to gender and occupation: women pensioners with a

university education were 19 times at greater odds for unhappiness than men, as were

retired professional women. Again, this relationship did not hold for life satisfaction.

Men with a general secondary education, one of the lowest educational levels, were at

much greater risk for unhappiness than women, controlling for interactions with age and

education. This was not the case for dissatisfaction with life. As a whole, pensioners

were over 10 times more likely to be unhappy than non-pensioners, but retired men with

university educations were about 90% less likely, although they were at greater risk for

poor health.

Men were 86% at greater risk than women for unhappiness due to poor marital

support and distress but less likely to report anomie. This suggests that distress and

anomie are not confounding concepts. Psychological distress should be considered as a

covariate with any measure of social cohesion.

Social networks in the form of supportive family and social participation were

160

positively associated with happiness. The odds were greater among those with only a

technical secondary education if they were union members, which was also evident in

better self-rated health. It is unclear why participating in a social or child related group

would increase the risk for unhappiness among all groups except professionals. The

same effect was not found for satisfaction.

In contrast to happiness, both men and women were at risk for dissatisfaction

with life because of poor marital relationships. Dissatisfaction with work was not

associated with lack of social cohesion or marital support (Table 22).

Pensioners with lowest and highest education had greater odds for a positive

association with life satisfaction, as well as job satisfaction. Pensioner with incomplete

higher and technical secondary education were dissatisfied with life. This may have

been due to the interaction between pension levels and occupational category, which

was most often dependent on educational credentials. There was an interaction between

professionals and education. Those with a lower general secondary education but

working in a professional capacity were more likely to be dissatisfied, as were those

with a university education. But those in professional occupations as a whole were 63%

less likely to be dissatisfied. White collar workers with higher education and women

white collar workers were more likely to be satisfied, but white collar men were 50%

more likely to be dissatisfied.

Finally, in distinction to health and happiness, union membership was associated

with life dissatisfaction among those with a mid-level educational attainment.

Participation in sports was the only health choice to be positively associated

161

TABLE 22 : LOGISTIC REGRESSION OF LIFE SATISFACTION BY QOL, CIVIC COMMUNITY,AND HEALTH CHOICES, ALL MOSCOW AREAS (N=1629; LOGIT b, ODDS RATIO e^b)

__________________________________________________________________________________________ Odds of dissatisfied vs. satisfied with life:

| b z P>|z| e^b % change__________________________________________________________________________________________HRQOLPoor rated health | 0.30493 2.232 0.026 1.3565 35.7Life unhappy | 1.10712 8.565 0.000 3.0256 202.6DEMOGRAPHICSUrban access | 0.11130 1.910 0.056 1.1177 11.8married | 0.30632 1.807 0.071 1.3584 35.8Educational LevelHigh ed6, complete | 1.35756 4.705 0.000 3.8867 288.7High ed5, incomplete |Tech sec ed4 |Gen sec ed3 |Incomplete sec ed2 |age * educationage*gen sec ed3 | 0.01634 2.545 0.011 1.0165 1.6age*incompl sec ed2 | 0.01491 2.304 0.021 1.0150 1.5Occupational Statusprofessional | -0.98371 -2.292 0.022 0.3739 -62.6white collar | 0.83136 2.385 0.017 2.2964 129.6pension | 1.04713 3.484 0.000 2.8494 184.9manual |Age * occupationAge*profession | 0.02224 2.569 0.010 1.0225 2.2Female * occupationF*white collar | -0.69631 -2.015 0.044 0.4984 -50.2Educational level * occupational statusHigh ed6*white col | -1.09048 -2.804 0.005 0.3361 -66.4High ed6*pension | -0.86890 -2.077 0.038 0.4194 -58.1Gen sec ed3*profession| 0.85945 1.818 0.069 2.3619 136.2Gen sec ed3*pension | -1.26593 -2.383 0.017 0.2820 -71.8Incom sec ed2*pension| -1.14019 -2.115 0.034 0.3198 -68.0HEALTH CHOICESAny smoking |Any alcohol |Overweight |Physical activity | -0.21716 -1.742 0.081 0.8048 -19.5CIVIC COMMUNITYInformal NetworksN family contact |N friends contact | -0.01478 -2.073 0.038 0.9853 -1.5Formal Networks/ group membershipsProfess/trade grp |Child/social grp |Union grp |Religious grp |education al level * formal network/group membershiphigh ed5,incompl*union| 0.66942 1.797 0.072 1.9531 95.3tech sec ed4*union | 0.96334 3.809 0.000 2.6204 162.0Social CohesionN anomie | 0.15636 4.071 0.000 1.1692 16.9N distress | 0.04933 2.761 0.006 1.0506 5.1Social SupportPoor Marriage | 0.11741 3.958 0.000 1.1246 12.5__________________________________________________________________________________________

Logistic model goodness-of-fit test 2 =1597.3 p>.5910; area under ROC curve = .7579;dependent variable linktest [hat coeff=.8649… p<.000][hatsq coeff=-.1300 p<.033];pseudo R2 = 0.1435; N obs = 1627; b = raw coefficient; z = z-score for test of b=0; P>|z| = p-value for z-test; e^b = exp(b) = odds ratio for unit increase in X; % = percent change in odds for unit increase in X;

`

162

with life satisfaction and self-rated health.

Job dissatisfaction was substantially affected by a gender, education, and

occupation interaction. Women were 91 times at greater risk for being dissatisfied with

their jobs than men. Older women and younger highly educated women were more

likely to be satisfied. Educated women with pensions were also satisfied with their jobs,

suggesting that they were deriving income and satisfaction from continued employment

after age 50. This is consistent with the Soviet era practice of permitting full pensions

from one position to be kept after moving to another position, especially among senior

management.

Although university educated men had greater odds for dissatisfaction with work

than women, white collar higher educated women and women employed in manual

labor with a secondary education were from 600% - 700% more likely than men in to be

dissatisfied with their jobs.

Women engaging in sports had greater odds to be satisfied with their jobs,

although they were also more likely to be in poor health. This is consistent with the

organization of Moscow health services around restricted employer-based clinics, in

addition to lower quality neighborhood polyclinics available to everyone. A further

indication of the effect of job based health facilities is the greater odds for

dissatisfaction with a willingness to pay out of pocket for medical services.

The effect of formal networks on job satisfaction was also related to three-way

interactions. Participating in professional or trade organizations was associated with

greater odds for job satisfaction, especially among women, having a general secondary

education, as well as those with professional occupations. Women who participated in

163

TABLE 23: LOGISTIC REGRESSION OF JOB SATISFACTION BY QOL, CIVIC COMMUNITY, ANDHEALTH CHOICES, IN ALL MOSCOW AREAS (N=1629; LOGIT b, ODDS RATIO e^b)

__________________________________________________________________________________________ Odds of dissatisfied vs. satisfied with job:

| b z P>|z| e^b % change__________________________________________________________________________________________HRQOLLife unhappy | 0.79712 5.546 0.000 2.2191 121.9DEMOGRAPHICSage | 0.04504 2.917 0.004 1.0461 4.6F * age | -0.08788 -4.457 0.000 0.9159 -8.4female | 4.52008 4.261 0.000 91.8431 9084.3Educational LevelHigh ed6, complete | 3.35140 3.853 0.000 28.5426 2754.3High ed5, incomplete | 2.27114 3.474 0.001 9.6904 869.0General second ed3 | 2.13510 2.110 0.035 8.4579 745.8Female * educational levelF*high ed6, complete | -4.19905 -3.394 0.001 0.0150 -98.5F*tech second ed4 | -2.58028 -2.520 0.012 0.0758 -92.4F*gen second ed3 | -4.48830 -3.253 0.001 0.0112 -98.9age * educationage*high ed6, complete| -0.07634 -4.291 0.000 0.9265 -7.3age*gen sec ed3 | -0.05650 -2.440 0.015 0.9451 -5.5age*incomplete sec ed2|-0.01593 -2.565 0.010 0.9842 -1.6female * age * educationF*age*high ed6, compl | 0.06714 2.954 0.003 1.0694 6.9F*age*tech sec ed4 | 0.02816 1.599 0.110 1.0286 2.9F*age*gen sec ed3 | 0.09492 3.185 0.001 1.0996 10.0Occupational StatusProfessional |White collar | 0.82631 2.721 0.007 2.2849 128.5Manual | 2.47169 2.723 0.006 11.8425 1084.2Pension | -0.91853 -2.013 0.044 0.3991 -60.1age * occupationage * manual | -0.03554 -1.811 0.070 0.9651 -3.5female * occupationF * professional | 1.33485 4.341 0.000 3.7994 279.9F * pension | 1.93627 3.417 0.001 6.9329 593.3educational level * occupational statushigh ed6*white collar | -1.38456 -2.625 0.009 0.2504 -75.0high ed5*white collar | -2.08648 -2.268 0.023 0.1241 -87.6gen sec ed3*white coll| -1.05453 -2.203 0.028 0.3484 -65.2high ed5*professional | -2.27820 -2.787 0.005 0.1025 -89.8tech sec ed4*pension | -1.27700 -2.501 0.012 0.2789 -72.1incom sec ed2*manual | 2.17342 3.116 0.002 8.7883 778.8Female * education * occupationF*highed6*white collar| 2.00740 3.117 0.002 7.4439 644.4F*high ed5*pension | -3.04751 -2.820 0.005 0.0475 -95.3F*tech sec ed4*manual | 2.08178 1.842 0.065 8.0187 701.9F*gen sec ed3*profess | -1.53787 -2.403 0.016 0.2148 -78.5F*gen sec ed3*pension | -1.73731 -2.377 0.017 0.1760 -82.4HEALTH CHOICESPay private MD | 0.38756 2.382 0.017 1.4734 47.3F*physical activity | -0.34181 -2.195 0.028 0.7105 -29.0CIVIC COMMUNITYfemale * formal network/group membershipF * union | -0.64787 -2.365 0.018 0.5232 -47.7F * profess/trade grp | -0.81873 -2.305 0.021 0.4410 -55.9educational level * formal network/group membershiphigh ed6,incom* union | 0.56170 1.928 0.054 1.7537 75.4tech sec ed4 * union | 0.84667 2.490 0.013 2.3319 133.2gen sec ed3 * union | 1.36718 3.541 0.000 3.9243 292.4gen sec ed3*prof/trade| -2.05136 -3.159 0.002 0.1286 -87.1incompl sec ed2*relig | 0.79740 1.844 0.065 2.2198 122.0occupational status * formal network/group membershipprofessional*relig grp| -0.43782 -2.118 0.034 0.6454 -35.5manual * union grp | -0.95121 -1.641 0.101 0.3863 -61.4Social CohesionDistress | 0.07666 3.985 0.000 1.0797 8.0

164

__________________________________________________________________________________________

Logistic model goodness-of-fit test 2 =1568.4 p>.7772area under ROC curve = .8163dependent variable linktest [hat coeff=1.062… p<.000][hatsq coeff=-.039 p<.477];pseudo R2 = 0.1210; Observations = 1627b = raw coefficient; z = z-score for test of b=0; P>|z| = p-value for z-test;e^b = exp(b) = odds ratio for unit increase in X;% = percent change in odds for unit increase in X__________________________________________________________________________________________

union were more satisfied with their jobs than men. But unions did not ensure job

satisfaction among the educated at the secondary or higher level (Table 23).

EFFECTS OF INEQUALITY ON QOLInequality of resource development between urban areas was associated with

unhappiness among area residents. There was almost a 19% greater likelihood for life

unhappiness among residents in new development areas, which lacked social and

material resources, than among other area residents. Life happiness was not related to

access as strongly to lack of resources. Life dissatisfaction was related to living in areas

with greater access to material resources. Experiencing lack of fulfillment or rising

frustrated expectations may underlie these seemingly disparate findings.

Low access and high development factor scores were two different indicators of

average inequality within urban areas. Although the two types of average inequality

were almost orthogonal, the composite factors were independent of each other and

varied among urban areas. Some areas could have high access/low development, high

development/low access, or high access to resources and high new development of

resources. The particular urban area of residence accounted for the variation in QOL

outcomes being associated with average inequality controlling for demographic and

psychosocial factors.

In order to investigate the effect of average inequality, logistic regressions were

calculated for each outcome within low access and high development areas (Table 24).

165

Seemingly unrelated estimation was used to check the significant differences between

the model for low access areas and high development areas with STATA6 –unest-

command (Stata6 v. 4, STB52, sg120). The regression coefficients of all eight logistic

regressions were significantly different within and between dependent variables, as well

as between areas. This was a further indication that there was a significant variation

between urban areas which could be examined in a multilevel model, even though the

between group variation (ICC) was small.

There were distinct patterns of determinants for health, happiness and

satisfaction, dependent upon the type of average inequality in the urban areas.

Informal networks, social support, and distress did not differ markedly by

inequality in areas. Life chances, social participation, health choices, and social

status/gender interactions varied significantly by type of inequality. HRQOL within

urban areas varied by social status and inequality: the low access areas had the worst

health status and high development areas had greater likelihood for job dissatisfaction

among some groups and life unhappiness among other groups. The variation of job

dissatisfaction among social status indicated the importance of the workplace context

and life control as an indirect effect on health through life happiness.

Generally, average inequality was associated with a discernible hierarchical

effect of social status on QOL. The odds of poor self-rated health varied by education

and occupation, as well as average inequality. Although not consistent among the eight

outcome groups because of complex interaction terms, higher education had better odds

for good health, happiness and job satisfaction than lower education.

166

TABLE 24: LOGISTIC REGRESSION OF HEALTH, HAPPINESS, AND SATISFACTION BYDEMOGRAPHICS AND TYPE OF AVERAGE INEQUALITY IN URBAN AREA (LOW ACCESS TORESOURCES; HIGH DEVELOPMENT OF NEW RESOURCES), ODDS RATIOS (|Z-STATISTIC|)

_______________________________________________________________________SELF-RATED HEALTH LIFE LIFE JOB

FAIR/POOR UNHAPPY UNSATISFIED UNSATISFIED____Low High Low High Low High Low Highaccess develop access develop access develop access develop

__________________________________________________________________________________________

HRQOLLow rated health 1.506 1.421 1.491

(2.29)* (2.15)*(1.81)^life unhappy 1.648 1.601 3.259 3.612 2.182 1.498

(3.05)* (2.08)* (7.84)** (6.48)** (5.83)** (1.90)^life unsatisfy 1.385 3.360 3.536

(1.88)^ (7.84)** (6.31)**job unsatisfy 2.153

(4.28)**physical health 1.067 1.059 1.013 1.019profile/n (9.19)**(5.78)** (1.96)*(2.28)*

URBAN AREAResources Access 3.715

(2.145)*Age*Access 0.975 1.004

(2.19)*(2.02)*Resources New Dev 1.25

(2.50)*Age*new dev 0.997 0.995 0.997

(1.69)^(2.18)* (1.87)^DEMOGRAPHICSfemale=1 5.793 12.815

(4.35)**(4.97)**age/yr 0.977 0.962 1.019 1.042

(3.46)**(3.51)** (2.94)**(4.22)**married=1 1.637

(2.47)*fem * age 1.025 0.969 0.975 0.957

(2.46)* (3.53)** (4.60)**(3.47)**EducationHigh ed6,compl 0.200 0.212 1.828 4.491 1.77 0.561

(2.50)*(3.32)** (2.00)* (4.74)**(2.59)* (2.21)*Hi ed5,incompl 0.069 2.946 3.91

(2.90)* (1.79)^(2.39)*Tech sec ed4 2.136 2.218

(2.53)* (2.23)*Gen sec ed3 2.966 3.579 0.094

(3.05)** (2.96)* (3.03)**incompl sec ed2 2.597

(2.31)*OccupationProfessional 2.894 5.369 4.677

(2.36)*(3.32)** (4.64)**white collar 3.16 3.686

(3.12)* (3.13)*manual 1.788 2.942

2.61)* (4.14)**pension 7.154 2.292

(3.85)** (2.33)*MobilityAny job change 0.586 0.634

(2.52)* (1.97)*Any housing change__________________________________________________________________________________________

167

The odds varied significantly by average inequality. The odds for life dissatisfaction

were greater for higher educated residents of low access areas. Involvement in union

activities increased the odds for good health for all educational groups in low access

areas but increased the odds for life dissatisfaction among union members with mid-

level education. Life happiness and job dissatisfaction were not markedly affected.

The effect of occupational status on QOL varied by inequality of area and was

moderated by education and social capital. Pensioners of all educational levels were at

greater risk for poor health if they lived in low access areas but not high development

areas. Professionals and university educated white collar workers had greater odds for

poor health in new development areas. Manual workers with incomplete secondary

school in low access areas were the highest risk group for poor physical health among

those working, but they were not the unhappiest group.

Occupational status and education were not linearly related in Moscow, thus the

effect on outcomes was not consistent over type of inequality. The odds for life

unhappiness were affected by a strong hierarchical relation between an interaction of

gender with occupation and inequality in new development areas. Although women

manual workers had 19 times greater odds than men and 16 times greater odds than

professional women for unhappiness, there was no similar effect on self-reported health,

life satisfaction, or job satisfaction.

Participation in formal networks by occupational status varied by type of

inequality and QOL outcome measure. A readily discernible and consistent pattern

could not be estimated due to the complexity of third or even fourth order interaction

patterns, some of which had null cells given the sample size (age * gender * education *

168

occupation * group membership) (Tables 25/26).

TABLE 25: LOGISTIC REGRESSION OF HEALTH, HAPPINESS, AND SATISFACTION BY CIVICCOMMUNITY, HEALTH CHOICES AND TYPE OF AVERAGE INEQUALITY IN URBAN AREA ,(LOW ACCESS TO RESOURCES; HIGH DEVELOPMENT OF NEW RESOURCES), ODDS RATIOS

(|Z-STATISTIC|)_______________________________________________________________________

SELF-RATED HEALTH LIFE LIFE JOBFAIR/POOR UNHAPPY UNSATISFIED UNSATISFIED____Low High Low High Low High Low Highaccess develop access develop access develop access develop

__________________________________________________________________________________________

CIVIC COMMUNITY

Informal NetworksN friends 0.980 0.977 0.980

(2.38)*(2.07)* (2.37)*N family 0.984

(1.81)^Formal Networks/ Group MembershipUnion group 8.79 2.364 2.106

(3.46)** (2.97)*(2.12)*child/social grp 2.392

(1.75)^religious grp 1.57 4.526 0.653

(1.80)^ (3.59)** (2.31)*Profess/trade grp 0.068 0.476

(4.14)**(1.78)^Social SupportPoor marriage 1.532 1.479 1.142 1.079

(7.63)**(7.31)** (3.86)**(2.16)*F * poor marriage 0.827

(3.50)**

Social CohesionAnomie/n 1.158

(3.17)**Distress/n 1.224 1.278 1.070 1.263 1.047 1.059 1.064 1.134

(7.26)**(6.00)** (2.96)**(3.03)** (2.23)*(1.98)* (2.93)*(3.82)**F * anomie 1.119 1.267

(2.52)* (3.68)**

HEALTH CHOICESPay private MD 1.578

(2.58)*Physical Activity 0.693 0.382 0.547

(2.31)*(3.43)** (2.85)**Any smoking 2.921

(2.51)*female * health choicesF * obesity 0.579

(2.02)*F * smoke 0.247

(2.78)*F * active 2.005 0.537 1.822 0.474

(2.61)* (2.75)* (2.94)** (2.84)*

______________________________________________________________________

169

TABLE 26: INTERACTIONS IN LOGISTIC REGRESSION OF HEALTH, HAPPINESS, ANDSATISFACTION BY TYPE OF AVERAGE INEQUALITY IN URBAN AREA, (LOW ACCESS TO

RESOURCES; HIGH DEVELOPMENT OF NEW RESOURCES), ODDS RATIOS (|Z-STATISTIC|)_______________________________________________________________________

SELF-RATED HEALTH LIFE LIFE JOBFAIR/POOR UNHAPPY UNSATISFIED UNSATISFIED____Low High Low High Low High Low Highaccess develop access develop access develop access develop

__________________________________________________________________________________________

Female * educational statusF*high ed6 2.254

(3.24)**F*high ed5 7.593 0.286

(2.44)* (1.98)*F*tech sec ed4 0.517

(2.12)*F*gen sec ed3 3.695 0.393 0.389

(3.38)** (2.56)* (2.79)*Female * occupational statusF*professional 3.100 0.414

(2.14)* (2.90)*F*white collar 4.939 0.312 0.342

(2.43)* (2.72)* (2.16)*F*manual 7.115 2.68

(2.75)** (2.07)*F*pension 19.532

(3.54)**female * formal networks - group membershipF*union 0.467 0.373 0.401

(2.49)*(2.49)* (2.69)*F*child/social grp 2.25

(2.17)*Educational level * occupational statusHigh ed6*professional 2.08

(2.28)*Techseced4*prof 0.147 2.508

(2.85)* (1.98)*Gen seced3*prof 2.52

(2.00)*High ed6*whitecollar 9.96 0.335 0.237

(3.54)** (2.65)* (2.41)*Techseced4*white 0.301

(2.45)*Gen sec ed3*white 0.342 0.276

(2.21)* (2.41)*Inc seced2*manual 3.461 0.289

(1.96)* (2.27)*High ed6*pension 5.133 0.238

(2.85)* (3.16)*High ed5*pension 11.37

(2.28)*Tech sec ed4*pens 4.988 0.134

(2.24)* (3.06)*Gen sec ed3*pens 3.799

(1.96)*Inc sec ed2*pens 4.522 4.129

((2.95)* (2.91)*

170

(CONTINUED) INTERACTIONS IN LOGISTIC REGRESSION OF HEALTH, HAPPINESS, ANDSATISFACTION BY TYPE OF AVERAGE INEQUALITY IN URBAN AREA, (LOW ACCESS TO

RESOURCES; HIGH DEVELOPMENT OF NEW RESOURCES), ODDS RATIOS (|Z-STATISTIC|)_______________________________________________________________________

SELF-RATED HEALTH LIFE LIFE JOBFAIR/POOR UNHAPPY UNSATISFIED UNSATISFIED____Low High Low High Low High Low Highaccess develop access develop access develop access develop

__________________________________________________________________________________________

Educational level * formal networks - group membershipHigh ed6*prof/trade grp 8.553

(3.03)*High ed6*union 0.173

(2.48)*High ed6*religious 0.172

(3.62)**High ed5*union 0.327 3.637

(2.46)* (2.27)*Tech seced4*union 0.292 2.740

(1.96)* (3.18)**Tech seced4*relig 0.185 0.363

(3.57)** (3.07)*Gen seced3*relig 0.202

(3.05)*Gen seced3*union 0.083 1.978 4.438

(3.64)** (2.36)*(3.12)*Incomseced2*union 0.0422

(3.95)**Incomseced2*relig grp 0.158

(3.07)*Occupational status * formal networks\ group membershipProfession*protradegrp 0.262

(2.10)*profession*relig grp 0.580

(2.12)*profession*union grp 2.279

(2.56)*whitecol*prof/trade grp 4.74

(2.42)*whitecol*social/child grp 0.262

(1.96)*whitecol*relig grp 0.481 2.49

(2.19)* (3.02)**whitecol*union grp 2.700 5.164 3.409

(2.74)** (3.55)** (3.29)**manual*prof/trad 5.371 14.65 0.130

(1.79)^(1.88)^ (2.39)*manual*soc/child grp 0.119

(1.98)*manual*union grp 2.827

(3.09)**__________________________________________________________________________________________Observations= 1236 688 1215 684 1229 681 1221 681R2sq.= 0.3176; 0.3635; 0.2495; 0.1728; 0.1616; 0.1728; 0.1056; 0.1661;Lfit=2= 1175.6; 597.7; 1190.6; 651.17; 1215.4; 668.00; 1220.8; 629.4;

(p<.66); (p<.49); (p<.42); (p<.49); (p<.44); (p<.49); (p<.25); (p<.16)Lroc= 0.8542; 0.8745; 0.8236; 0.8193; 0.7683; 0.7792; 0.7154; 0.7666;

Absolute value of z-statistics in parentheses;* p<.05; ** p<.001; ^ p<.051-.09

171

PHYSICAL HEALTH PROFILEThe social determinants of each health outcome, whether QOL, mortality,

disability, morbidity, self-rated health or functional limitations, has a specific set of

unique predictors which may not be generalized to other outcomes, although a subset of

predictors may be common for a variety of outcomes. It is therefore important to

considering the structure of physical dimensions as a whole, as well as specific chronic

and acute conditions or diseases, when looking at the social and lifestyle determinants

of population health status.

The Alameda Physical Health Profile was conceptualized as a spectrum of

health from disability or functional incapacity to excellent health or functioning with a

high level of energy. As described in the methods chapter, the physical profile was

aggregated in terms of specific diseases, symptoms, conditions, discomforts, or

disabilities (see Appendix for exact items/ questions). The physical health profile can be

disaggregated into its components, each of which can be examined separately in

conjunction with social determinants (Table 27).

The major components of the physical health profile are mutually exclusive and

include: any disability; no disability but any job dysfunction; no disability or job

dysfunction but one impairment or chronic condition; none of the previous categories

but two or more impairments or chronic conditions; no impairments or chronic

conditions but one or more acute symptoms; and last, those who report no specific

problems with their health and can function with a low level of energy or a high level of

energy in everyday activities. The link test of the physical health profile indicated that

inequality was related to the specification of the profile. The constructed profile was

correct for low access areas but not for high development areas.

172

TABLE 27: ORDERED LOGIT REGRESSION OF ALAMEDA PHYSICAL HEALTH PROFILE BYTYPE OF AVERAGE INEQUALITY IN URBAN AREA (LOW ACCESS TO RESOURCES; HIGH

DEVELOPMENT OF NEW RESOURCES) AND LIFE CHANCES, HEALTH CHOICES AND CIVICCOMMUNITY (LOGIT b, ODDS RATIO e^b)

--------------------------------------------------------------------------------------- low access to resources high development of new resources---------------------------------------------------------------------------------------Physical Health| b z P>|z| e^b b z P>|z| e^b---------------------------------------------------------------------------------------HRQOLLife unhappy | 0.235 1.930 0.054 1.265 0.312 1.931 0.053 1.366Low rated health | 1.306 10.109 0.000 3.692DEMOGRAPHICSUrban Access | -0.556 -2.140 0.032 0.572Female | 0.730 1.869 0.062 2.076 1.998 3.880 0.000 7.379married | 1.041 2.690 0.007 2.834

* Female | -1.176 -2.593 0.010 0.308age | 0.047 6.784 0.000 1.049 0.039 5.776 0.000 1.040

* Female | -0.017 -2.207 0.027 0.982Educational LevelIncompl seced2 | -1.195 -3.249 0.001 0.302 -1.006 -2.617 0.009 0.365Gen sec ed3 | -1.107 -2.899 0.004 0.330Tech sec ed4 | -1.005 -2.730 0.006 0.366Higher ed6, | -0.569 -1.738 0.082 0.565 -0.420 -1.899 0.058 0.656Tech sec ed4 | * Female | 0.678 3.066 0.002 1.971Occupational StatusPension | 0.958 3.499 0.000 2.608white collar | -0.460 -1.926 0.054 0.631Female * occupational status *manual | 0.687 2.297 0.022 1.989Educational level * occupational statusGen seced3*pens | 0.622 1.710 0.087 1.863Tech seced4*pens | 0.655 1.656 0.098 1.925Gen seced3*white | 1.028 2.765 0.006 2.797Tech seced4*white | 0.661 1.972 0.049 1.937High ed5*profess | -1.123 -2.059 0.039 0.325High ed6*profess | -0.499 -2.091 0.036 0.606MobilityMove house | -0.231 -2.003 0.045 0.793 -0.306 -1.914 0.056 0.735Move job | 0.269 1.715 0.086 1.309HEALTH CHOICESAlcohol | -0.461 -2.325 0.020 0.630

* Female | 0.381 1.615 0.106 1.463Obesity | 0.236 1.987 0.047 1.267

CIVIC COMMUNITYInformal NetworksN friends contact | 0.015 2.432 0.015 1.016Formal Networks/ group membershipsUnion group | -0.804 -2.364 0.018 0.447Social/child grp | 0.517 1.670 0.095 1.678occupation * formal networks/group membershipsPension*child grp | 1.075 1.969 0.049 2.930Manual*child grp | 0.943 1.719 0.086 2.569Manual*union grp | -0.445 -1.782 0.075 0.640Professional*union | 0.379 2.008 0.045 1.462education * formal network/group membershipIncoml seced2*union| 1.662 3.456 0.001 5.270 1.615 3.544 0.000 5.030Gen sec ed3*union | 1.042 2.402 0.016 2.837Tech sec ed4*union | 0.909 2.112 0.035 2.483Higher ed5*union | 0.602 1.623 0.105 1.826Higher ed6*union | 0.803 1.984 0.047 2.233Gen seced3*religious|-0.551 -2.143 0.032 0.576High ed6*religious | 0.677 2.537 0.011 1.969female * formal networks/group membershipsF*religious group | 0.512 3.675 0.000 1.670

173

(CONTINUED) ORDERED LOGIT REGRESSION OF ALAMEDA PHYSICAL HEALTH PROFILE BYTYPE OF AVERAGE INEQUALITY IN URBAN AREA (LOW ACCESS TO RESOURCES; HIGH

DEVELOPMENT OF NEW RESOURCES) AND LIFE CHANCES, HEALTH CHOICES AND CIVICCOMMUNITY (LOGIT b, ODDS RATIO e^b)

-------------------------------------------------------------------------------------- low access to resources high development of new resources

-------------------------------------------------------------------------------------Physical Health| b z P>|z| e^b b z P>|z| e^b-------------------------------------------------------------------------------------Social CohesionAnomie | 0.097 2.098 0.036 1.102Distress | 0.210 5.323 0.000 1.234 0.222 4.255 0.000 1.249

* Female| -0.095 -2.246 0.025 0.909 -0.101 -1.816 0.069 0.903

Social SupportPoor marriage| 0.140 2.085 0.037 1.151

* Female| -0.055 -2.410 0.016 0.946 -0.188 -2.479 0.013 0.828------------------------------------------------------------------------------------- Coef. Std.Err. Coef. Std.Err. _cut1 | -1.794 .487 -.031 .572 _cut2 | .331 .455 2.365 .521 _cut3 | 1.937 .460 3.768 .534 _cut4 | 3.466 .469 5.240 .551 _cut5 | 5.992 .483 7.595 .580 _cut6 | 6.980 .493 8.620 .598-------------------------------------------------------------------------------------Approximate likelihood-ratio test of proportionality of odds across responsecategories: chi2(111) = 259.81 Prob > chi2 = 0.0000 (Ho rejected; b not constant)--------------------------------------------------------------------------------------link test for model and dependent variable specification: N of obs = 1237; R2 = 0.1617; N of obs = 689; R2 = 0.1433;Nphyslow | Coef. Std.Err. P>|z| Coef. Std.Err. P>|z|---------+--------------------------------------------------------------------------- _hat | 1.142 .1327 0.000 1.646 .2959 0.000 _hatsq | -.022 .0195 0.256 -.0626 .0279 0.025(Ho not rejected for low access model; Ho rejected for high dev. model, transformdependent variable or respecify model)---------+--------------------------------------------------------------------------- b= raw coefficient; z = z-score for test of b=0; P>|z| = p-value for z-test; e^b = exp(b) = factor change in odds for unit increase in X-------------------------------------------------------------------------------------

The physical health profile, not adjusted for inequality of urban area, indicated

a strong hierarchical relationship with life chances, education and occupation,

especially among and between gender groups (see Appendix). Third-order

interactions with gender and second-order interactions were included and significant.

Professional women, on the whole, were five times more likely to report poor

physical health than other occupations but highly educated professional women were

from 72%-84% more likely than other education/occupation groups to have better

physical health. On the other hand, men with higher education, incomplete higher,

174

and technical secondary educations, employed in a professional capacity, had odds

from almost 4-7 times (OR=3.57, 5.95, 6.89) greater for poorer physical health.

Hierarchy in social status was related to physical health. The risk gradient for

professional men was lower for the most educated and higher for the least educated.

Professional men were at greater risk than women. This finding is consistent with

data on the mortality and life-expectancy gender gap in Russia, discussed previously,

which illustrates to what extent men have suffered the health effects of living in

transitional economies. In addition, obesity, distress, and lack of marital support were

associated in men with a significantly greater risk for poorer physical health than in

women. An anomalous finding was the greater risk of poorer physical health with

alcohol intake among women (OR=0.1.71) but not among men.

Mobility in housing and hierarchy in education were positively associated

with physical health. Higher status occupations, professionals and white collar

workers, who participated in professional and trade organizations were at

significantly less risk for poorer physical health by about 75% than others. Manual

workers involved with unions were also at half the risk for poorer health than others.

However, this positive effect on health was reversed for union members

within an educational gradient. Union members with the lowest educational level,

incomplete secondary school, were almost 6 times more likely to have poorer

physical health, those with general secondary school – 2.5 times, and union members

with completed higher education – 2 times. A third-order interaction between

education, occupation, and formal group membership could not be estimated due to

small cell sizes (5x4x4 matrix). It is likely that negative health of union members is

175

associated with specific work environments, as well as gender and education.

When type of inequality in urban area was controlled, these relationships

followed the educational and occupational gradient, but varied by type of inequality

(Table 28). Union members with the lowest educational level were at 5 times greater

risk for poorer physical health than other groups in both low access and high

development areas. Union membership among professionals or those with incomplete

higher education, as well as lack of social cohesion and mobility in workplace, were

risks only for residents of high development areas.

The education-occupation hierarchy was evident particularly in low access

areas. Professionals with higher education were 40%-70% less likely to have poorer

physical health, while white collar workers with secondary educations were almost 2-

3 times at greater risk than other groups.

The significant positive effect of mobility in housing on physical health

remained for both inequality areas. But women, as a whole, and married men were at

greater risk for poorer physical health if they lived in high development areas.

176

TABLE 28: ORDERED LOGIT REGRESSION OF DISABILITY BY TYPE OF AVERAGEINEQUALITY IN URBAN AREA (LOW ACCESS TO RESOURCES; HIGH DEVELOPMENT OF NEWRESOURCES) AND LIFE CHANCES, HEALTH CHOICES AND CIVIC COMMUNITY (LOGIT b, ODDS

RATIO e^b) low access to resources high development of new resources--------------------------------------------------------------------------------------- disability | b z P>|z| e^b b z P>|z| e^b---------------------------------------------------------------------------------------HRQOLLow rated health | 0.651 4.146 0.000 1.919Life unsatisfy | 0.281 2.007 0.045 1.329 0.435 2.303 0.021 1.545Job unsatisfy | -0.473 -2.346 0.019 0.623LIFE CHANCES/ DEMOGRAPHICSage | 0.051 5.816 0.000 1.054 0.038 6.100 0.000 1.039 * Female |-0.023 -2.335 0.020 0.977female | 1.062 2.012 0.044 2.895 1.579 2.136 0.033 4.854Urban New Development | 0.332 3.707 0.000 1.394 * age | 0.002 1.771 0.077 1.004Educational LevelHigh ed6, complete| -0.834 -2.368 0.018 0.433Occupational StatusManual | -0.899 -2.337 0.019 0.406female * occupationF * white collar | 0.616 1.790 0.073 1.858educational level * occupational statusHigh ed6 * professional| 0.627 1.806 0.071 1.872Gen sec ed3 * white | -1.340 -2.631 0.009 0.261Gen sec ed3 * pension | 0.724 1.852 0.064 2.062Incomplete seced2*pens| -0.983 -2.311 0.021 0.373HEALTH CHOICESF * physical activity | 0.277 1.722 0.085 1.324CIVIC COMMUNITYFormal Networks/ group membershipsProfessional/trade grp| 0.743 2.862 0.004 2.109Social/Child group | -0.618 -1.824 0.068 0.530 -2.590 -2.363 0.018 0.075Religious group | 0.322 2.386 0.017 1.386 -0.433 -1.664 0.096 0.648educational level * formal network/group membershipHigh ed6*religious grp| 0.816 2.052 0.040 2.261female * formal networks/group membershipsF*social/child group | 3.037 2.546 0.011 20.856Occupational status * formal networks/group membershipWhite collar*union | -0.982 -2.789 0.005 0.375White collar*religious| 0.611 1.850 0.064 1.842Manual * religious | 1.345 2.495 0.013 3.839Pension*social/child | 1.995 3.174 0.002 7.359Social CohesionAnomie | 0.390 3.419 0.001 1.478F * anomie | -0.354 -2.706 0.007 0.701distress | 0.138 3.143 0.002 1.149F*distress | -0.092 -1.954 0.051 0.915 0.076 2.754 0.006 1.080--------------------------------------------------------------------------------------- coeff. std.err. coeff. std.err. _cut1 | 4.208 .4466 _cut1 | 4.375 .7524 _cut2 | 6.787 .7850 _cut2 | 6.473 .4762

-------------------------------------------------------------------------------------Approximate likelihood-ratio test of proportionality of odds IMPAIR across responsecategories: chi2(37) = 47.88; Prob > chi2 = 0.1085; (Ho not rejected; b constant)--------------------------------------------------------------------------------------link test for model and dependent variable specification: N of obs = 1237; R2 = 0.1161 N of obs = 689; R2 = 0.1389------------------------------------------------------------------------------ Oimpair | coef. z P>|z| coef. z P>|z|---------+-------------------------------------------------------------------- _hat | .585 1.548 0.122 .350 0.800 0.424 _hatsq | .057 1.109 0.268 .083 1.488 0.137(Ho not rejected for low access model; Ho not rejected for high dev. model)------------------------------------------------------------------------------------- b = raw coefficient; P>|z| = p-value for z-test;e^b = exp(b) = factor change in oddsfor unit increase in X

177

DIMENSIONS OF PHYSICAL HEALTHIndividual agency in combination with social factors change health outcomes.

Downward drift or social selection, due to initial or changing health status, are not

necessarily the antecedent causes of proximal risk factors for all outcomes of disease.

One set of determining factors cannot be generalized from one outcome to various

other outcomes unless specifically demonstrated within the same model. The factors

which were associated with disability and chronic conditions varied by inequality

context and psychosocial determinants.

The structures of disability and chronic diseases, as opposed to acute

symptoms or daily functioning with high energy, illustrate to what extent global

indicators need to be supplemented by information from specific indicators. Specific

indicators also need to be balanced by global assessments of health, which do not

tend to exclude multiple, complex, polymorphous illness states affecting and affected

by the functional levels of everyday activities.

DISABILITY AND CHRONIC CONDITIONSAn international comparison of disability-free life expectancy in 7

industrialized countries, before 1990, indicated that the proportion of disability years

varied between a low 11% in men to a high of 24% in women. Disability was greatest

among the poorest, and was affected the most by diseases of the circulatory system

(Robine, et al., 1991). The ranking of diseases which affect mortality, morbidity, or

disability are not the same. This order can vary by region and country. The most

important diseases affecting disability-free life expectancy in the West were first,

skeletomuscular and locomotor disorders, second, circulatory disorders, and third,

diseases of the respiratory system.

The threefold increase in disability, between 1980 and 1995 in Russia, was

178

due primarily to an increase in diseases of the circulatory system among the working

age population. Almost one in ten adults was classified as disabled by survey data,

substantially greater than official statistics. This was most likely due to the contrast

between self-reported assessments of the surveys and medically certified data

collected by local polyclinics for government agencies.

The reasons for disability in Russia were not completely consistent with the

rank order of disease-specific causes of death. Trauma, accidents, and homicides

ranked first among causes of death while diseases of the circulatory system accounted

for the greatest number of disabled. Disability and mortality for neoplasms, the third

major cause of death and the second cause of disability, have not shown any major

trend changes for this period. While disability has increased only slightly for diseases

of the respiratory and digestive systems, mortality rates, however, have more than

doubled.

Disability and chronic diseases were intimately linked in Moscow because

57% of all mortality was due to cardiovascular disease, the primary cause of

disability, and the prevalence rates for heart disease were higher in Moscow than in

Russia as a whole. Occupational accidents did not account for increased disability in

Moscow. This implies that personal behavior may be relevant to mortality from

external causes, but that disability is related to other than personal factors.

Conceptualization and measurement of disability in Russia has traditionally

been linked to physical dysfunction in the performance of a socially necessary role.

Disability has been defined as the loss of working capacity either completely or over

a protracted period because of trauma or chronic disease. There is some overlap

179

between the physical health profile conceptualization of disability and that of the

official Russian statistical agency which maintain public health data. Disability was

considered by each as impairment of working capacity by illness or injury .

The definition used by the Alameda Physical Health Profile was constructed

from multiple self-reported items, summarized in Tables 15 and in detail in

Appendix 10. The disability category was the assumed to have the worst health, and

consisted of 6 questions covering functional limitations, mobility, as well as being

unable to work or needing to cut down on hours worked due to some illness or injury.

Chronic conditions consisted of a check list enumerating specific observable

impairments or conditions.

Hypertension, respiratory and heart disease were prevalent among 34% of

adults in the HRQOL survey. Arthritis, hepatitis, and gastritis occurred in 10% to 16%

of Moscow adults. Type of inequality was associated with specific diseases: 40% of all

hypertension, respiratory and heart disease cases were located in low development-low

access urban areas; 9% in high development-high access areas; and 16% in low

development-high access areas; tuberculosis was significantly more prevalent in low

access urban areas. Poor self-rated health was a significant risk for disability and

chronic conditions in low development-low access areas.

Women were almost 5 times more likely than men to have disabilities if living

in high development areas and 3 times more likely if living in low access areas. There

was no direct gender effect for chronic conditions (Table 29).

180

TABLE 29: ORDERED LOGIT REGRESSION OF CHRONIC CONDITIONS BY TYPE OFAVERAGE INEQUALITY IN URBAN AREA (LOW ACCESS TO RESOURCES; HIGH DEVELOPMENTOF NEW RESOURCES) AND LIFE CHANCES, HEALTH CHOICES AND CIVIC COMMUNITY (LOGITb, ODDS RATIO e^b)____________________________________________________________________ low access to resources high development of new resources--------------------------------------------------------------------------------------- N chronic conditions| b z P>|z| e^b b z P>|z| e^b-------------+-------------------------------------------------------------------------HRQOLLow rated health | 1.181 8.922 0.000 3.259Life unhappy | 0.426 2.616 0.009 1.531Job unsatisfy | -0.237 -1.809 0.070 0.788LIFE CHANCES/ DEMOGRAPHYage | 0.041 8.537 0.000 1.041 0.052 9.151 0.000 1.054Educational LevelGen second ed3 | -0.398 -2.276 0.023 0.671Tech second ed4 | -0.573 -2.581 0.010 0.563Higher ed6, completed | -0.902 -2.671 0.008 0.405Occupational StatusProfessional | -0.824 -2.339 0.019 0.438educational level * occupational statusIncompl seced2*pension| -0.521 -1.783 0.075 0.593Tech seced4*pension | 0.829 1.940 0.052 2.292Tech seced4*profession| 0.949 2.995 0.003 2.584High ed5*white collar | -1.593 -2.190 0.028 0.203High ed6*white collar | -0.700 -2.267 0.023 0.496MobilityMove house | 0.304 1.794 0.073 1.355HEALTH CHOICESObesity | 0.227 1.830 0.067 1.254Any smoking | 0.215 1.599 0.110 1.240Any alcohol | -0.599 -3.402 0.001 0.549 -0.398 -2.568 0.010 0.671

* Female | 0.296 1.680 0.093 1.344CIVIC COMMUNITYInformal NetworksN friends contact | 0.011 1.601 0.109 1.011Formal Networks/ group membershipsUnion group | -0.502 -1.672 0.095 0.605 -0.701 -2.682 0.007 0.496Religious group | 0.328 2.822 0.005 1.388Social/Child group | 0.657 1.979 0.048 1.929female * formal network/group membershipF * union group | 0.679 3.219 0.001 1.972educational level * formal network/group membershipIncompl sec ed2*union | 0.977 2.852 0.004 2.658Tech sec ed4*religious| 0.663 2.732 0.006 1.941High ed6*religious grp| 0.548 1.972 0.049 1.730High ed6*union group | 0.793 2.114 0.035 2.212occupation * formal networks/group membershipsManual * union group | 0.740 2.337 0.019 2.096Pension * union group | 0.956 2.530 0.011 2.602 1.093 2.844 0.004 2.983White col*social/child| 0.978 2.449 0.014 2.660White collar*union grp| 0.892 2.787 0.005 2.440Profession*union group| 0.883 1.977 0.048 2.420Social CohesionAnomie | 0.106 2.173 0.030 1.112Distress | 0.148 8.343 0.000 1.160 0.174 6.794 0.000 1.191-------------------------------------------------------------------------------------

coef. std.err. coef. std.err. _cut1 | 2.250 .353 _cut1 | 2.956 .435 _cut2 | 3.550 .363 _cut2 | 4.164 .451

_cut3 | 4.951 .375 _cut3 | 5.498 .470-------------------------------------------------------------------------------------Approximate likelihood-ratio test of proportionality of odds CHRONIC across responsecategories: chi2(123) = 154.30; Prob > chi2 = 0.0294; (Ho rejected; b not constant)--------------------------------------------------------------------------------------link test for model and dependent variable specification: N of obs = 1237; R2 = 0.2017; N of obs = 689; R2 = 0.1959;Ochronic | coef. std.err. P>|z| coef. std.err. P>|z|

181

---------+----------------------------------------------------------------------------- _hat | 1.087 .158 0.000 1.159 .253 0.000 _hatsq | -.014 .023 0.564 -.019 .031 0.513(Ho not rejected for low access model; Ho not rejected for high dev. model)---------+----------------------------------------------------------------------------- b = raw coefficient; z = z-score for test of b=0; P>|z| = p-value for z-test; e^b = exp(b) = factor change in odds for unit increase in X---------------------------------------------------------------------------------------

There was no consistent hierarchical relation of education or occupational

status with disability or chronic conditions, controlling for type of inequality.

However there was a significant interaction between inequality, education and

occupation. University and higher educated professionals had the greatest risk for

disability in high development areas but least risk for chronic diseases in low access

areas.

Men were at greater risk than women for disability due to psychological

distress in low access areas and due to lack of social cohesion in new development

areas. Housing mobility, anomie and distress were risk factors for chronic conditions

in both men and women in new development areas.

Membership in various social groups is associated with social trust, a more

effective government, and public well-being (Putnam, 1993). Alameda County, and

other longitudinal studies, have demonstrated, since 1965, the important protective

effects which the number and size of informal and formal group networks have upon

longevity, even under circumstances of relative social and economic deprivation

(Mossey, 1982; Idler, 1992; Marmot, 2001; Marmot, 2002).

Participation in social groups has been related to better health in Russia (Rose,

1998). The New Russian Barometer reported that the odds for poor physical

functioning were greater for those persons who could rely only on formal institutions,

such as the state, employer, charity, or church, for help with problems (Rose, 1998).

182

This indicates that participation as a risk in Moscow might be due to the lack of other

resources. This emphasizes the importance of family and friends to health status.

Chronic conditions were more likely among those who reported more

interactions with friends (but not family), and formal groups, such as unions and

religious groups. This indicates that, given the ubiquity of the extended family,

friends may be more important in procuring help. It is possible that only certain group

memberships may be important for health, such as those which provide

psychologically or economically-related support.

On the whole, social participation in new development areas had a protective

effect against disability and chronic conditions, except for women’s involvement with

social or child-related groups. These women were 21 times more likely to have

disabilities. It is unclear what this significant relation really indicates unless group

memberships are health related and the children are also disabled. Data on children

was not available in this sample.

Those participating in religious groups were on the whole at greater risk for

disabilities and chronic conditions. White collar union members in low access areas

were less likely to have disabilities, but more likely to have chronic conditions. Union

members in all occupational groups were at greater risk for chronic conditions if they

lived in low access areas, as did union members with the lowest and highest

educational levels in high development areas.

Several lifestyle factors, associated with mortality, were not found to be

associated with disability in Moscow. Obesity, smoking, and alcohol consumption,

were associated with chronic conditions in low access areas. Drinking alcohol

183

predicted health outcome independently of other factors in a direction inconsistent

with mortality studies. Drinking alcohol was related to less risk for chronic conditions

among men but not women, and less risk for acute symptoms for both men and

women.

On the whole, the extent to which social factors, such as marriage, education,

social integration, as well as health choices, influence physical health is determined,

not only by model specification (Duncan et al., 1993), but by a larger culturally-

specific context (Blaxter, 1990; Macintyre, et al., 1996; Seeman, 1988). The disabled

were 12% of the sample and chronic conditions were the most prevalent, comprising

58% of the sample. This was consistent with other surveys and Goskomstat data

(Maximova, 1995). Type of average inequality, life chances, lack of social cohesion

and social participation were significant risks for disability and chronic conditions.

ACUTE SYMPTOMSGenerally, there were significant interactions between gender, education,

occupation. Women in new development areas were almost 5 times more likely to have

acute symptoms than men. Health choices, distress, poor marital support, and mobility

in workplace were associated with greater risk for acute symptoms in all urban areas.

Drinking alcohol reduced the risk for acute symptoms, as did smoking for women.

An inequality and education gradient effect was significant among pensioners.

Those with incomplete secondary school had 50 times greater odds for acute symptoms

than pensioners with a completed higher education in new development areas (Table

30).

184

TABLE 30: ORDERED LOGIT REGRESSION OF ACUTE SYMPTOMS BY TYPE OF AVERAGEINEQUALITY IN URBAN AREA (LOW ACCESS TO RESOURCES; HIGH DEVELOPMENT OF NEWRESOURCES) AND LIFE CHANCES, HEALTH CHOICES AND CIVIC COMMUNITY (LOGIT b, ODDS

RATIO e^b) low access to resources high development of new resources

------------------------------------------------------------------------------------- N Acute Symptoms | b z P>|z| e^b b z P>|z| e^b-------------------------------------------------------------------------------------HRQOLLife unhappy | -0.383 -2.814 0.005 0.681 -0.359 -1.972 0.049 0.698Low rated health | 1.298 10.062 0.000 3.663LIFE CHANCES/ DEMOGRAPHICSage | 0.020 4.764 0.000 1.020 0.027 3.163 0.002 1.028female | 1.584 6.487 0.000 4.876married | 0.517 3.084 0.002 1.678 0.420 1.864 0.062 1.522Urban New Dev | 0.470 1.871 0.061 1.600 * age | -0.009 -1.636 0.102 0.990Educational LevelIncomplete second ed2| 0.708 1.692 0.091 2.030 -4.790 -3.636 0.000 0.008Gen second ed3 | -1.651 -2.070 0.038 0.191Tech second ed4 | -2.121 -2.512 0.012 0.119High ed5, incomplete | -2.331 -2.621 0.009 0.097High ed6, complete | -3.973 -4.193 0.000 0.018Female * educational statusF * incompl sec ed2 | -0.967 -2.035 0.042 0.380Occupational StatusManual | -0.750 -2.358 0.018 0.471MobilityMoving jobs | 0.277 2.300 0.021 1.320 0.558 3.298 0.001 1.748Female * occupational statusF * manual | -0.775 -1.799 0.072 0.460F * pension | -2.019 -3.696 0.000 0.132educational level * occupational statusInc second ed2*white | 1.479 1.646 0.100 4.391Inc second ed2*pension| 4.018 3.246 0.001 55.619Gen second ed3*pension| 1.339 1.998 0.046 3.818Tech sec ed4*pension | 1.487 1.782 0.075 4.426High ed6*pension | 3.139 3.930 0.000 23.102High ed6*white collar| 2.110 3.256 0.001 8.255High ed6*professional| 1.718 2.972 0.003 5.577Health ChoicesAlcohol | -0.204 -1.685 0.092 0.814 -0.321 -1.879 0.060 0.724Any smoking | 0.695 3.342 0.001 2.004 0.672 2.392 0.017 1.959 * Female | -0.754 -2.976 0.003 0.470 -0.580 -1.642 0.101 0.559F *Physical Activity | 0.278 1.964 0.050 1.320CIVIC COMMUNITYFormal Networks/ group membershipsUnion group | -0.508 -2.377 0.017 0.601 -2.456 -2.401 0.016 0.085Social/child group | 1.107 2.526 0.012 3.025female * formal networks/group membershipsF*social/child group | 1.563 2.336 0.020 4.773F * religious group | 0.321 2.358 0.018 1.379F * union | 0.626 2.746 0.006 1.870occupation * formal networks/group membershipsPension*union group | 0.511 1.617 0.106 1.667 1.012 1.845 0.065 2.752Manual*union group | 0.690 1.876 0.061 1.994 0.704 1.948 0.051 2.022White col*religious | -0.692 -2.251 0.024 0.500White col*social/child|-1.263 -2.074 0.038 0.282Profession*soc/child | -0.994 -1.735 0.083 0.369education * formal network/group membershipIncompl sec ed2*union| 5.077 3.542 0.000 160.339Gen sec ed3 * union | 1.778 1.693 0.090 5.918Tech sec ed4*religious| 0.891 2.613 0.009 2.438Tech sec ed4*union | 2.244 2.062 0.039 9.438High ed5,incom*union | 2.793 2.374 0.018 16.336High ed6,compl*union | 2.226 2.089 0.037 9.263High ed6,compl*relig | 0.481 1.680 0.093 1.617

185

Social CohesionF * anomie | 0.053 1.608 0.108 1.054Distress | 0.237 10.728 0.000 1.267 0.359 10.660 0.000 1.432Social SupportPoor marriage | 0.070 2.376 0.018 1.073 0.117 2.977 0.003 1.124---------------------------------------------------------------------------------------- Coef. Std.Err. Coef. Std.Err. _cut1 | .966 .316 _cut1 | -.264 .918 _cut2 | 2.091 .3196 _cut2 | .996 .920 _cut3 | 3.071 .325 _cut3 | 1.912 .923 _cut4 | 3.922 .3331 _cut4 | 2.745 .925----------------------------------------------------------------------------------------Approximate likelihood-ratio test of proportionality of odds ACUTEacross response categories: chi2(123) = 136.71 Prob > chi2 = 0.1879(Ho not rejected; b constant) linktest: link test for model and dependent variable specification: N of obs = 1237; R2 = 0.1745; N of obs = 689; R2 = 0.2023;---------------------------------------------------------------------------------- Oacute | Coef. Std.Err. P>|z| Coef. Std.Err. P>|z|---------+-------------------------------------------------------------------- _hat | 1.203 .167 0.000 1.224 .127 0.000 _hatsq | -0.031 .024 0.204 -0.051 .024 0.033(Ho not rejected for low access model; Ho rejected for high dev. model, transformdependent variable or respecify model)---------+---------------------------------------------------------------------------- b= raw coefficient; z = z-score for test of b=0; P>|z| = p-value for z-test; e^b = exp(b) = factor change in odds for unit increase in X--------------------------------------------------------------------------------------

The risk for acute symptoms varied by type of average inequality in urban areas

and by interactions between education and occupations and between education and

group memberships. Women participating in unions (in low access areas) and religious

groups or social/child-related groups (in new development areas) were at greater risk

for acute symptoms. Manual workers or pensioners who were union members had

greater odds in new development areas for acute symptoms. Professional and white

collar workers who participated in social or child-related groups (in low access areas)

and white collar workers involved in religious activities (in high development areas)

were at half the risk for acute symptoms.

The hierarchical and interaction effects of education were more pronounced

with union participation in high development areas. Union members with incomplete

secondary education (OR=160.34) were over 150 times at greater risk for acute

symptoms than union members with a completed higher education (OR=9.26).

186

TABLE 31: ORDERED LOGIT REGRESSION OF LOW ENERGY LEVELS BY TYPE OFAVERAGE INEQUALITY IN URBAN AREA (LOW ACCESS TO RESOURCES; HIGH

DEVELOPMENT OF NEW RESOURCES) AND LIFE CHANCES, HEALTH CHOICES AND CIVICCOMMUNITY (LOGIT b, ODDS RATIO e^b)

--------------------------------------------------------------------------------------- low access to resources high development of new resources---------------------------------------------------------------------------------------N low energy symptoms | b z P>|z| e^b b z P>|z| e^b---------------------------------------------------------------------------------------HRQOLLow rated health | 1.077 8.609 0.000 2.94Life unsatisfy | 0.303 1.856 0.064 1.35Job unsatisfy | 0.450 3.521 0.000 1.57 0.521 3.080 0.002 1.68LIFE CHANCES/ DEMOGRAPHICSage | 0.027 5.247 0.000 1.027F * age | 0.018 4.719 0.000 1.019F * married | 0.241 1.774 0.076 1.273Educational LevelIncomplete second ed2 |-1.529 -2.178 0.029 0.216 -2.917 -4.705 0.000 0.054Tech second ed4 |-0.828 -3.141 0.002 0.436Higher ed6, completed | 0.867 2.616 0.009 2.379Female * educational levelF * incomplete sec ed2| -1.121 -2.236 0.025 0.325F * gen second ed3 | -0.980 -3.594 0.000 0.375F*high ed6, completed | -1.063 -3.508 0.000 0.345Occupational StatusManual | -0.965 -3.323 0.001 0.381White collar | -0.733 -2.191 0.028 0.480female * occupational statusF * manual | 0.889 2.280 0.023 2.433F * white collar | 0.813 2.982 0.003 2.255 1.088 2.857 0.004 2.970F * professional | 0.955 3.083 0.002 2.600educational level * occupational statusInc sec ed2*pension | 1.651 3.006 0.003 5.213 1.989 3.338 0.001 7.309Gen sec ed3*pension | 0.918 2.343 0.019 2.506Gen sec ed3*manual | -1.373 -3.653 0.000 0.253Gen sec ed3*profess | -1.271 -2.472 0.013 0.280Tech sec ed4*pension | 1.664 3.572 0.000 5.285Tech sec ed4*manual | 1.284 3.359 0.001 3.612High ed5*white collar| -1.105 -1.758 0.079 0.331High ed6*white collar| -1.153 -3.195 0.001 0.315High ed6*profession | -0.827 -2.608 0.009 0.437HEALTH CHOICESPay private MD | 0.375 2.403 0.016 1.455Any smoking | -0.419 -2.622 0.009 0.657

* Female | -0.288 -1.919 0.055 0.749Physical Activity | -0.737 -3.464 0.001 0.478

* Female | 0.766 3.534 0.000 2.151

CIVIC COMMUNITYInformal NetworksN family contact | 0.016 2.450 0.014 1.016N friends contact | 0.018 2.102 0.036 1.018Formal Networks/ group membershipsSocial/child group | -0.531 -2.090 0.037 0.587 -0.633 -1.822 0.068 0.530Union group | 0.767 3.045 0.002 2.153occupation * formal networks/group membershipsPension*union | -0.561 -1.626 0.104 0.570Manual*social/child | -1.304 -1.605 0.109 0.271Manual*religious grp | -0.624 -2.062 0.039 0.535Whitecollar*religious| -0.596 -2.461 0.014 0.551Profession*religious | -0.542 -2.133 0.033 0.581Profession*union grp | -0.868 -3.104 0.002 0.419education * formal network/group membershipInc second ed2*union | 1.619 2.935 0.003 5.051 2.019 3.386 0.001 7.532Gen sec ed3*soc/child| 1.033 1.651 0.099 2.809Gen sec ed3*union grp| 0.650 2.247 0.025 1.915High ed5*union group | -0.656 -1.959 0.050 0.518

187

High ed6*religious grp| 0.545 2.055 0.040 1.724female * formal networks/group membershipsF * religious grp | -0.605 -3.515 0.000 0.545F * union group |-0.677 -2.584 0.010 0.508

MobilityMove housing | -0.240 -2.021 0.043 0.785Social CohesionDistress | 0.156 8.377 0.000 1.168 0.157 6.620 0.000 1.170------------------------------------------------------------------------------------- _cut1 | .0202325 .3062224 -1.937179 .4033964 _cut2 | 1.520407 .3080538 -.2646692 .3614377 _cut3 | 3.239231 .3190224 1.18773 .358052 _cut4 | 2.944496 .3740368-------------------------------------------------------------------------------------Approximate likelihood-ratio test of proportionality of odds LOENERGY

across response categories:

chi2(123) = 136.88

Prob > chi2 = 0.1852(Ho not rejected; b constant)

link test for model and dependent variable specification:. linktest: N of obs = 1237; R2 = 0.1397; N of obs = 689; R2 = 0.1076;------------------------------------------------------------------------------Oloenerg | Coef. Std. Err. P>|z| Coef. Std. Err. P>|z|---------+-------------------------------------------------------------------- _hat | 1.191366 .1546489 0.000 1.037461 .1859431 0.000 _hatsq | -.0410135 .0308527 0.184 -.0096686 .0437886 0.825(Ho not rejected for low access model; Ho not rejected for high dev. model)---------+-------------------------------------------------------------------- b= raw coefficient; z = z-score for test of b=0; P>|z| = p-value for z-test; e^b = exp(b) = factor change in odds for unit increase in X------------------------------------------------------------------------------

The religiously active with technical secondary school (40% of white collar workers) or

with higher education (23% of white collar workers) had 1.5 to 2.5 times greater risk

for acute symptoms in new development areas, in contrast to white collar workers in

religious groups. Not all the interactions between education, occupation, and formal

networks were estimable due to small cell sizes.

Job satisfaction was not a significant risk for acute symptoms, although it was a

risk for chronic conditions in high development areas.

ENERGY LEVELSSocial determinants of high and low energy levels varied by type of

inequality. Poor self-rated health, life unhappiness, life dissatisfaction, and job

dissatisfaction were significant risks for low energy. Self-rated health and life

satisfaction, although associated with low energy, were not related to high energy.

188

Life unhappiness and job dissatisfaction were about two-thirds less likely to be

associated with high energy than happiness and satisfaction. High and low energy

dimensions were defined by the same variables but with different response categories,

the patterns can be interpreted much as the responses of happiness and satisfaction

(Figure 23) which were affected by a large midpoint category (“somewhat”),

precluding an inverse relationshop.

On the whole, living in low access than in high development areas was a greater

risk for low energy. A significant interaction between gender, education, occupation

and group membership also followed a hierarchical gradient in low access areas.

Women in all occupations were over 2 times more likely to have low energy than

employed men. However, men in the lowest and highest educational categories,

excluding mid-level education, were more at risk for low energy functioning than

women of the same educational levels (Table 31).

In low access areas, lower educated pensioners (incomplete secondary, general

secondary and technical secondary school) were 2.5 to 5.3 times at greater risk for low

energy than higher educated professionals (OR=0.44) or white collar workers

(OR=0.32). There was also a significant hierarchical effect of social status on low

energy: union members with incomplete secondary school had more than 5 times

greater odds for low energy (OR=5.1) than those with higher education (OR=0.52).

Participating in social networks was inversely associated with low energy to the

same extent among professionals, pensioners, manual and white collar workers who

were involved in either union or religious groups. Being less likely to have low energy

was not synonymous with having better health. On the contrary, it could be interpreted

189

TABLE 32: ORDERED LOGIT REGRESSION OF HIGH ENERGY LEVELS BY TYPE OF AVERAGEINEQUALITY IN URBAN AREA (LOW ACCESS TO RESOURCES; HIGH DEVELOPMENT OF NEWRESOURCES) AND LIFE CHANCES, HEALTH CHOICES AND CIVIC COMMUNITY (LOGIT b, ODDS

RATIO e^b)----------------------------------------------------------------------------------------- low access to resources high development of new resources-----------------------------------------------------------------------------------------N High Energy symptoms| b z P>|z| e^b b z P>|z| e^b-------------+---------------------------------------------------------------------------HRQOLLife unhappy | -0.283 -2.212 0.07 0.75 -0.405 -2.436 0.015 0.66Job unsatisfy | -0.424 -3.192 0.001 0.65 -0.460 -2.565 0.010 0.63LIFE CHANCES/ DEMOGRAPHICSage | -0.015 -3.108 0.002 0.98 -0.029 -5.383 0.000 0.97F * married | -0.250 -1.738 0.082 0.77 -0.494 -2.358 0.018 0.69Educational LevelIncomplete second ed2 | 0.908 2.829 0.005 2.47 0.967 3.493 0.000 2.63High ed6, completed | -0.700 -2.475 0.013 0.49female * educational levelF * high ed, completed| 0.907 2.821 0.005 2.47 0.524 2.433 0.015 1.68Occupational StatusManual | 1.301 4.397 0.000 3.67White collar | 0.921 2.473 0.013 2.51Professional | 1.441 4.395 0.000 4.22female * occupational statusF*pension | -0.932 -3.259 0.001 0.39F*manual | -0.904 -2.424 0.015 0.44F*white | -1.169 -2.863 0.004 0.31F*professional | -1.236 -3.297 0.001 0.29educational level * occupational statusgen sec ed3*manual | 1.296 3.870 0.000 3.65gen sec ed3*profession| 1.096 2.226 0.026 2.99tech sec ed4 * manual | -0.804 -2.371 0.018 0.44MobilityMoving jobs | 0.211 1.734 0.083 1.23HEALTH CHOICESAny smoking | 0.411 2.438 0.015 1.58Physical Activity | 0.362 1.742 0.081 1.43 0.640 2.544 0.011 1.89

* Female | -0.405 -1.694 0.090 0.66 -0.676 -2.349 0.019 0.58CIVIC COMMUNITYInformal NetworksN family contacts | -0.017 -2.595 0.009 0.98N friends contacts | -0.021 -2.296 0.022 0.97Formal Networks/ group membershipsSocial/child group | 0.863 2.514 0.012 2.37occupation * formal networks/group membershipspension*union group | -1.417 -3.474 0.001 0.24white collar*profession| 0.846 1.812 0.070 2.33education * formal network/group membership ed2union | -1.002 -2.480 0.013 0.36 ed6relig | -0.418 -1.868 0.062 0.65female * formal networks/group membershipsF * religious group | 0.315 2.058 0.040 1.37 0.706 3.801 0.000 2.02Social CohesionF * anomie | 0.090 2.447 0.014 1.09distress | -0.193 -10.030 0.000 0.82 -0.170 -6.582 0.000 0.84Social SupportF * Poor marriage | -0.084 -2.351 0.019 0.91----------------------------------------------------------------------------------------- Coef. Std.Err. Coef. Std.Err. _cut1 | -2.284 .338 _cut1 | -3.222 .413 _cut2 | -0.666 .331 _cut2 | -1.457 .395-----------------------------------------------------------------------------------------Approximate likelihood-ratio test of proportionality of odds HIENERGY across responsecategories: chi2(123) = 146.85; Prob > chi2 = 0.0702 (Ho rejected; b not constant)-----------------------------------------------------------------------------------------link test for model and dependent variable specification: N of obs = 1237; R2 = 0.1342; N of obs = 689; R2 = 0.1470;

190

Coef. Std.Err. P>|z| Coef. Std.Err. P>|z| _hat | 1.136 .1488 0.000 .821 .265 0.002 _hatsq | 0.039 .0388 0.309 -.035 .051 0.487(Ho not rejected for low access model; Ho not rejected for high dev. model)----------------------------------------------------------------------------------------- b= raw coefficient; z = z-score for test of b=0; P>|z| = p-value for z-test; e^b = exp(b) = factor change in odds for unit increase in X__________________________________________________________________________________________

as being more likely to have chronic or acute conditions, unless the energy scale was part

of the physical profile as the positive tail. Further, women involved with unions or

religious groups had smaller odds for low energy than men, consistent with the

interpretation that a decrease in low energy meant more chance of poor health.

Greater likelihood for having high energy levels in everyday activities was

associated with a gradient in occupational status among men, but not women, living in

low access areas, and among men who engaged in physical activities (Table 32). High

energy was also more likely with mobility in workplace; involvement in religious

activities for women and professional/trade groups for white collar workers.

EFFECT OF INEQUALITY ON PHYSICAL HEALTHThe effect of educational or occupational hierarchy on physical health may have

been due to the contextual effect of inequality or to variation in the composition of

urban areas. The latter is consistent with the patterns of demographic mobility and

urban growth in Moscow. The new development of employment and housing markets in

peripheral areas of Moscow selected for a different demographic profile of people than

in central areas. However, the effect of residing in peripheral or new development areas

on physical health was not only dependent on individual characteristics but also on the

workplace, housing, or neighborhood context. A logistic model cannot distinguish

between these effects of context and composition.

The indirect effects of composition and context on physical health were

particularly evident in the moderating the effect of social status. The effects of

191

psychosocial factors were consistently significant as major determinants of poor health.

In comparison, health choices had a negligible effect, although positive for physical

fitness, smoking and alcohol consumption. A multilevel model of this relationship will

be explored further in the next chapter.

DISCUSSIONThe quality of the experience of living in urban areas characterized by inequality

was dependent on gender and social status in Moscow, rather than health behaviors or

the prevalence of a civic community. A social status effect was significant in relation to

inequality and physical health, controlling for health choices, social cohesion, and social

participation. Generally, those living in low access areas were more at risk for a poor

quality of life than people in high development areas.

Occupational mobility was associated with greater job satisfaction and better

health in high development areas. This suggests that life control or workplace context

were associated with health, consistent with other survey findings (Marmot et al., 1997;

Bosma et al., 1997).

In the QOL models and the five sub-models of the physical health profile, both

the global and specific health indicators have significant and independent effects on

each other, demonstrating that they measure unique health aspects, and, although

overlapping, are not synonymous. Self-rated health was significantly predicted by

gender, age, distress and physical health, as well as participation in formal organizations

such as religious groups or unions.

This basic pattern was consistent with those found on examining logistical

models of the separate components of the physical health profile: high energy levels,

low energy levels, acute symptoms, chronic conditions, and disability. Age was a

192

consistent predictor, in the expected direction, for all five physical health components.

Gender was less consistent: although women and men had the same odds for reporting

disabilities or acute symptoms, and some gender-education-occupation interactions

were significant, men were more likely to function at a high energy level.

The difficulty in obtaining a unified picture of which individual determinants

were most significant for specific outcomes was due to the multiple interactions

between micro effects, contextual models, and several outcomes. To adequately

estimate all effects, a single regression model is required which includes direct and

interaction effects at three levels: multiple outcomes within individuals at level 1; micro

determinants of persons at level 2; and urban level inequality at level 3. Such a

multivariate multilevel model is further complicated by different measurements of the

outcome variables, which are binary, ordinal, and polytomous. The following chapter

will describe a preliminary, simpler multilevel model, which assumes the continuous

dependent variable to be physical health.

EFFECTS OF HEALTH CHOICESRemarkably, several lifestyle factors, associated with mortality, were not

associated with poor physical health. Problems with sleeping were measured as part of

the distress factor, which was a significant risk for all HRQOL outcomes. Hours of

daily sleep, snacking, or eating breakfast regularly were not significantly related to

health status. Being overweight, sedentary, not drinking alcohol or smoking, were

risks for health disorders. Except for sports, health choices did not significantly change

the odds for happiness or satisfaction. Physical fitness, in particular, was associated

with gender: greater odds for good self-rated health, less odds for disability and acute

symptoms, among men; greater odds for job satisfaction among women; and greater

193

odds for life satisfaction and high energy for both men and women. Interactions

between health choices, such as smoking drinkers or overweight smoking drinkers,

were not estimated due to insufficient cell sizes.

Nutritional status was measured by the bodymass index which was calculated

following WHO recommendations. Those with high energy and no physical health

problems were most likely to have normal weight (BMI=18.6-25.0 wt/ht2), while the

overweight and obese were more likely to be disabled. Being underweight or having

inadequate food intake was not a problem in Moscow before 1992. A nutritional study

of the elderly between 1992 and 1994, based on the Russian Longitudinal Monitoring

Survey which applied the same BMI categories in national samples, found overall

increases in weight between 1992 and 1994. There were statistically significant

increases among the obese, over 60 years old, between 1992-1993; and increases (not

significant) among the normal and underweight groups (Popkin et al., 1996).

Drinking more alcohol was related to having acute symptoms, while drinking

less predicted chronic conditions. Drinking any amount was not a risk for poor self-

rated health. Similar to the findings in this survey, the New Russian Barometer survey

established, in 1996, that current smokers and those drinking alcohol had better physical

functioning (Rose, 1998). Residence in Moscow or in an urban center, having less than

a higher education, drinking at least twice a month, and reporting most severe material

deprivation were associated with being a current smoker (Bobak, 1998b). These

relationships were related to a lack of life control in some social groups.

Comparisons of odds ratios could not be made directly with the New Russia

Barometer survey because alcohol consumption was measured as monthly frequency

194

rather than quantity by the latter. Reexamination of the alcohol frequency variable for

the HRQOL data (coded as frequency of drinking once/week to 4 times/week) found a

significant odds for having chronic conditions among those drinking less frequently

(0.87 [95%CI:0.76-0.99, p<.049]), independently of other factors.

Not smoking predicted high energy but not good self-rated health. It is

noteworthy that the odds for poor physical function and smoking in the New Russia

Barometer survey (OR=0.92; 95%CI=0.56-1.52) was similar to that found in this data

for disability (OR=0.79; 95%CI=0.53-1.19) and high energy/ no health disorders

(OR=0.70; 95%CI=0.45-1.08). The odds were not significant for smoking and poor

self-rated health in the New Russia Barometer (OR=1.29; 95%CI=0.87-1.89), as well as

in the HRQOL study (OR=0.87; 95%CI=0.65-1.16).

Having a high energy level and no significant health disorders was associated

with making a smaller number of healthy choices than those with acute symptoms,

chronic conditions, or disabilities. Total lifestyle effect was calculated as a count of all 7

risk factors: body mass index, smoking, hours of sleep at night, daily snacking, eating

breakfast, participation in sports, and alcohol consumption. Those with a greater

number of low risk behaviors were more likely to have disabilities and chronic

conditions. The differences between the health groups, however, were not great, and

significance was marginal. This is consistent with similarly anomalous results, due to

smoking and alcohol intake, from multiple community surveys in Russia, as described

earlier.

For example, a longitudinal investigation of personal health practices among the

elderly, replicating the operational definitions of the Alameda County Studies, found no

195

overall relation of lifestyle to mortality if the elderly had reached their seventh decades.

Among elderly men, not one of the personal health choices was associated significantly

to mortality, while among women, only never having smoked predicted longevity. The

disparity between lifestyle effect in middle aged adults and the elderly was conjectured

to be due to an initially better health status, and secondly, to the diminished importance

of health practices after a certain stage in life had been reached (Branch, 1984). This

may also be the case in Russia, except that the lifespan is shorter and the lack of

lifestyle effect or variations in expected effect may also be evident at earlier ages.

Patterns of personal lifestyles have been demonstrated to have varying effects in

different social status groups as well as cultures (Bourdieu, 1990; Cockerham, 1997).

Dietary, sleep, and work patterns are culturally specific and particular biomedical

consequences are neither necessary nor uniform across social groups or nations. The

prolonged experience of poor health may develop resilience, not only chronic stress

syndrome, in host immune systems.

Similarly, some choices, such as drinking alcohol, which have a dose-response

relationship, may have a beneficial effect when combined with social factors under

specific conditions (Marmot, 1996). Whereas “drinkers” may be lonely drinkers in

Western cultures, in Russia, for example, sharing a drink is a custom linked with

hospitality, friendship and sociability.

EFFECTS OF CIVIC COMMUNITYCivic community dimensions, on the whole, appeared to be related to happiness

and health but not satisfaction. Distressed men were at greater risk for poor health.

Union membership was positively associated with health at lower educational levels.

The effect of professionals who were union members or manual workers who were

196

professional/ trade group members could not be estimated because the group sizes were

not sufficiently large in the sample. A complex interaction between gender, education,

and occupation with participation in formal networks was reflected in the various effect

patterns for different QOL outcomes.

Life control or workplace context had an indirect effect on poor health through

unhappiness, if dissatisfaction with one’s job is interpreted as an indicator for lack of

control or unsatisfactory workplace setting.

Social cohesion had a direct effect independently of psychological distress and

was consistently related in the same direction for health, happiness, and satisfaction.

Among the indicators of social capital, adjusted for inequality, membership in religious

and professional/ trade organizations was associated predominantly with disability,

whereas union membership was a risk for overall poor HRQOL. Union members were

at greater risk for poor self-rated health, dissatisfaction with their job, unhappiness with

life, chronic conditions, acute symptoms, low energy, and less likely to experience high

energy in daily activities.

Friends, rather than family members, were associated with having chronic

conditions, low energy, but satisfaction with life and job.

EFFECTS OF LIFE CHANCESA persistent effect of inequality and hierarchy in occupation and education was

significant across HRQOL. Although the gradient was significant, no consistent pattern

was associated for all QOL outcomes. Lowest educated manual workers had 9 times

greater odds for job dissatisfaction than other educational-occupational groups, although

this group did not report life unhappiness or life dissatisfaction. Professionals were most

likely to be in poor health but satisfied with their jobs and life. In contrast, those with

197

university and higher educations were likely to have good self-reported health, but the

greatest likelihood for job and life dissatisfaction. Professional men were at risk for

poor health due to stress. This suggests that gender and education interactions were

important qualifiers to the direct effect of inequality and occupation on QOL.

There was a dissatisfaction gradient related to gender and occupational group

within educational level. Women, as a whole, were at greater odds to be dissatisfied

with their jobs than men. Women in all occupational groups were at greater odds for life

unhappiness than men if they lived in high development areas.

The occupational structure of socialist Moscow in 1991 had not yet begun its

reorganization to a market transition at the time of the survey. It is important to keep in

mind that the most important professional or trade organization in the Moscow of 1991

was still the Communist Party. The Party was linked to occupations through

“voluntary” membership and the most accepted and prevalent mode for economic

mobility. Party membership and the Komsomol, a social group for children and

adolescents, similar to the Scouts, may have confounded the formal network measures.

On the other hand, party affiliation provided rewards which could be seen on health

status. The confounding may exist in that formal networks in Moscow have diverse

effects, the specific group is not known and party affiliation was not asked.

As was described in earlier chapters, several socioeconomic factors have had a

negative impact on health in post-Perestroika Moscow: the disintegration of a centrally

planned economy, including the health sector, and general impoverishment of the

working population. About three-fifths of the population were the working poor. The

average pension was near the poverty level and average salaries were three times lower

198

than the subsistence minimum. Low wages and short cash supply from being paid in

kind, inflated prices, marketisation of medicine, housing, electricity, water, telephones

and other utilities, increased the cost of the minimum standard of living for many

families after 1992. Such increases in economic inequality could only exacerbate the

effect of a social status hierarchy on HRQOL.

The factors associated with poor HRQOL raise some doubts about lifestyle,

especially alcohol consumption, as the primary mechanism of premature mortality in

Moscow. The pattern of HRQOL challenges the definition of the health crisis in Russia

as a personal problem, and reconceptualizes it as a social problem.

199

CHAPTER 8: RESULTS OF A MULTILEVEL MODEL OF HEALTH

The working hypothesis posits that social status (education, occupation), health

choices, social support, and social capital, such as participation in formal networks

(religious groups; professional or trade groups) are related to physical health differently

in some urban areas due to the distribution of average and relative inequality across

urban areas.

Multilevel analysis explores the direct effect of individual level variables and

urban level inequality on physical health. Furthermore, the effect of urban inequality on

individual physical health can be examined as a moderating effect on individual level

predictors. The Random Coefficient cross-level model illustrates the moderating effect

of urban factors on the specific individual factors which impact on physical health.

Based on the conceptual model discussed in Chapter 1 (Figure 1) and tested in logistic

and ordinal logit regression models in Chapter 7, four hypotheses were examined in the

multilevel model:

1.) it was expected that low status individuals living in high inequality areas

would have poorer physical health than if they lived in low inequality areas, controlling

for poor health choices;

2.) it was expected that poor health choices among resident in high inequality

areas would have a greater impact on physical health than among residents in low

inequality areas;

3.) it was expected that social capital, social support, social cohesion would have

a protective or buffering effect on the physical health on individuals living in high

inequality areas, controlling for social status and health choices; the buffer effect was

200

expected to be greater for low status individuals living in high inequality areas;

4.) finally, it was expected that average inequality would have an overall effect

on physical health, and relative inequality would increase the hierarchical effect of

individual social status on physical health.

There were poverty pockets in Moscow which were related to social status and

geographic location. The distribution of inequality mirrored a geographical gradient

from the city center to the city outskirts. There were some compositional differences

between areas. The peripheral areas of Moscow, with predominantly lower educated

blue-collar residents, also had a greater proportion of younger, married families, and

less single, older women than the central areas around the Kremlin. About three-fifths

of all families in Moscow were estimated to be poor. Three-quarters of larger than

average families, with three or more children, lived below the poverty level; and one-

third of poor households were larger than average due to the presence of adult

dependents or disabled family members (Shaw et al., 1994) The average family size in

1989 was 3.1 members. Urban areas with larger than average families (over 5 members)

were associated with smaller than average apartments, greater ratio of lower educated to

higher educated residents, and greater numbers of blue collar to white collar workers

(Table 33).

Relative inequality was also related, particularly in the periphery, to the two

measures of average inequality in urban areas. Urban areas with a higher proportion of

lower educated and manual workers were also undergoing new development. Greater

mean alcohol consumption areas were also strongly associated with larger families and

blue-collar workers. There was substantial average and relative inequality in Moscow at

201

the close of the Perestroika period, controlling for compositional differences.

The geographic gradient of central and peripheral areas was primarily due to the

historical significance of the areas surrounding the Kremlin (in the center of the city).

Administrative and government functions were located around the Kremlin, while some

large petrochemical and automobile factories were established in outlying areas. With

urban growth, older factories were also located closer to heavily populated areas.

On the whole, the two average inequality measures were geographically

orthogonal: centrally-located high access areas did not have any new development, and

development of new resources occurred in low access peripheral areas. There were 4

outliers to this general pattern which may have affected the average inequality effect on

physical health and the normality assumption. Normality was not affected significantly

(see Appendix 2). Since relative inequality measures were not affected, and because

maintaining the largest number of groups was necessary to detect cross-level

interactions, the outliers remained in the sample.

The two outliers in the central areas (cE, cN1) exhibited high access to social

resources and high development of new resources (Figure 22). This was consistent with

the Mayor’s environmental report, which indicated that the central eastern

administrative areas had the poorest ecological condition (www.mos.ru). Two

peripheral areas (pNW, pE) had low access to resources and low development of new

resources. This also was consistent with the reported concentration of historically

established facilities in northwest, and with the location of industrial plants and

radioactive contamination in the peripheral east and southeast (Figures 20, 21).

The bivariate correlation matrices of macro (Table 33) and micro variables

202

(Table 19) indicate that there was only a subset of 5 education and blue collar measures

which had a significant covariation between p=0.4 and 0.8. Other macro variables were

not significantly intercorrelated. Therefore, census rates for separate education and

occupation status by gender for each urban area were included in the multilevel model.

203

TABLE 33: Bivariate correlation matrix of urban level variables, Moscow census, 1989 (n=33)* significance level = p<.05

acc dev obese etoh smoke sport anomi progr kidgr union relig famsiz blm blf apt edf edm----------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------

AVERAGE INEQUALITY

Access to Resources 1.00

New Development -0.121 1.00

AGGREGATE URBAN AREA MEANS

Bodyweight 0.364 -0.220 1.00

Alcohol 0.147 0.176 -0.074 1.00

Smoking 0.197 0.080 0.246 0.323 1.00

Sports -0.032 0.101 -0.224 0.116 -0.091 1.00

Anomie -0.046 -0.240 0.345* -0.266 -0.231 0.053 1.00

Prof/Trade grp -0.027 -0.160 -0.164 -0.287 -0.249 0.084 0.392* 1.00

Child/Social grp -0.329 -0.091 0.040 -0.183 0.103 0.407* 0.128 -0.047 1.00

Union grp -0.294 -0.057 0.035 0.035 0.034 0.153 0.270 0.146 0.351* 1.00

Religious grp -0.020 0.016 -0.065 0.129 0.241 0.177 -0.291 -0.061 -0.010 0.234 1.00

POVERTY

Family Size -0.492* 0.280 0.024 0.293 0.202 0.113 -0.216 -0.261 -0.061 0.106 0.160 1.00

RELATIVE INEQUALITY

Blue Col M -0.559* 0.530* -0.118 0.219 0.034 0.086 -0.275 -0.310 0.116 0.121 0.158 0.614* 1.00

Blue Col F -0.457* 0.568* -0.122 0.251 0.056 0.059 -0.329 -0.329 0.050 0.112 0.150 0.634* 0.969* 1.00

Apt Size -0.556* 0.239 0.031 0.237 0.023 -0.129 -0.233 -0.280 0.068 0.132 0.149 0.643* 0.809* 0.807* 1.00

Lo Educ F -0.642* 0.038 -0.071 0.168 0.075 0.022 -0.015 -0.197 0.281 0.265 0.117 0.461* 0.611* 0.593* 0.721* 1.00

Lo Educ M -0.631* 0.109 –0.001 0.163 0.068 0.091 0.011 –0.202 0.164 0.257 0.163 0.501* 0.703* 0.694* 0.751* 0.912* 1.00

204

SPECIFICATION OF THE MULTILEVEL MODELThe multilevel model was estimated for the Moscow sample. Physical health was

modeled as a normally distributed variable in a two-level hierarchical linear model using

HLM version 5.04 software.

A random coefficient model was used to test indivdual level and urban inequality

variables and compare the unexplained variance at each hierarchical level. Since each level

in a hierarchical model is assumed to be a separate population, urban areas were assumed to

be a sample from a hypothetical population of areas, characterized by average and relative

inequality, about which inferences could be made. The random coefficients represent the

residual or unexplained variance at each level.

Individual level predictors were allowed to be random to test between area

differences in average physical health. It was expected that the effect of individual level

social status, health choices, social capital, social support, and social cohesion on health

would be different across urban areas because of average and relative inequality in urban

areas. The means-as-outcome model estimated the effect of urban inequality on individual

physical health; the slopes-as-outcome model estimated the effect of individual level

predictors, within and between urban areas; the random cross-level model estimated the

moderating effect of urban inequality on the individual level predictors.

Table 34 compares the random intercept models estimated, while Table 35 describes

the random coefficient cross-level interaction model. Continuous variables were centered

around the grand mean. A group centered model which included aggregate group means

(Table 33) was estimated to test for compositional effects. The aggregate variables were

found to be nonsignificant and group centering was therefore not applied (Kreft and de

Leeuw, 1998). All macro level variables were centered around the grand mean.

205

The linear regression assumption of homogeneity of level 1 variances was tested

with each estimated model. Heteroscedasticity of residual within-area variances was not

found to be significant.

The relationship between the power to find significant relationships with a dataset is

related to sample size and method for estimating standard errors. Level 1 estimates depend

on the total number of observations (n=1629 in this case), while level 2 depends on the total

number of groups (N=33 in this case). Simulation studies have indicated that about 30 (but

at least 20) groups with 30 observations/ group are required to find significant cross-level

interactions. Although smaller group sizes or a smaller number of groups decrease the power

of finding a significant relationship, simulation studies have also shown that small group

sizes (n=10) have a lower probability of making a Type I error than larger group sizes. There

were an adequate number of groups to conduct a multilevel analysis. Group sizes across the

33 urban areas varied from one low of 9 to a high of 123, with average size of 49.4

individuals. The smallest group of 9 individuals was dropped as a separate observation and

included in a geographically adjacent area, for a total of 32 macro level groups. Maximum

likelihood estimates have been shown to be the most efficient method for maximizing power

and estimating standard errors in multilevel models. Therefore all random coefficient

models and comparison of deviance between models were estimated with ML.

Even if a larger sample produces more robust results, the significant relations found

in this smaller study indicated that a more balanced sample with less sampling or response

bias could produce stronger contextual effects (deLeeuw, 1998; Hox, 1998; Raudenbusch,

1992).

RANDOM COEFFICIENT MODELThe multilevel analysis progressed in three basic steps: first a null random intercept

206

model was estimated, then a random coefficient means-as-outcome model, followed by a

random coefficient slopes-as-outcome, and finally, the random coefficient means-and-

slopes-as-outcome model. The null model provides the intraclass correlation coefficient as a

baseline for assessing “explained” in contrast to “unexplained” variance for subsequent

HLM models. Both the random effects ANOVA (Table 18 )and the Random Intercept

model indicated that the random variation between urban groups was significant (Table 34,

model 1).

The ICC [ = τ00/(τ00 + 2)] for the null model and the RANOVA was equal to

0.008, suggesting that, although significant, urban area heterogeneity was minimal and a

negligible proportion of variance was due to group structure. About 99% of the variance in

physical health was due to individual level effects. Because the ICC is affected by

unbalanced designs, number of groups and group sizes, the statistical effect may not be

large. However, the theoretical model which specifies the contextual effects of average and

relative inequality on physical health can still be tested with this data given a significant

ICC.

The grand mean of physical health across areas was 4.184 ± 1.96 (.0394). The 95%

confidence interval was 4.1068 ≤ γ00 ≤ 4.2612. The variability of urban area means in

physical health was τ00 = u0j = 0.0143. The null hypothesis that τ00 = 0 was rejected at p<.05

level, indicating that average physical health was significantly different between areas,

although the ICC was less than 1%.

A significant intraclass correlation coefficient also indicates that if the clustered data

is not taken into account, the violation of the linear regression model assumption of

independent observations within groups will result in inflated standard errors of coefficients,

207

and the likelihood of seeing nonexistent relationships (Type I errors). Even as small an

intraclass correlation as 0.01, with about 50 observations within each group, will inflate the

Type I error from the posited =.05 to about =.11, which increases to about =.17 for 100

observations per group (Kreft and de Leeuw, 1998; Snidjers, 2000).

Goodness of model fit was assessed by the deviance statistic with a 2 distribution.

A random coefficient means-as-outcome model was investigated (Table 34: model 2),

unconditional at level 1, included all the urban area predictors, to estimate how much

between group variance in the mean of physical health was due to the average and relative

inequality factors.

A random-slopes-as-outcome model, unconditional at level 2, included all the

individual level predictors, was estimated to obtain an average intercept and slope across the

32 urban areas, and to examine the amount of variation in individual coefficients across

areas (Table 34: model 3). Individual level interactions were not entered due to small cell

sizes, continuous variables were centered around the grand mean, dummy variables were not

centered.

RESULTS FOR CONTEXTUAL EFFECTSProportional reduction in variance [τ00 (RANOVA) - τ00 (model 2)] / τ00 (RANOVA)

= 0.9864] indicated that 99% of between urban area variance in the mean of physical health

was accounted for by the macro factors of average and relative inequality (Table 34: model

2). There was no longer a significant random variance across areas, indicating a good fit by

the macro model. The conditional ICC, adjusted for average and relative inequality,

indicated that all the variation in physical health was across individuals within groups,

except for 0.01%. The null hypothesis that physical health is constant between urban areas,

given average and relative inequality as the macro predictors, could not be rejected.

208

The variation of average inequality measured by new development in urban areas

had a significant direct effect on the mean of physical health in areas, whereas inequality of

access to resources did not. It was removed from subsequent models. Relative inequality,

measured by global ratios of blue collar to white collar residents in an area, lower educated

to higher educated residents, and smaller apartment sizes to larger apartments impacted on

average urban physical health. Poverty, measured by the global ratio of family size, did not

vary significantly across groups and did not significantly affect physical health as a direct

effect. It was dropped from subsequent models.

RESULTS FOR FIXED EFFECTSTo examine how much individual level regressions varied between urban areas, in

both their intercepts and slopes, the random coefficient slopes-as-outcome model was

estimated as unconditional at level 2, fitting all the components of social status, health

choices, social cohesion, social support, and social capital as random coefficients (Table 34:

model3). Those variables which did not have a significant random effect were rerun as

fixed, and if not significant again, dropped from the cross-level interaction model

(incomplete higher education, marital status). All other nonsignificant variables were

conceptually important and retained in the subsequent model as nonrandomly varying. The

goodness-of-fit deviance statistic indicated that the individual level variables were a

significantly better fit than the null model to account for the variance of physical health.

The equation for modeling the within-area variability of individual level predictors

on physical health took the following form: Level-1 Model:

Y = B0 + B1*(social support) + B2*(social cohesion) + B3*(social capital-informal network-nfamily) + B4*(social capital-informal network-nfriends) + B5*(social capital-formal network-childgroup member) + B6*(social capital-formal network-professional/trade group member) + B7*(socialcapital-formal network-union group member) + B8*(social capital-formal network-religious groupmember) + B9*(health choice – pay doctor) + B10*(health choice – drink alcohol) + B11*(healthchoice-bodyweight) + B12*(health choice-smoke) + B13*(health choice-sports) + B14*(female) +

209

B15*(age) + B16*(married) + B17*(life chance-mobility in housing) + B18*(life chance-mobility inwork) + B19*(life chance-occupation-professional) + B20*(life chance-occupation-manual) +B21*(life chance-occupation-white collar) + B22*(life chance-occupation-pension) + B23*(lifechance-education-complete higher) + B24*(life chance-education-incomplete higher) + B25*(lifechance-education-technical secondary school) + B26*(life chance-education-general secondary school)+ B27*(life chance-education-incomplete secondary) + R

The equation for modeling contextual effects separated the intercept and area-

dependent residual in the following form: Null Level 2 model: β0j = γ00 + u0j .

Average physical health was significantly different between urban areas, leaving an

unexplained variation between areas, given the direct effect of individual level predictors.

Two dimensions of social capital had a direct effect on physical health within areas,

with random effects: participation in professional or trade groups and being active in

religious groups. The significant random slopes indicated that the direct effect on health was

not constant across groups, a variation not entirely explained by group participation.

Social capital as membership in either child, professional, or religious groups, was

associated with poor physical health, without a random effect, which was consistent with the

basic results of the logistic regressions. Although social capital is hypothesized to have a

positive relationship to health, the Russian context may have cultural aspects which explain

this apparent reversal in expected results, and will be addressed in a later section.

The direct effect of educational or occupational status on physical health was null, on

average, within areas but varied significantly between areas. In some urban areas, specific

educational and occupational groups had significantly different physical health status. A

nonsignificant fixed effect and a significant random effect indicated a contextual interaction

could be expected in the cross-level model.

Mobility in work or place of residence, on average, were significantly related to

physical health within urban areas, a relationship which was constant across areas. Mobility

210

in employment was positively associated with poor health, while mobility in place of

residence was negatively associated with physical health. People who had moved from jobs

were, on average, in poorer health than people who moved their housing accommodations.

Social cohesion had a significant direct effect on health, while social support had a

random effect and no direct effect.

Among health choices, paying a physician out-of-pocket for medical care was

positively associated with poor physical health within areas and varied between areas.

Obesity also had a random effect. Participating in sports and drinking any alcohol, on

average, were negatively associated with poor physical health within groups. Those who

were active in sports or drank alcohol were in better health than those who did not,

regardless of the urban area of residence.

211

TABLE 34: RANDOM COEFFICIENT MODELS OF PHYSICAL HEALTH

Physical Health Profile (scoring from high energy to severe disability)

Random InterceptNull model

Null at L1Null at L2Model 1

Random CoefficientMacro variables forMeans-as-outcomeNull at L1All at L2Model 2

Random CoefficientMicro variables forSlopes –as-outcomeAll at L1 ^Null at L2Model 3

Est. (SE) prob est. (SE) prob. est. (SE) prob est. (SE) prob

Fixed effects Individual Level 1 Fixed effect of slope within areas

Random effect of slope between areas

physical health 4.184 (.039) p<.000 4.193 (.029) p<.000 3.387 (.185) p<.000

COVARIATESSex (female=1) 0.441 (.056) p<.000 0.0002 (.014) p<.014Age* 0.025 (.003) p<.000 0.213 (.461) p<.001Married=1EDUCATIONAL LEVELHigher, completed =1Higher, incomplete =1Tech. Secondary =1General Secondary =1Secondary, incomplete =1

0.356 (.596) p<.016

0.293 (.541) p<.0360.261 (.511) p<.0380.311 (.558) p<.082

OCCUPATIONAL STATUSProfessional =1White collar =1Manual =1Pensioner =1 0.418 (.153) p<.011

0.204 (.451) p<.0820.152 (.389) p<.0340.211 (.459) p<.0720.344 (.587) p<.086

MOBILITYAny change in place of work =1Any change in residence =1

0.125 (.054) p<.029-0.108 (.063) p<.097

HEALTH CHOICESObese =1Smokes =1Any sport *Drinks Alcohol/mos*Pays MD =1

-0.129 (.044) p<.007-0.116 (.051) p<.0330.158 (.083) p<.067

0.064 (.253) p<.007

0.093 (.305) p<.049SOCIAL COHESIONAnomie * 0.079 (.019) p<.000SOCIAL SUPPORTPoor marriage * 0.003 (.055) p<.030SOCIAL CAPITALSees N family/mos*Sees N friends/mos*Union grp member =1Child/social grp member =1Professional/trade grp =1Religious grp member =1

0.399 (0.091) p<.0000.266 (0.114) p<.0260.249 (0.072) p<.002

0.237 (.487) p<.0910.106 (.325) p<.031

^ the overall intercept is interpreted as the expected outcome for a man, average age of45.22 years, with a primary education, not working; lacking formal networks but having averageinformal networks; average social support; not overweight, not smoking, not paying for medicalcare, drinking on the average 0.264 liters/month, engaging in average number of physicalactivities; experiencing average number of anomie symptoms; not married, not mobile, etc.

212

CON’T: RANDOM COEFFICIENT MODELS OF PHYSICAL HEALTHPhysical Health Profile (scoring from high energy to severe disability)

Random InterceptNull model

Null at L1Null at L2Model 1

Random CoefficientMacro variables forMeans-as-outcomesNull at L 1All at L2Model 2

Random CoefficientMicro variables forSlopes-as-outcomeAll at L1Null at L2Model 3

Est. (SE) prob est. (SE) prob. est. (SE) prob

Fixed Effects Urban Area Level 2POVERTYRatio large/average Family size

RELATIVE INEQUALITYRatio small/average Apt size

Ratio Blue/White collar MRatio Blue/White collar FRatio Low/High Ed MRatio Low/High Ed F

AVERAGE INEQUALITYNew Dev of resources, area Access to resources, area

-0.036 (.027) p<.202

-0.155 (.057) p<.012

-0.053 (.139) p<.706 0.303 (.138) p<.038-0.235 (.071) p<.003 0.278 (.066) p<.000

-0.074 (.036) p<.048 0.019 (.031) p<.530

Random effectsLevel 1 residual(within areas) variance σ 2 (sd) 1.754 (1.33) 1.749 (1.32) 1.025 (1.013)Level 2 residual(between areas)intercept variance (sd) chi sq. (df) prob. level

0.0147 (.121)χ2=44.97 (df=2)p<.05

0.0002 (.014)χ2=27.87 (df=23) p<.22

0.396 (.629)χ2=18.06 (df=19)p>.50

Deviance

Var-Cov components testChi. sq. stat (df) p-value

5552.91 (df = 2) 5531.52 (df =11)

χ2=21.39 (df=9)p<.011

5008.39 (df=407)

χ2=544.52 (df=405)p<.000

Homogeneity of L1 var Chi-sq. stat (df) p-value 25.599 (31) p>.500 25.59 (31) p>.500 20.199 (19) p<.383

*centered around grand mean; all level 2 predictors are centered around grand mean;

SUMMARYAverage physical health varied significantly between urban areas. The mean of

physical health was not constant across urban areas. Average and relative inequality, as a

characteristic of urban areas, accounted for the variation in average physical health across

areas, without including the effect of any individual level predictors. Poverty due to family

size and average inequality as access to resources were not associated with physical health

in urban areas and were dropped from the subsequent cross-level model.

213

The relationship of individual level education and occupational status to physical

health varied significantly between urban areas. Among health choices, the effects on

physical health of obesity and paying out-of-pocket for medical care were significantly

different between urban areas. In addition, social support and social capital were

significantly associated with health which varied between areas. This indicated that

contextual effects were present which could be investigated in a cross-level model by the

distribution of inequality across urban areas.

The cross-level model, discussed in the following section, posits that the direct

effect on physical health of individual level life chances, health choices (obesity, paying for

medical care), social support, and participation in formal networks varies by urban area of

residence due to the area level of average and relative inequality.

THE INTERCEPT-AND-SLOPES-AS-OUTCOMES MODELThe random intercept-and-slopes-as-outcomes model includes cross-level

interactions (Table 35) to fit average and relative inequality as predictors of the random

variation in physical health between urban areas, controlling for the individual level

variables.

The cross-level random coefficient model took the following general form:Alameda Physical Health Profile = [00 + 10 (women) + 20 (age) + 30 (educ) + 40

(occup) + 50 (health choices) + 60 (social cohesion, social support) + 70 (social capital)+ 01 (urban average inequality) + 02-05 (urban relative inequality) + .. + 111 (level 1 *level 2 interactions)] + [ u0jXpij + u0j + eij ]; and where:

physical health is scored between 1=high energy to 4=chronic conditions and7=severe disability;

eij is the individual level residual term; the macro level residual terms [u1jXpij + u0j]; the overall mean effect is 00 ; and u0j is the conditional deviation of each urban area from the overall

mean, which remains unexplained, given the urban level predictors and individuallevel predictors in the model (context effect) ;

214

the overall slope effect is 0q , the direct effect of the urban level predictorson the overall mean outcome;

and u1j is deviation of each urban area slope from the overall slope, whichremains unexplained, given the urban predictors, individual predictors, and micro-macro interactions in the model (context effect).

TABLE 35: RANDOM COEFFICIENT MODELS, MEANS-AND-SLOPES-AS-OUTCOMES, URBANAREA INEQUALITY EFFECT ON MEAN PHYSICAL HEALTH

Random CoefficientMeans-and-Slopes-as-OutcomeCross-level InteractionsModel 4aFixed effect Coeff Std.Err prob

Urban Level Direct Effects on Average Physical Health*MEAN PHYSICAL HEALTH G00

POVERTYRatio large/average Family size

RELATIVE INEQUALITYRatio small/average Apt size G02

Ratio Blue/White collar M G03Ratio Blue/White collar F G04Ratio Low/High Ed F G05Ratio Low/High Ed M G06

AVERAGE INEQUALITYNew Dev of resources, area G01Access to resources, area

3.471 0.178 0.000

-0.243 0.484 0.618

-0.717 0.682 0.304 0.334 0.624 0.594 0.157 0.455 0.733 0.415 0.456 0.371

0.494 0.223 0.036

Total Random effectsLevel 1 residual(within areas) variance σ 2 (sd) 0.9945 (0.989)Level 2 residual(between areas)intercept variance (sd) chi sq. (df) prob. level

0.158 (.398)χ2=20.39 (df=18)p<.310

Deviance

Var-Cov components testChi. sq. stat (df) p-value

4772.79 (df=534)

χ2=235.60 (df=127)p<.0000

Homogeneity of L1 var Chi-sq. stat (df) p-value 31.472 (21) p<.066

* all level 2 predictors are centered around the grand mean; effects are controlled for all othervariables in Table 35 and Table 38; (due to size the single cross-level model was split into fourseparate tables)

215

RESULTS FOR FIXED AND INTERACTIVE EFFECTSURBAN AREA AVERAGE PHYSICAL HEALTH

Average inequality was positively associated with average area physical health,

whereas none of the measures of relative inequality were significantly related (Table 35:

model 4a). The overall average of physical health for the base comparison category

equaled was 3.47 for men of average age (44.74 years), with a primary education,

average social support, average anomie, average numbers of family and friends, drinking

an average of 0.266 liters of alcohol in the past month. Physical health was poorer in

areas undergoing development of new resources. Mean health in areas was not affected

directly by more smaller than average apartments, by more blue collar than white collar

residents, or by more low educated than high educated residents in areas. Because in

multilevel models there may be an overall average null fixed effect while variation across

levels can still be present, interactions between the individual and urban areas can be

significant. Although the intercept of model 4 was not significantly reduced by the

variables in model 4, the deviance indicating model fit was significantly smaller. There

were a number of interactions which indicated that the level of relative inequality in

urban areas had significant effects on physical health by moderating the effect of several

measures of social capital, health choices, occupational status, and education.

Women were on average in significantly poorer health than men ( = 3.92), and

the health of women was different across urban areas (Table 36: model 4b). As expected,

there was a significant direct and random effect of age on average physical health.

LIFE CHANCES

On the average, there were no significant direct effects of educational level or

occupational status on mean physical health. However, all educational and occupational

216

levels varied significantly across areas. There was an interaction effect which increased

217

TABLE 36: RANDOM COEFFICIENT MODEL, MEANS-AND-SLOPES-AS-OUTCOME, EFFECTS OFURBAN INEQUALITY AND LIFE CHANCES ON AVERAGE PHYSICAL HEALTH*

Random Coefficient Means-and-Slopes-as-OutcomeCross-level InteractionsModel 4bFixed effect Coeff StdErr prob.

Random effect SD(uij) Var(uij) prob(uij)

PHYSICAL HEALTH INTRCPT1, B0 G00 3.471 0.178 0.000

U0 0.378 0.143 0.045

COVARIATES

Sex (female=1) INTRCPT2, G130 Slope B13

0.448 0.075 0.000U13 0.155 0.024 0.104

Age* INTRCPT2, G140 Slope B14

0.027 0.003 0.000U14 0.012 0.0005 0.001

EDUCATIONAL LEVELHigher, completed =1 INTRCPT2, G210 Slope B21

Tech. Secondary =1 INTRCPT2, G220 Slope B22

General Secondary =1 INTRCPT2, G230 Slope B23

Secondary, incomplete =1 INTRCPT2, G240 Slope interaction B24 *rel.inequality LOW ED F G245

-0.010 0.144 0.944

0.077 0.131 0.562

-0.104 0.133 0.441

0.143 0.1619 0.386

0.811 0.458 0.089

U21 0.505 0.255 0.001

U22 0.374 0.140 0.015

U23 0.387 0.150 0.002

U24 0.454 0.206 0.050

OCCUPATIONAL STATUSProfessional =1 INTRCPT2, G170 Slope B17

White collar =1 INTRCPT2, G190 Slope B19

Manual =1 INTRCPT2, G180 Slope interaction B18 *rel.inequality APT SIZE G182

Pensioner =1 INTRCPT2, G200 Slope interaction B20 *rel.inequality LOW ED F G205 LOW ED M G206

-0.136 0.119 0.266

-0.110 0.124 0.382

-0.095 0.136 0.490

0.898 0.432 0.048

0.275 0.167 0.112

-0.826 0.441 0.073 0.797 0.443 0.084

U17 0.220 0.048 0.033

U19 0.259 0.067 0.009

U18 0.240 0.057 0.063

U20 0.454 0.206 0.017

MOBILITYChange in place of work =1 INTRCPT2, G150 Slope interaction B15 *rel.inequality APT SIZE G152 BLUE COLLAR F G154

Any change in residence =1 INTRCPT2, G160 Slope interaction B16 *rel.inequality APT SIZE G162

-0.070 0.069 0.324

0.623 0.193 0.004-0.708 0.248 0.009

0.127 0.068 0.076

-0.326 0.184 0.089

U15 0.124 0.015 >.500

U16 0.1494 0.022 0.073

* controlling for all other variables in Table 35 through Table 38; (due to size the single cross-level model was split into four separate tables)

218

the risk for poor physical health by almost 23% among men with the lowest educational

level, incomplete secondary school, who lived in urban areas which had greater ratios of

lower educated women as residents, and 36.3% for women. If manual workers lived in

areas with smaller than average apartments, the risk for poor health was increased by

26% for men , and by an additional 13% for women. The overall effect on physical health

of being a manual worker, holding urban inequality constant, was not significant.

The logistic regressions indicated that two-and three-level interactions were

significant. These could not be included in the multilevel model due to sample size

limitations and model complexity. However, given the random variances for education

and occupation, it is likely that a larger sample and a better specification of occupation

and gender-occupation-education interactions at both the micro and macro levels could

give a better model fit.

Relative inequality affected the risk for poor physical health among pensioners by

an increase of over one-fifth if they lived in urban areas with more poorly educated men,

and a decrease of nearly one-fourth if they lived in urban areas with more poorly

educated women. This apparent inconsistency is most likely due to the prevalence of

extended families with members providing economic and social support for pensioners,

which had a beneficial effect on male health in areas with more low to high educated

women (Ymen=2.65) and a negative effect in areas with more low to high educated men

(Ymen=4.27). If the pensioner was a woman, the risk for poor physical health was 13%

greater than for retired men in either area (YwomenY= 3.09, YwomenY=4.72, respectively).

The gender effect is consistent with the multiple caretaker roles of women which take

their toll on women’s health, especially if the surrounding neighborhood also lacks the

219

support provided by other women.

Mobility in housing had a negative main effect on physical health, increasing poor

health given a constant urban inequality. However, changing housing and living in areas

with smaller than average apartments had a beneficial interactive effect on health

indicating that people probably moved to better conditions. Mobility in employment was

not associated with a significant increase in average poor physical health, holding

inequality constant. However, if men changed jobs and lived in urban areas with smaller

than average apartments, their risk increased by 18% and for women by 31%. This could

be an indicator of downward drift in employment. However, mobility in employment

varied significantly across areas with more blue-collar women: reducing the risk for poor

health by 20% for men and by 7.5% for women. This could be an indication of upward

mobility in blue-collar neighborhoods. Living in areas with small apartments and more

blue collar women, reduced the risk for men by only 2.5% if they changed jobs and

increased for women by 10.5%. Although mobility in employment may be an indicator

of misfortune in Moscow rather than changing to more favorable circumstances, the

urban gender effect is consistent for that observed in occupations. The presence of

women in a Moscow neighborhood rather than men is associated with acting positively

on individual physical health for those changing jobs.

HEALTH CHOICES

Three out of five health choices varied significantly across urban areas, and four

were significantly associated with relative inequality in areas (Table 37: model4c); only

paying for medical services out-of-pocket varied directly with poor physical health,

holding urban inequality constant. Smoking did not have a fixed effect and varied

marginally across areas at p<.09. Regardless of the type of urban inequality, health

220

choices had a positive effect on physical health, with one exception. Obesity, drinking

alcohol, physical activity, and paying for a private physician decreased the risk for poor

physical health.

TABLE 37 : RANDOM COEFFICIENT MODEL, MEANS-AND-SLOPES-AS-OUTCOME, EFFECTSOF URBAN INEQUALITY AND HEALTH CHOICES ON AVERAGE PHYSICAL HEALTH*

Random Coefficient Means-and-Slopes-as-OutcomeCross-level InteractionsModel 4cFixed effect Coeff StdErr prob.

Random effect SD(uij) Var(uij) prob(uij)

HEALTH CHOICESPays MD =1 INTRCPT2, G90 Slope interaction B9 *rel.inequality BLUE COLLAR F G94 LOW ED M G96

Obese =1 INTRCPT2, G100 Slope interaction B10 *rel.inequality BLUE COLLAR F G104

Smokes =1 INTRCPT2, G110 Slope B11

Any sport * INTRCPT2, G120 Slope interaction B12 *ave.inequality G121 *rel.inequality APT SIZE G122

Drinks Alcohol/mos* INTRCPT2, G250 Slope interaction B25 *rel.inequality APT SIZE G252

0.174 0.082 0.045

0.468 0.282 0.109-0.712 0.241 0.007

0.058 0.072 0.430

-0.605 0.269 0.033

0.003 0.074 0.959

-0.083 0.062 0.191

-0.143 0.081 0.087

-0.299 0.174 0.098

-0.170 0.111 0.137

-0.527 0.3162 0.107

U9 0.118 0.013 0.188

U10 0.207 0.043 0.011

U11 0.152 0.023 0.088

U12 0.083 0.006 0.439

U25 0.308 0.095 0.025

* all level 2 predictors are centered around the grand mean; effects are controlledfor all other variables in Table 35 through Table 38; (due to size the single cross-levelmodel was split into four separate tables);

The one exception was for those who paid for private physician services and lived

in neighborhoods with a high ratio of blue to white collar women : mean physical health

for men was 3.94 and for women, 4.39. This effect was constant and marginal (p<.109)

across urban areas and may have been due to poorer initial health status or model

instability. For those paying medical costs out-of-pocket and living in areas with a higher

221

ratio of low to high educated men significantly reduced the mean of average physical

health (γ = -0.712) . This is consistent with a positive health choice effect on individual

physical health. However, it is also likely that the choice effect may be confounded by

individual economic status: people had greater economic resources at their disposal and

could afford to pay for physicians in urban areas with a high ratio of low to high educated

men, or there were more intact families with dual incomes in these areas. This is also an

indicator of a moderating effect by an urban context on individual health choice.

Obesity, however, was positively associated with physical health in areas with

more blue to white collar women (γ = -.605). Relative inequality in apartment size was

also positively associated with health for those who were physically active or drank any

alcohol. The effects were however marginal and need to be re-examined with a larger

sample. Those who were physically active and lived in urban areas characterized by

average inequality had a significantly smaller risk for poor physical health (Ymen = 3.33;

Y women = 3.78). This relationship did not vary across urban areas.

SOCIAL COHESION AND SOCIAL SUPPORT

Social cohesion had a significant negative effect on average physical health,

holding urban inequality constant, and varied across urban areas (Table 38). The effect of

social support on average physical health varied by the level of average inequality and

relative inequality within urban areas and was constant across urban areas. Inequality

significantly moderated the social support effect on health and provided an adequate

explanation for this effect. Health was poorer among those with unsatisfactory marriages

who lived in areas with educational inequality among women.

SOCIAL CAPITAL

Table 38 illustrates the measures of social capital with significant fixed and

222

random effects on average physical health: membership in social or child associated

groups, professional or trade organizations, and religious groups. Average inequality in

urban areas was not significant but ratios of low educational levels and blue collar

residents increased the risk for poor health among those participating in religious or

professional/trade groups. Religious group membership compounded the risk on poor

health as a direct and interactive effect, especially among women living in urban areas

with greater ratios of low to high educated men: Y men = 4.151; Y women = 4.599.

Urban area inequality in apartment size was positively related to average health

among men and increased the risk for poor health among women with greater family

contacts. This effect of social capital may also be due to the benefits provided by an

extended family to male members. Union membership increased the risk for poor health,

holding inequality constant. Having friends to rely on when necessary was not a

significant predictor within or between groups.

223

TABLE 38: RANDOM COEFFICIENT MODEL, MEANS-AND-SLOPES-AS-OUTCOME, EFFECTS OFURBAN INEQUALITY AND CIVIC COMMUNITY ON AVERAGE PHYSICAL HEALTH*

Random Coefficient Means-and-Slopes-as-OutcomeCross-level InteractionsModel 4dFixed effect Coeff StdErr prob.

Random effect SD(uij) Var(uij) prob(uij)

SOCIAL COHESION*Anomie INTRCPT2, G20 Slope B2

0.087 0.0213 0.000U2 0.068 0.004 0.042

SOCIAL SUPPORT*Poor marriage INTRCPT2, G10 Slope interaction B1 *ave.inequality G11 *rel.inequality APT SIZE G12 LOW ED F G15

-0.006 0.011 0.586

-0.028 0.014 0.072

-0.092 0.035 0.017 0.075 0.039 0.036

U1 0.021 0.0005 0.353

SOCIAL CAPITALSees N family/mos* INTRCPT2, G30 Slope interaction B3 *rel.inequality APT SIZE G32 LOW ED F G35F LOW ED M G36

Sees N friends/mos* INTRCPT2, G40 Slope B4

Child/social grp =1 INTRCPT2, G50 Slope B5

Professional/trade grp =1 INTRCPT2, G60 Slope interaction B6 *rel.inequality BLUE COLLAR M G63

Union grp member =1 INTRCPT2, G70 Slope B7

Religious grp member =1 INTRCPT2, G80 Slope interaction B8 *rel.inequality LOW ED M G86F LOW ED F G87

0.002 0.003 0.940

-0.016 0.009 0.086F 0.026 0.009 0.009

-0.020 0.010 0.037

0.002 0.004 0.561

0.422 0.138 0.006

0.139 0.131 0.326

0.826 0.481 0.098

0.168 0.083 0.053

0.226 0.077 0.008

0.454 0.224 0.053 0.371 0.211 0.091

U3 0.009 0.00008 0.348

U4 0.011 0.001 0.312

U5 0.317 0.101 0.079

U6 0.358 0.128 0.003

U7 0.231 0.053 0.315

U8 0.279 0.078 0.003

* all level 2 predictors are centered around the grand mean; effects are controlled for all othervariables in Table 35 through Table 38; (due to size the single cross-level model was split intofour separate tables);*L1 centered around grand mean: social support=2.78 items; anomie=5.25 items;nfamily=16.11 persons; nfriends=16.25 persons; age=44.74 years; alcohol=0.266 liters/pastmonth; all level 2 predictors are centered around grand mean

RANDOM EFFECTS AND SUMMARYAverage physical health was only partially explained by average inequality, the

224

residual variance was significant across urban areas. Relative inequality did not

account for overall average health but did vary significantly with specific slopes of

social status, mobility in housing, health choices, and social capital. The ratio of

apartment size was the most prevalent measure of relative inequality. The measures of

average and relative inequality accounted for different types of variation in physical

health: the main contextual effect on the overall intercept (average inequality) and the

direct and interaction effects of within area slopes which varied across areas (relative

inequality).

The addition of average and relative inequality to the Random Coefficient

Slopes-as-Outcome model 3 (Table 34) significantly decreased the variance in the

mean of physical health, and the residual between areas was no longer significant.

When the deviance was compared between model 4 and model 3, the cross-level model

was a significantly better fit (Table 35).

LIFE CHANCES

The effect of educational level or occupational status was not explained entirely

by the average and relative inequality measures. A significant residual variation

remained between urban areas. The expectation that low status individuals living in

high inequality areas would have poorer health was observed among pensioners, manual

workers, and the lowest education group.

HEALTH CHOICES

Among health choices, obesity and drinking alcohol had significant between

area residual variances, indicating that the contextual effects of inequality, as in the case

of life chances, contributed to explaining a portion of the effect on average health. The

random variation for smoking disappeared when the model was rerun with fixed effects

225

specified for those variables with insignificant random effects.

The proposition that poor health choices among residents in high inequality

areas would have a greater risk for poor health was observed for choosing to pay for

private medical services, but was not evident in the expected direction for obesity,

drinking alcohol, or smoking. Urban areas with high inequality of apartment size

differed only marginally from urban areas with low inequality in the strength of the

positive effect of individual alcohol consumption and physical activity on individual

health. Other factors remain to be considered in explaining why alcohol had a beneficial

effect on the health of men. The effect of physical activity on health was in the expected

direction but living in high inequality areas did not reverse this direction. Obesity had a

positive effect on women’s and men’s health in high inequality areas. The random

effect indicates that perhaps additional nutritional factors need to be considered to

explain this anomaly. Paying for private medical services was moderated by relative

inequality: the risk for poor health was decreased even in areas with greater ratios of

low to high educated residents. This suggests that this positive choice cushioned the

negative effect of urban inequality on physical health, consistent with the research

hypothesis.

ALCOHOL ANOMALY

The alcohol anomaly was examined in greater detail, given the significant

unexplained random residual for alcohol in the multilevel model and the established

relationship of alcohol to education and poor health and premature mortality in Russia

(Cockerham, 2000; Shkolnikov, 2001). To explain the residual, an aggregated group

mean was calculated for alcohol consumption in each area and included as a random L2

variable.

226

The effect of education on physical health status was found to be significantly

moderated by average alcohol consumption levels in urban areas. Individual differences

in educational level and physical health were influenced by macro alcohol consumption

levels as follows: in low consumption areas, physical health was poorer among the

higher educated and better among the lower educated; in high consumption areas,

physical health was better among the higher educated and poorer among the lower

educated.

At the individual level, mean physical health within educational category was

poorer with increased alcohol consumption among the lowest educational groups, and

those with incomplete higher education (Figure 24). This is evidence of a hierarchical

effect of social status on physical health, given alcohol consumption.

FIGURE 22: MEAN ALCOHOL CONSUMPTION AND MEAN PHYSICAL HEALTH BYEDUCATIONAL LEVEL, MOSCOW, 1991

Mean alcohol consumption and mean Physical Health

sta

nd

ard

ize

d v

alu

es

educational level-.2

-.1

0

.1

.2 Standardized values of (etohn) Standardized values of (xphyslow

primary incomple general technica incomple higher

227

The interaction between average and relative inequality and urban area mean

alcohol can be better illustrated by Figure 25. Areas with high access to resources but

low inequality in education among residents tend to also be average alcohol

consumption areas, and these are mostly central areas of Moscow. High inequality in

education among area residents is also associated with living in above average alcohol

consumption areas, which are peripheral areas, and have the least access to material

resources. At the individual level, higher alcohol consumption prevalence is related to a

lower educational level and urban inequality in access to resources. Alcohol

consumption is even more prevalent among those with incomplete higher education but

greater access to material resources.

It appears that for those lower educated individuals living in low alcohol

consumption areas, better physical health is related to a social advantage effect at the

macro level: greater educational egalitarianism in areas, as well as residing in areas with

greater access to material resources. For individuals with higher education living in high

alcohol consumption areas, better physical health was related to an advantage effect of

higher education at the micro level. Those with completed higher education had

prevalence rates of alcohol consumption close to the mean. It was also likely that these

individuals resided in areas which had higher access to material resources.

Poor physical health among the lower educated was associated with living in

high alcohol consumption areas, which were on the periphery of Moscow and had

greater relative inequality and less access to material resources, a cumulative

disadvantage effect.

228

FIGURE 23: MEAN ALCOHOL CONSUMPTION AND MEAN ACCESS TO SOCIAL RESOURCES BYEDUCATIONAL LEVEL, MOSCOW, 1991

[between areas, top 3 graphs, and within areas, bottom histogram]area ratio loed/hied lo inequality=0 hi inequality=1

me

an

are

a a

lco

ho

l co

nsu

mp

tio

ng

ran

d m

ea

n=

0.2

69

Educational inequality and alcohol by Moscow areasstd. values of mean access to social resources

s_edmf2==0

-0.92 0.00 1.00 2.00 3.24.1

.2

.269.3

.4

cen cenw

cenw

cen

cen

swcen

e

cencen

n+zel

cen

cen

cen cen

s_edmf2==1

se ene

e

n+zel

nw

s

w

cen

ne

ne

senw

s

n+zel

sw

Total

-0.92 0.00 1.00 2.00 3.24.1

.2

.269.3

.4

cen cenw

cenw

cen

cen

swcen

e

cencen

n+zel

cen

cen

cen cen

se ene

e

n+zel

nw

s

w

cen

ne

ne

senw

s

n+zel

sw

mean alcohol consumption and mean access to social resources

stan

dard

ized

val

ues

educational level-.2

-.1

0

.1

.2 Standardized values of (etohn) access33

primary incomple general technica incomple higher

229

Finally, those higher educated persons with poor physical health who lived in

lower alcohol consumption areas may have been those with incomplete higher

education and greater prevalence of individual level drinking, a negative personal

effect.

SOCIAL COHESION, SOCIAL SUPPORT, SOCIAL CAPITAL

Social support and group membership in child oriented, religious, or

professional/trade groups had significant residual variation remaining after the

contextual effect of average and relative inequality on physical health. High inequality

in urban areas increased the risk for poor health given social capital measures, contrary

to expectations. There was no beneficial effect on health from formal group

membership in either a direct or interaction effect. The increased risk for poor health

was observed across inequality measures and across social capital measures.

Aggregate variables for all formal network types were included in a separate

model to test for group mean and compositional effects. None of the area mean

aggregates of child, union, professional/trade, or religious groups was a significant

predictor. The random variation suggests that other factors, such as inadequate

measurement of occupational status, or additional interactions between gender,

education, occupation, and social capital at the individual level should be considered.

Some political groups prevalent in post-Coup Moscow, such as the Communist Party,

Komsomol youth groups, other party affiliations, unsanctioned religious affiliations,

were important influences on occupational status and obtaining access to resources but

could not be specifically included in the survey questionnaire.

230

CONTEXTUAL EFFECTS

The significant contextual effects were average inequality in urban areas,

characterized by the geographic gradient of new development of resources and relative

inequality: peripheral areas, characterized by low access and development of new

resources, had a greater ratio of small apartments to large apartments, a greater ratio of

blue collar to white collar workers, and low to high educated residents. These areas

were also associated with having larger families and higher than average alcohol

consumption. The contextual effect of urban inequality on individual physical health

was significant and independent of micro level factors.

The geographic distribution of average inequality acted directly on health, while

relative inequality acted indirectly by moderating the effects of life chances, health

choices and civic community measures. A strong gender effect suggests that perhaps

separate models for physical health should be tested, considering gender –specific

personal and contextual determinants: the health of women was poorer on the average

for all variables but contextual variation, as well, was related to gender-specific

inequality.

A contextual effect was found for the area mean of alcohol consumption as an

interaction with educational level: individuals with less education, controlling for

alcohol intake, had poorer physical health if they lived in urban areas with higher

average alcohol consumption. The risk for poor physical health was decreased if these

individuals lived in urban areas with low mean alcohol consumption. The effect of

urban area mean alcohol consumption contributed an additional effect on health above

that of individual alcohol intake.

The urban inequality effect varied nonrandomly on physical health with mobility

231

in housing, paying private physicians, engaging in physical activities, social support,

and informal family networks. For all other variables, random variation remained to be

explained by factors in addition to average and relative inequality.

232

CHAPTER 9: CONCLUSIONSChapters 1, 2 and 5 described the theoretical multilevel problem of linking

individual, social and cultural parameters; Chapter 6 outlined the methodological issues

of atomistic and ecological fallacies in measuring these psychosocial and structural

parameters in a community survey. Chapter 3 described one possible solution by

applying subjective, self-perceived, self-reported indicators and two types of urban-

level inequality measures to investigate the multiple pathways of effects on multiple

health-related quality of life (HRQOL) outcomes in multilevel models. The results

reported in chapters 7 and 8 indicated that HRQOL was associated with the distribution

of average and relative inequality between the urban areas of Moscow, independently of

the following individual determinants: health choices, life chances, and parameters of a

civic community. The distribution of health was due not to statistical artifact, data

collection procedures, or sampling variation but to the unequal distribution between

urban areas of average access to material resources and relative social status.

SUMMARY OF HYPOTHESESThe basic research questions addressed in this paper were:

1.) to identify the micro and macro level risks significant for poor HRQOL

among individuals in the city of Moscow;

2.) to identify whether the determinants of each of the HRQOL outcomes

indicated the same pathway of effect or different sets of factors;

3.) to assess the additive and interactive effects on HRQOL of two dimensions

of micro level risks – personal health choices and psychosocial behaviors, such as a the

influence of social relations of a civic community, formed by social cohesion, social

support, and social capital; and

233

4.) to examine the broad hypothesis that the distribution of average and relative

inequality at the urban level moderated the physical health of individuals as a main and

joint effect with personal health choices and psychosocial behaviors.

The debate surrounding whether health effects were primarily due to absolute

material position or a comparative socioeconomic position in relation to others was

reformulated to specify that these were two aspects of social position, which had effects

on health at two different points in the distribution of social status (Marmot, 2002;

Marmot and Wilkinson, 2001): the intercept (the mean effect) and the slope (the

strength of association).

It was conjectured that absolute position influenced average health while relative

position varied with multiple social groups and multiple psychosocial determinants. It

was assumed that the average level of HRQOL was directly influenced by a social

context of average inequality and through the psychosocial pathway of health choices,

low social status, and inadequate social integration. Indirect effects arose through the

pathway of relative inequality in urban areas, which moderated the effects of

psychosocial behaviors within areas.

MAJOR FINDINGS▪ Significant individual level effects and contextual effects of average and

relative inequality on health-related quality of life were found. These micro and macro

effects varied by whether outcome was measured as life happiness, life satisfaction, job

satisfaction, self-rated health, or physical health;

▪ The distribution of inequality was an urban-level contextual determinant of

individual health independent of individual risk factors such as social status and health

choices;

234

▪ There was a significant hierarchical effect of life chances on HRQOL,

controlling for health choices; people with less education or lower status occupations

were at greater risk for poor HRQOL regardless of which health choices were made;

▪ Poor health choices did not vary consistently with urban inequality,

controlling for psychosocial behaviors;

▪ The direct hierarchical effects of individual life chances on HRQOL were

independent of the effects of a civic community, in the form of social cohesion, social

support, formal and informal networks, controlling for health choices. Overall

interactions between education with social group membership and occupation with

social group membership were significant but not generally in the direction of having a

buffer effect on health. Social group membership in professional/trade groups, unions,

or social/child-related groups increased the risk for poor health in urban areas of high

relative inequality.

▪ The risk for poor HRQOL was a cumulative function of living on the periphery

of Moscow, in high inequality areas, which had a lack of social and material resources,

were undergoing new development, and more likely to have residents with low social

status;

▪ Average inequality had an overall average effect on mean physical health, and

relative inequality moderated the hierarchical effect of individual social status on

physical health. In sum, personal physical health was affected by where people lived as

well as how they lived.

▪ Although the ICC was so small that it could have been effectively ignored, the

logistic regression results did not agree with multilevel results on many dimensions

235

because of the significant variation between urban areas. The urban context of

inequality made a significant contribution to explaining poor physical health, and

provided an insight into which and how specific psychosocial behaviors change due to

inequality. This study demonstrated that the variable inequality, comprised of two

measures of average inequality and four measures of relative inequality, may be applied

as a macro global independent variable for social class, which combines (as suggested

in Susser, 1973) the components of education, occupation, and place of residence.

SUMMARY OF CONCLUSIONSCONTEXTUAL EFFECTS OF INEQUALITY ON QOL

This study demonstrated that there was substantial variation in the different sets

of contextual and individual psychosocial determinants of multiple outcome indicators

of HRQOL in post-Coup Moscow. Although a multivariate multilevel model could not

be estimated with this dataset, seemingly unrelated estimation (-suest-)was used to

check the significant differences between the models by type of average inequality in

urban areas. Logistic regressions were calculated for each outcome within low access

and high development urban areas. Significant variation between urban areas was

examined further in a multilevel model, which differentiated between the contextual

effects of average and relative inequality on psychosocial determinants of health

outcome.

The logistic models indicated distinct patterns of psychosocial determinants for

physical health, self-reported health, happiness and satisfaction, dependent upon the

type of average inequality in urban areas. The effect of different types of inequality on

life happiness, life satisfaction, or physical and self-rated health varied with

psychosocial behaviors.

Life chances, participation in social groups, health choices, and social

236

status/gender interactions varied significantly by type of average inequality in urban

area. HRQOL varied with social status and inequality: low access areas had the worst

self-rated health status and high development areas had a greater likelihood for job

dissatisfaction and life unhappiness. Job dissatisfaction was a strong risk factor for life

unhappiness. Life unhappiness was a significant factor in life dissatisfaction; both

dissatisfaction and unhappiness with life were risks for poor self-rated health,

controlling for physical health. Women were 12 times more likely in new development

areas and 5 times more likely in low access areas to be dissatisfied with their jobs than

men. The variation of job satisfaction indicates that well-being in work environment

and life control have significant indirect effects on health, which can explain a portion

of the overall gender effect in the health status of women. Job satisfaction as a health

effect was particularly relevant in socialist and post-socialist Moscow where there was

little unemployment among women.

Generally, average inequality was associated with a discernible hierarchical

effect of social status on QOL. The odds of poor self-rated health varied by education

and occupation, as well as average inequality. Although not consistent among the eight

outcome groups because of complex interaction terms, higher education had better odds

for good health, happiness and job satisfaction than lower education. Education showed

a clearer linear effect than occupation. Hierarchical effects on HRQOL were also

evident in the interaction between occupation with education, occupation with social

group memberships, and education with social group memberships.

The lack of a consistent pattern among occupational groups was most likely due

to the collapse of 17 groups into 4 large categories. The measurement of occupation in

237

this dataset by a functional classification system preferred by Moscow researchers and

+did not correspond to a traditional hierarchical grouping used in western studies. The

usefulness of the occupation measure was further complicated by the inclusion of

pensioners as part of the classification, without also asking about a second job. This

group comprised 20% of the sample and had a distinct age (99% ≥ 45 years old), gender

(24% women) and educational distribution (60% with primary, incomplete secondary,

and general secondary education). It was therefore examined as a separate category

which affected the consistency of hierarchical effects.

CONTEXTUAL EFFECT OF INEQUALITY ON PHYSICAL HEALTHThe effect of social status hierarchy on physical health was pronounced for the

interaction between education and occupation and between education and social group

membership with average inequality in urban areas. The hierarchical effect may have

been due to the urban context of inequality or to the variation in the composition of the

sample obtained from urban areas with high inequality. The latter is consistent with the

patterns of demographic mobility and urban growth in Moscow. The new development

of employment and housing markets in peripheral areas of Moscow encouraged specific

people to relocate. A causal pathway can only be suggested in a cross-sectional survey.

Although the relationship between health and social hierarchy may be recursive, the

effect of residing in peripheral areas on physical health was not only dependent on

individual characteristics but also on the workplace, housing, and other neighborhood

factors not subject entirely to personal control.

The effects of social factors were consistently significant as a major determinant

of poor health in comparisons across several logistic regressions. Health choices had

inconsistent effects: physical activity, smoking, obesity and alcohol consumption did

238

not act in the expected direction for all outcomes within and across inequality areas.

The quality of the experience of living in urban areas characterized by inequality

was dependent on the interaction between type of inequality, gender and social status in

Moscow. On the average, those living in low access areas were more at risk for a poor

quality of life than people in high development areas, while physical health risks were

more prevalent in new development areas. Codifying the results of multiple logistic

models was not a parsimonious or accurate method for assessing the differential

consequences for health of direct and interaction effects between macro and micro

determinants, for which the multilevel model is more appropriate.

MULTILEVEL INTERACTIONSThe multilevel model posited that the direct effect on physical health of individual

level life chances, health choices, social support, and participation in formal networks

varied by urban area levels of average and relative inequality. Average physical health

was found to vary significantly between urban areas.

The measures of average and relative inequality accounted for different types of

variation in physical health: the main contextual effect on the overall intercept (average

inequality) and the direct and interaction effects of within area slopes which varied

across areas (relative inequality). The ratio of apartment size was the most prevalent

significant measure of relative inequality. The urban inequality effect varied

nonrandomly on physical health with mobility in housing, paying private physicians,

engaging in physical activities, social support, and informal family networks. For all

other variables, random variation remained to be explained by factors in addition to

average and relative inequality.

The addition of average and relative inequality to the null Random Intercept

239

model significantly decreased the variance in the mean of physical health, and

accounted for the residual variation between areas. The deviance of the cross-level

model as compared to the null model or to the slopes-as-outcome model (the basic

logistic regression) indicated a significantly better fit.

LIFE CHANCES

The hierarchical effect of social status on health was not explained entirely by

the interaction effect of relative inequality. A significant residual variation in social

status remained between urban areas. The expectation that low status individuals living

in high inequality areas would have poorer health was observed among pensioners,

manual workers, and the lowest education group.

HEALTH CHOICES

Among health choices, obesity and drinking alcohol were only partially

explained by relative inequality. The proposition that poor health choices among

residents in high inequality areas would have a greater risk for poor health was observed

for choosing to pay for private medical services, but was not evident in the expected

direction for obesity or drinking alcohol. The effect of physical activity on health was

in the expected direction but living in high inequality areas did not reverse this

direction, indicating the lack of an additive effect of inequality. A random effect for

obesity indicated that possibly additional nutritional factors should be considered.

Paying for private medical services was moderated by relative inequality: the

risk for poor health was decreased even in areas with greater ratios of low to high

educated residents. This suggests that this positive choice cushioned the negative effect

of urban relative inequality on physical health, consistent with the research hypothesis.

It is possible that this health choice is also an indicator of economic status. The

240

association between average inequality, poor HRQOL and willingness to pay for

medical services may, however, have measured the ability to purchase services, as well

as need for services which could not be obtained from the “free” public health system in

any other way.

ALCOHOL ANOMALY

The alcohol anomaly was examined separately given a significant random

residual for alcohol in the multilevel model and the established relationship of alcohol

to education and premature mortality in Russia. Urban area mean alcohol consumption

levels, an aggregate variable, was added to the model.

A hierarchical effect of social status on physical health, given alcohol

consumption levels in urban areas, was observed. There was an interaction effect

between average and relative inequality with urban area mean alcohol consumption

levels. For those lower educated individuals living in low alcohol consumption areas,

better physical health was related to a social advantage effect at the macro level: lower

urban inequality due to smaller educational differences between residents, as well as

greater access to material resources. For individuals with higher education living in high

alcohol consumption areas, better physical health was related to an advantage effect of

higher education at the micro level. Poor physical health among the lower educated was

associated with living in high alcohol consumption areas, which were on the periphery

of Moscow and had greater relative inequality and less access to material resources, a

cumulative disadvantage effect.

SOCIAL COHESION AND SOCIAL SUPPORT

Social cohesion in Moscow included assessments of social chaos and conflict, as

well as a psychological sense of powerlessness or loss of control. There was a

241

convergence of factors which reflected the interaction of perceived anomie and

membership in various social groups on health status. The variation between health and

the slope of social support, measured by the quality of the marital relationship, was

accounted by psychosocial factors and the level of average and relative inequality in

urban areas.

Anomie is not a construct which reflects an internal locus of control or a

particular coping stratagem of the individual, but a personal view of the structural

components of social order: expected and organized rules, values, or norms of social

behavior as means to attain individual goals. In this respect, perceived social cohesion

measures the structural order or disorder of social life from the perspective of the

individual. Individual reports of cohesion are not internal states within the individual

but aspects of the environmental context in which individual and social life occurs and

which are perceived as controllable, to some extent, by the individual to meet personal

goals. A lack of social cohesion was consistently significant for poor health-related

quality of life with all types of urban inequality present.

The structure of a civic community is salient for the theory of health-related

quality of life and institutional cohesion, demonstrating how the integration, rather than

segmentation of the public and private spheres is essential for sustained community

health. It is important to note that lack of social cohesion was the sole civic community

measure which was consistently related to poor HRQOL in all the logistic and

hierarchical linear models. Lack of social support was not significant in the logistic

regression of physical health, but did have a small, direct, and significant effect in the

random coefficient regression. The formal networks of social capital exhibited

242

significant effects on physical health when urban area effects were controlled.

SOCIAL CAPITAL

Residual variation remained in addition to the slope effect of group membership

in child oriented, religious, or professional/trade groups on physical health, which was

not entirely explained by inequality. This random effect may have been due to

inadequate measurement of occupational status, or omitted interactions between gender,

education, occupation, and social capital. High inequality in urban areas increased the

risk for poor health given social capital measures, contrary to expectations. There was

no beneficial effect on health from formal group membership in either a direct or

interaction effect. A general increased risk for poor health was observed across

inequality measures and across social capital measures.

It is likely that the negative effect on physical health was due in part to

professional and trade group membership being tied to occupational status. The

government bureaucracy in Moscow in 1991 was still intact and a dominant employer

of educated professionals. It is possible that professional group membership was also

connected to communist party membership either directly or through party control of

the organization, and that social/child-related groups were linked to the Komsomol.

Respondents may have assumed the group membership question to mean party

affiliation. Communist party membership was more common among the established,

older elite, or among pensioners and veterans, which would be consistent with poorer

health status. In addition, distress levels and poor health were significant in logistic

regression among professional men who were probably suffering from trauma and

genuine threats to personal security during the August Coup.

Another perspective, which is inherent in the dimensions of the civic

243

community, suggests an explanation for the lack of a buffer effect on health of social

capital, as measured in this dataset. Explanation of poor health and premature mortality

have been related to cardiovascular diseases rates in Moscow and Russia, as well as

other Eastern European countries undergoing rapid social change and economic

transition.

Occupational stress can result in increased cardiovascular disease. Individual

self-regulation, self-efficacy, self-worth and self-integration are disrupted by a change

in obtaining social rewards from occupational and economic status (in the case of

Moscow - loss of savings from inflation, loss of wages from unemployment or unpaid

labor, inhibition of access to material and social resources due to social status or

increased prices). Self-integration is affected by the inability to retain social capital due

to loss of social status and access to material resources. The psychosocial elements of

loss of self-regulation, self-efficacy, self-worth, and self-integration after the August

Coup of 1991 may be reflected in the overwhelming sense of normlessness measured by

the social cohesion scale of the Moscow health profile (Siegrist, 1995; Putnam, 1992;

Rose, 1999; Marmot, 2002; Wilkinson, 2001).

The lack of social cohesion was noted in public opinion polls of Muscovites as

predicting lower voter turnout, which in turn retards the growth of a civic society.

Instead of participation in public life, Muscovites have turned to informal means to

achieve their goals. When individual choice and collective choice are diametrically

opposed, the growth of an “uncivil” society is fostered where citizens are alienated

from the political arena which is either nonresponsive or oppressive. They do not vote

even if enabled, separate themselves from public decision-making, and inadvertently

244

cultivate the continuation of the very social mechanisms which create greater relative

inequality.

It has been pointed out by the New Russia Barometer Social Capital Survey that

in a complex “anti-modern” society, like Russia, formal organizations, although

numerous, fail to operate according to the rule of law. As a consequence, alternative

methods for obtaining goods and services were formed by Russians, such as the social

networks of a “second economy” and a “second polity” in a civil society, separated

from the state by self-interested, rather than civic activities. The poor health

consequences of a civil society can be seen in the Moscow health profile, while the

persistence of such “anti-modern” civil networks have retarded the development of a

civic community.

IMPLICATIONSThe main implication derived from the logistic and multilevel models was a

consistent demonstration of how average and relative urban inequality was associated

with cumulative disadvantage. Relative inequality in the composition of social status

between urban areas, at the micro and macro level, was associated with average

inequality in having access to material resources, which resulted in a cumulative

disadvantage for residents of peripheral areas in Moscow.

Structural influences exist independently of individual level behaviors, so that

the social distribution of health is affected by the social distribution of inequality

between geographic areas. Those who lived in peripheral Moscow areas were

surrounded by the effects of a greater prevalence of small apartments, low educational

level of residents, larger than average families, and a large proportion of blue-collar

workers, who were also more likely to live in higher than average alcohol consumption

245

areas. These multiple factors can act as cumulative risks for the adoption of certain

types of unhealthy lifestyle behaviors more so than for those who reside in areas where

there is a lower prevalence of cumulative risk. Individuals living in the “poor health”

areas may have less personal choice in lifestyle due to their place of residence than

persons without such pernicious social influences, holding other factors constant.

The study findings broadly indicate that community context should be

incorporated into public health interventions. Public health policy, traditionally affected

by the methodological individualism of the biomedical model, can benefit from a

multilevel approach which places the individual back in the social milieu as an active

participant. The purposive action of individuals is constrained and shaped by their

institutional setting, which is a macro-micro connection that is also recursive. Although

the parameters of a civic community were not demonstrated by the study to be

significant for a better quality of life in Moscow due to the cultural specificity of the

construct, social connections still provide the context within which individuals are able

to influence public policy.

A second macro-micro problem in public health is to specify the causal

pathways by which health policy can influence health outcomes of individuals and

communities. The outline of the WHO research agenda as set out in the Healthy Cities

program provides a broad outline within which to couch the empirical macro-micro-

macro elements of a multilevel health profile. This is especially critical if individual

preferences are to be considered in formulating collective choices and public policy.

A further macro-micro problem centers on the issue of who is responsible for

individual level and community level health given that individuals influence the

246

communities in which they live, as well as the relative position of other individuals

within the existing context of their neighborhoods. Both of these multilevel issues were

addressed by the Moscow health profile in defining relative inequality in terms of social

status (education, occupation, income/apartment size) and average inequality in terms of

access to material resources. Inequality was defined as having average access to

material resources in comparison to the available distribution of resources between

areas. Both measures are related to the position of individuals in the social structure of a

community rather than to isolated characteristics of individual residents themselves.

The definitions of average and relative inequality in this study sought to operationalize

several of the criteria which the World Health Organization used in distinguishing

between the inequality and inequity of health gaps.

The existence of these health gaps in themselves point to the fact that they are

avoidable and preventable; that persons living under the social conditions specific to

Russia, for example, are more at risk for poor health than persons living in western

Europe. These greater risks are not due to individual behaviors but to the inequality in

nationally and culturally distributed conditions within and between countries.

Broad social strategies to increase equity and decrease inequality in public

health include the quality of living and working conditions; education about healthy

lifestyles; community participation in local, decentralized public health decision-

making; and assessments of policy and program effectiveness.

Although cross-national surveys often indiscriminately compare items across

time and countries, it is necessary to maintain the distinctiveness of periods and

cultures. It is likely that the prevalence of anomie reported by this study was a true

247

cultural effect specific to the structure of the urban community of Moscow, as well as a

reflection of the political crisis occurring at the time the survey was conducted.

STUDY LIMITATIONSThe prevalence of health-related quality of life in Moscow was based upon self-

reported perceptions, derived exclusively upon the survey responses of individuals

themselves. No medical diagnoses of illness or disease were obtained, no physiological

measurements of parameters were taken, no linking with polyclinic medical records was

possible at the time due to political instability. It was therefore not possible to

distinguish between the “subjective” perceptions of illness and “objective” medical

diagnoses in order to obtain a more balanced and rounded profile of Moscow’s public

health.

It was argued that nonprofessional, lay assessments are crucial for the evaluation

of health-related quality of life. However, the relationship of lay assessment to expert

assessment is in itself an important part of public health policy which would have been

included had it not been for the chaotic atmosphere and pervasive angst. The

consistency of responses could have been analyzed, as well as the possible reasons for

inconsistencies between independent sources of health data. The health profile is thus a

picture which the community residents gave of themselves, balanced by data from the

city census on social status and access to material resources in urban areas.

The limitations of the data are affected not only by the sampling frame and

survey errors, as outlined in chapter 4, but by the response error due to the tumultuous

period of the August 18-21, 1991 Coup, which effectively closed the Soviet era in

Moscow. The impact of the Coup on responses could have been substantial in relation

to the self-assessment of health-related quality of life, in particular upon the dimensions

248

of a civic community. However, this cross-sectional picture in time was obtained at a

significant period for future assessments. It will be invaluable to validate health

outcomes with follow-up research, linking individuals to medical records and the city

death register.

The stability of multilevel models are sensitive to the number of variables

included, the number of random parameters, and the number of macro groups. Larger

samples of groups provide greater stability and reliability. Several changes in the

variables tested in the model produced different results. The final model was based on

testing the conceptual dimensions as defined in Figure 1. Although this was a complete

universe of urban areas in Moscow, a better fit for the model may be had with a larger

sample within groups, as well as greater attention to random selection with known

probability of selection. The probability of selection was unknown for this dataset,

which was clearly biased by a large response and interviewer error (Appendix 7).

Another major limiting factor which might have affected responses was the use

of an instrument developed in the United States, a society based on different political,

economic, and social principles than Soviet Russia. Although concrete questions and

the physical health questions were the easiest to translate, any psychosocial items and

the social cohesion dimension presented a different set of concerns. The responses for

most subjective assessments were in the negative direction. In cumulating responses,

life happiness or life satisfaction were insignificantly small as compared to life

unhappiness and life dissatisfaction. This was also true for poor/fair rather than

good/excellent self-rated health, and disability/chronic conditions in contrast to acute

symptoms/no health problems of the Alameda Physical Health Profile. This negative

249

picture of Moscow residents may be a cultural predisposition to attribute a darker hue to

reality than exists, in contrast to other cultures, like the United States, where it is taboo

to casually disclose in public how bad things really are in one’s life. In any case, it is

more likely that responses reflected the social context, given the regularity of subjective

perceptions between other dimensions like social cohesion.

FUTURE RESEARCHNot sufficient progress has been made in the problem of measuring the

difference between concepts of inequity and inequality as relative and average

inequality, respectively. The monitoring of inequity in health has been part of WHO

policy every six years. Inequity has been made an explicit policy component in

formulating community health profiles and deriving community health policies.

However little advances have been made in refining the definition and development of

measures of inequity and their application in either health surveys or health profiles.

In those countries which have been regular participants in the WHO Healthy

City strategy, periodic health profiles in the 1990’s were progress reports which rarely

included measures of macro inequality effects on health. In Russia and other states of

the Former Soviet Union, health profiles have been drawn for the first time as baseline

assessments. A comprehensive national health profile of Russia has still not been

published at the end of the millennium. At the present time, this study remains the only

baseline profile of the city of Moscow which addresses contextual effects of urban

inequality on individual health status outcomes.

250

APPENDICESAPPENDIX 1: Total permanent Moscow population, 1989, unweighted andweighted sampling distributions, Moscow, 1991, by sex and urban area_______________________________________________________________________________________URBAN CENSUS* UNWEIGHTED WEIGHTED** AREA Total(N) M(%) F(%) SAMPLE(N) M(%) F(%) M(%) F(%)_______________________________________________________________________________________MOSCOW 8875579 44.9 55.1 1629 30.3 69.7 44.2 55.8______________________________________________________________________01-Babushkinski 339184 1.7 2.1 56 .9 2.5 1.3 2.002-Baumanski 83550 .4 .5 17 .3 .7 0.4 0.603-Volgogradski 334056 1.7 2.1 64 1.1 2.8 1.7 2.404-Gagarinski 286961 1.5 1.7 46 .7 2.1 0.9 1.905-Dzerzhinski 143528 .7 .9 22 .5 .9 1.0 0.506-Zhelezhno. 157692 .8 1.0 33 .5 1.5 0.8 1.207-Kalininski 98916 .5 .6 18 .5 .6 0.7 0.508-Kievski 158129 .8 1.0 33 .4 1.6 0.6 1.509-Kirovski 582524 3.0 3.6 123 2.5 5.1 4.0 3.910-Krasnopres. 134560 .7 .8 41 .8 1.7 1.1 1.411-Krasnogvard. 655941 3.4 4.0 116 3.1 4.1 4.7 3.312-Kuibishevsk. 344241 1.7 2.2 65 1.2 2.8 1.6 2.113-Kuncevski 384643 1.9 2.4 69 1.2 3.0 1.7 2.314-Leningradsk. 320526 1.6 2.0 66 1.0 3.0 1.4 2.515-Leninski 106703 .5 .7 31 .6 1.2 0.9 0.916-Lyublinski 311749 1.6 1.9 59 .8 2.7 1.1 2.117-Moskvoretsk. 103059 .5 .7 17 .4 .6 0.8 0.518-Octiabrski 218456 1.1 1.4 35 .7 1.3 0.9 0.919-Pervomaiski 363926 1.8 2.3 62 1.0 2.8 1.3 2.220-Perovski 458525 2.3 2.9 77 1.2 3.5 1.7 3.021-Proletarski 239873 1.2 1.5 40 .7 1.7 1.0 1.522-Sverdlovski 115588 .6 .7 19 .4 .7 0.6 0.623-Sevastopol. 286146 1.4 1.8 53 1.1 2.2 1.7 1.724-Sovetski 452568 2.3 2.8 98 1.9 4.1 2.6 3.325-Sokolnicheski 97437 .5 .6 9 .1 .4 0.2 0.426-Solncevski 113460 .6 .7 18 .4 .7 0.6 0.627-Taganski 146822 .7 .9 25 .5 .8 0.6 0.728-Timiriazev. 413296 2.1 2.6 61 .9 2.8 1.2 2.229-Tushinski 260501 1.3 1.6 52 1.3 1.9 2.0 1.530-Frunzenski 161267 .8 1.0 40 .6 1.8 0.9 1.531-Horoshevski 339238 1.7 2.1 59 .9 2.7 1.2 2.132-Cheremushk. 503699 2.6 3.1 78 1.5 3.3 2.2 2.633-Zelenograd 158815 .9 .9 27 .4 1.3 0.6 1.0______________________________________________________________________

Notes to Appendix 2 :*The distribution of the permanent 1989 Moscow population of working age, 20 years andolder (n=6673984), was not available for each urban area, therefore these frequencies bydistrict also include children of all ages;** Because the sampling distribution was based only on adults aged 18 years and older,the referent Moscow population for weighting was calculated from all those age 15years(N=7213121) and was therefore more than the permanent working age Moscowpopulation; age-sex proportions were derived by dividing the number in age-sex groupsby the base referent population 15 years;+Weights=the proportion referent population in age-sex cells divided by proportion of

sample distribution in age-sex cells.

251

APPENDIX 4: Ordered Logit Regression of Physical Health Profile by LifeChances, Health Choices, and Civic Community

(logit b; odds ratio e^b; % change in OR))

------------------------------------------------------------------------------------- Odds of: poor/fair>m vs <=m Physical Health (scored from high 7 to low 1) b z P>|z| e^b/ %-------------+------------------------------------------------------------------------HRQOL lohealt2 | 1.229 11.006 0.000 3.421 242.1 lifehap2 | 0.234 2.193 0.028 1.263 26.4DEMOGRAPHICS i_age | 0.036 8.559 0.000 1.037 3.8 agenewdv | 0.001 1.626 0.104 1.001 0.2Educational Level educ6 | -0.916 -2.800 0.005 0.400 -60.0 educ5 | -0.698 -1.666 0.096 0.497 -50.3 educ4 | -0.998 -2.482 0.013 0.368 -63.2 educ3 | -0.970 -2.653 0.008 0.378 -62.1 educ2 | -1.253 -3.517 0.000 0.285 -71.4Female * age * educational level Fageed5 | 0.047 3.413 0.001 1.049 4.9 Fageed4 | 0.011 2.081 0.037 1.012 1.2Occupational Status profess2 | -1.809 -2.877 0.004 0.163 -83.6Education * occupation Xed6prof | 1.271 1.943 0.052 3.567 256.7 Xed5whit | -1.234 -2.159 0.031 0.291 -70.9 Xed5prof | 1.782 1.834 0.067 5.946 494.7 Xed4prof | 1.929 2.576 0.010 6.888 588.8 Xed3pens | 1.591 1.925 0.054 4.908 390.9Female * occupational status femprof | 1.684 2.316 0.021 5.391 439.2 femmanua | 0.489 1.732 0.083 1.631 63.1female * educational level * occupational status Fed6prof | -1.258 -1.703 0.089 0.284 -71.6 Fed5prof | -3.451 -2.843 0.004 0.031 -96.8 Fed5pens | -3.103 -2.637 0.008 0.044 -95.5 Fed4prof | -1.824 -2.116 0.034 0.161 -83.9 Fed3pens | -1.383 -1.631 0.103 0.250 -74.9Mobility i_naddr2 | -0.178 -1.762 0.078 0.836 -16.3

HEALTH CHOICES femfat | -0.332 -1.661 0.097 0.717 -28.3 femetohn | 0.536 2.951 0.003 1.710 71.0 fat2 | 0.412 2.419 0.016 1.510 51.1 etohn2 | -0.525 -3.311 0.001 0.591 -40.9CIVIC COMMUNITYInformal Networks i_npalC | 0.010 1.887 0.059 1.010 1.1Formal Networks/ group memberships uniongr2 | -0.684 -2.140 0.032 0.504 -49.6 kidgrp2 | 0.594 3.065 0.002 1.812 81.3 i_protr2 | 1.373 2.959 0.003 3.947 294.7occupation * formal networks/group memberships Xwhitpro | -1.561 -2.769 0.006 0.209 -79.0 Xprofpro | -1.368 -2.578 0.010 0.254 -74.6 Xpenspro | -1.350 -1.863 0.062 0.259 -74.1 Xmanuuni | -0.676 -2.899 0.004 0.508 -49.2education * formal network/group membership ed6union | 0.699 1.862 0.063 2.013 101.3 ed6relig | 0.477 2.643 0.008 1.612 61.2 ed4union | 0.676 1.748 0.080 1.966 96.6 ed3union | 0.888 2.237 0.025 2.430 143.0 ed2union | 1.756 3.871 0.000 5.790 479.0female * formal networks/group memberships femrelig | 0.281 2.239 0.025 1.324 32.5Social Cohesion

252

femstres | -0.103 -3.006 0.003 0.901 -9.8 distress | 0.225 6.775 0.000 1.252 25.3Social Support femnegsp | -0.049 -2.477 0.013 0.951 -4.8________________________________________________________________________________ e^b = exp(b) = factor change in odds for unit increase in X % = percent change in odds for unit increase in X b = raw coefficient z = z-score for test of b=0 P>|z| = p-value for z-test % = percent change in odds for unit increase in X

253

APPENDIX 5: Type of Inequality in urban area by Physical Health Profile andQOL: Low-High development of new urban resources by Low-High access to urbanresources in residence areas of Moscow, 1991;(n=1629) percent-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

AREA INEQUALITY 1=LODEV-HIACCESS 2=HIDEV-HIACC 3=LODEV-LOACC 4=HIDEV-LOACC

UNHAPPY WITH LIFENPHYSLOW |other yes |other yes |other yes |other yes-----------+----------------------+----------------------------------------------------- energy | 25 4 | 12 4 | 62 15 | 46 22 | 13.37 5.33 | 14.46 8.51 | 12.86 7.65 | 13.57 10.00-----------+----------------------+------------------------------------------------------ acute< | 34 16 | 14 4 |103 31 | 73 30 | 18.18 21.33 | 16.87 8.51 | 21.37 15.82 | 21.53 13.64-----------+----------------------+------------------------------------------------------ acute> | 49 10 | 26 8 |113 47 | 90 50 | 26.20 13.33 | 31.33 17.02 | 23.44 23.98 | 26.55 22.73-----------+----------------------+----------------------------------------------------- chronic | 61 32 | 22 23 |159 74 | 94 85 | 32.62 42.67 | 26.51 48.94 | 32.99 37.76 | 27.73 38.64-----------+----------------------+------------------------------------------------------ impair | 18 13 | 9 8 | 45 29 | 36 33 | 9.63 17.33 | 10.84 17.02 | 9.34 14.80 | 10.62 15.00-----------+----------------------+------------------------------------------------------ Total |187 75 | 83 47 |482 196 |339 220 % Percent |100 100 |100 100 |100 100 |100 100----------------------------------------------------------------------------------------- chi2(4) = 11.55 (P=.021)| chi2(4) = 9.96 (P=.04)| chi2(4) = 10.25 (P=.036)| chi2(4)=13.73 (P=.008)-----------+----------------------+------------------------------------------------------UNSATISFIED WITH LIFE-----------+----------------------------------------------------------------------------- energy | 20 9 | 12 4 | 60 17 | 57 11 | 11.98 9.47 | 14.81 8.16 | 13.07 7.76 | 14.88 6.25-----------+----------------------+---------------------------------------------------- acute< | 32 18 | 15 3 | 97 37 | 80 23 | 19.16 18.95 | 18.52 6.12 | 21.13 16.89 | 20.89 13.07-----------+----------------------+----------------------------------------------------- acute> | 43 16 | 23 11 |113 47 | 99 41 | 25.75 16.84 | 28.40 22.45 | 24.62 21.46 | 25.85 23.30-----------+----------------------+------------------------------------------------------ chronic | 49 44 | 24 21 |139 94 |105 74

| 29.34 46.32 | 29.63 42.86 | 30.28 42.92 | 27.42 42.05-----------+---------------------+------------------------------------------------------- impair | 23 8 | 7 10 | 50 24 | 42 27 | 13.77 8.42 | 8.64 20.41 | 10.89 10.96 | 10.97 15.34-----------+----------------------+----------------------------------------------------- Total |167 95 | 81 49 |459 219 |383 176 |100 100 |100 100 |100 100 |100 100----------------------------------------------------------------------------------------- chi2(4) = 8.86 (P=.06) | chi2(4) = 9.67 (P=.04) | chi2(4) = 12.55 (P=.01) | chi2(4) = 21.63 (P<.000)-----------+----------------------+-----------------------------------------------------DISSATISFIED WITH JOB-----------+----------------------+------------------------------------------------------ energy | 12 17 | 3 13 | 22 55 | 17 51 | | 12.50 10.24 | 7.89 14.13 | 11.34 11.36 | 11.97 12.23 |-----------+----------------------+------------------------------------------------------ acute< | 15 35 | 2 16 | 30 104 | 22 81 |

| 15.62 21.08 | 5.26 17.39 | 15.46 21.49 | 15.49 19.42 |-----------+----------------------+----------------------------------------------------- acute> | 23 36 | 15 19 | 28 132 | 41 99 | | 23.96 21.69 | 39.47 20.65 | 14.43 27.27 | 28.87 23.74 |-----------+----------------------+------------------------------------------------------ chronic | 38 55 | 10 35 | 89 144 | 45 134 | | 39.58 33.13 | 26.32 38.04 | 45.88 29.75 | 31.69 32.13 |-----------+----------------------+----------------------------------------------------- impair | 8 23 | 8 9 | 25 49 | 17 52 | | 8.33 13.86 |21.05 9.78 | 12.89 10.12 | 11.97 12.47 |-----------+----------------------+----------------------------------------------------- Total | 96 166 | 38 92 |194 484 |142 417 |

254

|100 100 |100 100 |100 100 |100 100 |----------------------------------------------------------------------------------------- chi2(4) = 3.65 (P=.45) | chi2(4) = 11.03 (P=.02) | chi2(4) = 23.66 (P=.000) | chi2(4) = 2.04 (P=.73)----------------------------------------------------------------------------------------------------+----------------------+----------

POOR OR FAIR SELF-RATED HEALTH |other yes |other yes |other yes |other yes |-----------+----------------------------------------------------------------------------- energy | 25 4 | 13 3 | 65 12 | 61 7 | | 21.37 2.76 | 24.53 3.90 | 22.89 3.05 | 24.70 2.24 |-----------+----------------------+------------------------------------------------------

acute< | 29 21 | 12 6 | 92 42 | 66 37 | | 24.79 14.48 | 22.64 7.79 | 32.39 10.66 | 26.72 11.86 |-----------+----------------------+------------------------------------------------------ acute> | 34 25 | 18 16 | 77 83 | 72 68 | | 29.06 17.24 | 33.96 20.78 | 27.11 21.07 | 29.15 21.79 |-----------+----------------------+------------------------------------------------------ chronic | 25 68 | 8 37 | 37 196 | 36 143 | | 21.37 46.90 | 15.09 48.05 | 13.03 49.75 | 14.57 45.83 |-----------+----------------------+------------------------------------------------------ impair | 4 27 | 2 15 | 13 61 | 12 57 | | 3.42 18.62 | 3.77 19.48 | 4.58 15.48 | 4.86 18.27 |-----------+----------------------+------------------------------------------------------ Total |117 145 | 53 77 |284 394 |247 312 | |100 100 |100 100 |100 100.00 |100 100 |------------------------------------------------------------------------------------------ chi2(4) = 52.41 (P=.000) | chi2(4) = 33.72 (P=.000) | chi2(4) = 181.94 (P=.000) | chi2(4) = 138.79 (P=.000)-----------+----------------------+------------------------------------------------------

APPENDIX 6: Kish Tables

APPENDIX 7: Overall Design EffectIn a rapidly changing sociopolitical climate, it is especially important to provide

an accurate image of Russia. Surveys are a major source of information for popular

consumption, published by the media and professional journals, often used by business

and government agencies for foreign and economic policy development (McKeehan,

1993a). It is thus important to see to what extent the picture we have of Russia is

accurate, and to what extent the validity and reliability of reported statistics describing

public opinion and attitudes are affected by components of survey error, such as mode

and interviewer effects.

Although international comparisons between survey modes and error indicate

differential effects upon population parameters, there is a dearth of published data

255

concerning survey modes, survey quality, and survey error in Russia (McKeehan,

1993b). The importance of attending to the quality of survey data by estimating

systematic bias in data collection is particularly acute in nations new to the free flow of

information, such as the Newly Independent States of the Former Soviet Union. A

systematic effect on survey results may be estimated by measurement bias (interviewer,

mode, instrument, and response effects); and by bias of nonobservation (frame

noncoverage, sampling error of subgroup statistic, unit and item nonresponse).

To sift through the differential effect of error sources with any accuracy in

relation to interviewer error and a particular mode of data collection requires an

experimental survey. If multiple survey results are not accessible for a more accurate

meta-analysis of systematic bias, the measurement of specific bias in one survey, may

still be reported as an indicator of the more complex underlying issue of survey quality.

Kish (1962) proposed the (int), the between interviewer variability, as a partial

measure of survey quality. The (int) is the unit free effect on the total variance of a

measure, where (int) depends upon between and within mean square interviewer error

and interviewer workloads (the number of cases assigned to each interviewer), as well

as sample size and total number of interviewers. The model assumes a random

assignment of interviewer workload. The (int) also defines the Design Effect which

reflects the effect of interviewer error on the variance of survey statistics.

The following section discusses two types of measurement bias which affected

the results of the 1991 Moscow Quality of Life survey: interviewer effect, as measured

by the Kish between-interviewer variability, and telephone mode effects. Moscow

University’s research firm, ‘OPINIO’, provided interviewer staff, data collection,

256

supervision, and the telephone sampling frame.

INTERVIEWER EFFECTInterviewer error was calculated according to the Kish additive model of int ,

in a oneway analysis of variance where interviewers are the single factor and

int = {(Va - Vb) m} {[Va - Vb) m] + Vb};

where Va = between interviewer error; Vb = within interviewer error; and m is

workload, the number of interviews assigned to each interviewer. The design effect of

interviewer error on the sample mean was calculated as: deffint = 1+int (m-1).

Interviewer bias has most often been studied as the effect of individual

interviewer attitudes or ideology upon obtained responses. The (int) has been shown

to be a useful indicator of several components of interviewer effect other than

interviewer ideology: workload size; number of interviewers; selection bias due to

interviewers; outliers among interviewers with respect to some questions and with

respect to other interviewers. (int), has been suggested as a measure of survey quality

indicating how specific components of interviewer error may be reduced by survey

design and other components by interviewer training. (int) has often interpreted as

reflecting the professional quality of interviewer performance. Interviewers affect

sampling variance by selecting respondents properly; response error by delivering the

instrument properly; missing values and nonresponse error through appropriate coding

and follow-up procedures, etc. High mean values of (int) may indicate the use of

poorly prepared staff as interviewers, as well as those factors unrelated to individual

subjective behavior, such as random assignment and workload size.

Kish (1987) illustrates that increased sample validity is equivalent to decreased

257

bias: survey quality is increased when variance is decreased by designing larger sample

sizes. Insofar as values of (int) are also related to sample size, it is a measure of

survey quality independent of interviewer effect. Because (int) and deff are calculated

based on the number of interviewers and size of workload, these two factors are useful

in monitoring interviewer effect in survey designs, in addition to sample size and

interviewer training.

The expected values of a cumulative distribution of (int), obtained by Groves

and Magilavy (1986) for selected variables over replicated samples, were found to

decrease when number of interviewers increased; decrease when number of cases

increased; decrease when workload size decreased. The cumulative distribution of (int)

may be seen as a normative range for the values of (int). Any.values outside this range

may be due to outliers: extreme questions or interviewers. About 70% of surveys,

analyzed by Groves, had mean values of (int) <0.04; about 85% of surveys had mean

values <0.06; and about 98% had mean values of (int) <0.08.

The Moscow QOL survey had a mean value of (int)=0.03 for 33 selected

variables. While the range of (int) in the Moscow survey was within the average range

reported by Groves, suggesting the applicability of (int) in cross-cultural contexts, the

design effect indicated an increase in the variance of survey statistics by an average

factor of 2.2. Some implications of this inflated deff follow.

In examining the interaction between interviewer effect and other variables upon

the magnitude of (int), past research has investigated the relevance of several

structural factors in questionnaires and respondents: a.) question form and content:

258

whether open/closed; whether factual/attitudinal and b.) respondent traits: gender,

education, age, region of residence (Groves and Magilavy, 1980).

There have been inconsistent findings with respect to question type. Freeman

and Butler (1976) report increased variance in the mean (int) due to interaction

between interviewer, respondent, and question type: open-ended or emotive questions

(O’Muircheartaigh, 1976,1979), and cognitively ambiguous or difficult questions,

concerning income or employment (Hansen, et al, 1961). Kish (1962) reports no

difference in the mean values of (int) between factual and attitudinal questions or

between open and closed questions. Groves and Magilavy (1986) looked at nine surveys

and also found little effect of question type upon between interviewer variance.

Respondent traits, such as sex and age, have been reported as having equivocal

influence on mean values of (int). Groves found respondent education did not increase

(int) variability as expected for respondents with less than 12 years of education.

However, Freeman and Butler found that younger interviewers had smaller values for

(int) than older interviewers. The largest (int) held for old interviewers paired with old

respondents, whereas (int) values decreased for younger interviewers (31 years) who

had younger respondents. Tucker, as well as Groves, concluded that (int) increased for

older respondents, due mostly to greater nonresponse rates among the elderly in

telephone surveys rather than interviewer ideology. Tucker (1983) also found increased

variance between interviewers by geographic regions when questions were related to

that region. Tucker found that (int) size was the product of interaction between

interviewer, respondent, and item, as well as due to the nonrandom assignment of

259

workload by region.

Kish (1962) postulates an additive model of (int) across subclasses of

respondent traits such as sex and age, assuming random assignment of respondents to

interviewers. If the (int) is significant only for the total sample but not within

respondent subclasses, the value of the total (int) is due to the nonrandom distribution

of subclasses across interviewers and not to subjective interviewer behavior, like

ideology. Kish shows that if interviewer bias is the same for respondent subclasses, the

interviewer bias cancels out across subclasses, and (int) tends towards zero. Thus

(int) is additive across subclasses if bias is constant across subclasses: there is no

interaction between respondent trait and interviewer bias. Interaction occurs when

interviewer effects vary across subclasses. If both the total (int) and subclass (int)

are significantly large, then interviewer bias is related to that subclass for a particular

item. If interviewer bias is significant in subclasses but not in the total, it is too weak to

appear in the total.

The specific distribution of (int) was investigated initially by respondent sex

and age, given a lack of prior published data for comparison on (int) in Moscow

surveys and only tangentially for question form and content.

Table 34 shows (int) and deff for selected variables. The magnitudes of (int)

obtained for the total sample and subclasses lie below the range of “maximum” values

(above which 10% of (int) values are greater) derived by Groves from past research

for similar items. The magnitude of total interviewer variance (int) was reduced when

respondent gender was controlled. However, the persisting large difference of (int)

260

between respondent gender groups indicates an interaction between interviewers and

women respondents for specific items.

Two items, the frequency of changing place of employment (nwork) and the

number of days hospitalized (nhosp), had a larger interviewer bias among women

respondents than in the total sample. The (int) for only two items (age and height)

was greater among men than in the total sample. Interviewer bias was not greatly

reduced in several items for women reporting divorces (ndivorce), chronic conditions

(nchronic), symptoms (nsymptom), or social group memberships (ngroup).

The (int) of three sensitive, attitudinal scales were within the ranges for factual

items, except for a large interviewer bias among women responding to an anomie scale

consisting of 7 items. As a whole, the open-ended items did not have larger total (int)

than closed items.

261

TABLE 39: INTERVIEWER EFFECTS FOR SELECT VARIABLES BY SEX, MOSCOW HRQOLSURVEY, SEPTEMBER, 1991

___________________________________________________________________VARIABLE TOTAL MEN WOMENNAME TYPE * pint deff pint deff pint deff

___________________________________________________________________DEMOGRAPHICSAGE F-O -.0089 0.66 .0027 1.10 -.0099 0.62EDUC F-C .0179 1.69 .0060 1.23 .0108 1.42MARRIAGE F-C -.0035 0.86 -.0034 0.87 .0007 1.03NDIVORCE F-O .0188 1.72 -.0014 0.95 .0155 1.59NWORK F-O .0294 2.13 .0015 1.06 .0363 2.39HEALTH CONDITIONSNCHRONIC F-C .0351 2.35 .0059 1.23 .0323 2.24NSYMPTOM F-C .0477 2.84 .0119 1.45 .0426 2.64HT F-O -.0095 0.63 .0099 1.38 -.0003 0.99WT F-O .0063 1.24 .0085 1.33 .0100 1.39HEALTH A-C .0101 1.39 .0029 1.12 .0075 1.29MEDICAL UTILIZATIONDAYMD F-O .0072 1.28 .0076 1.29 .0081 1.31NHOSP F-O .0293 2.13 -.0078 0.70 .0336 2.29MDSATISF A-C .0447 2.72 .0299 2.16 .0264 2.02ATTITUDESNANOMIE A-C .0866 4.34 .0175 1.68 .0819 4.16NSATISF A-C .0184 1.71 .0038 1.14 .0135 1.52NSPOUSE A-C .0295 2.14 .0151 1.58 .0258 1.99NUMBER OF SOCIAL CONTACTSNGROUP F-C .0412 2.59 .0073 1.28 .0365 2.41NFRIENDS F-O .0209 1.80 .0120 1.46 .0125 1.48NSEENMOS F-O .0398 2.53 .0285 2.09 .0240 1.93___________________________________________________________________*a=attitude; f=factual; o=open; c=closed

262

In looking at the (int) among respondent age groups, Table 35 indicated that,

as compared to the gender groups, many total (int) values were reduced within age

groups often by almost half. This is consistent with the interviewer effects found in

other western surveys. Several (int) disappear completely. The interviewer effect

among men for height (ht) was eliminated when age was controlled. Interviewer bias

among women for specific items (nhosp, nwork, nanomie) persisted within age groups,

and were often concentrated in one specific age group.

TABLE 40: INTERVIEWER EFFECTS FOR SELECT VARIABLES BY AGE, MOSCOW HRQOLSURVEY, SEPTEMBER, 1991

___________________________________________________________________VARIABLE <19-29yrs 30-49yrs 50-64yrs >64yrsNAME pint deff pint deff pint deff pint deff

___________________________________________________________________DEMOGRAPHICSSEX -.0127 0.51 -.0037 0.86 -.0026 0.89 -.0079 0.69EDUC -.0027 0.89 .0051 1.19 .0212 1.82 .0033 1.13MARRIAGE -.0004 0.99 -.0071 0.73 .0029 1.11 -.0039 0.85NDIVORCE .0017 1.07 .0054 1.21 .0149 1.58 .0005 1.02NWORK .0104 1.40 .0117 1.45 .0136 1.53 .0366 2.41*HEALTH CONDITIONSNCHRONIC -.0026 0.89 .0112 1.43 .0222 1.86* .0148 1.57NSYMPTOM .0049 1.19 .0224 1.87* .0209 1.81 .0059 1.23HT -.0088 0.66 .0083 1.32 -.0014 0.94 .0075 0.71WT -.0259 0.01 -.0006 0.98 -.0005 0.98 .0135 1.52HEALTH -.0076 0.71 .0146 1.56 -.0015 0.94 -.0005 0.98MEDICAL UTILIZATIONDAYMD .0071 0.73 .0059 1.23 .0096 1.37 -.0033 0.87NHOSP 0 1.0 .0732 3.82* .0315 2.21 -.0095 0.64MDSATIS .0400 2.54 .0115 1.44 .0103 1.39 .0281 2.08ATTITUDESNANOMIE .0018 1.07 .0604 3.33* .0214 1.82 .0152 1.59NSATISF -.0026 0.89 .0049 1.19 -.0009 0.96 .0042 1.16NSPOUSE -.0011 0.96 .0060 1.23 .0091 1.35 .0208 1.80NUMBER OF SOCIAL CONTACTSNGROUP .0104 1.40 .0209 1.80* .0087 1.34 .0062 1.24NFRIEND -.0039 0.85 .0011 1.04 .0018 1.07 .0126 1.49NSEENMOS .0219 1.85 .0126 1.49 .0311 2.19 .0018 1.07___________________________________________________________________* significant interviewer outlier

263

Those items which had a high interviewer bias among women respondents, also

had a higher interviewer bias among two age groups. Persistent high (int), above 0.02

with concomitant deff above 1.5, were clustered in two age groups: 30-49 years and 50-

64 years.

If the total and subclass (int) are significant (as in nhosp, nwork, nanomie) the

subsample (int) causes the increase in the total interviewer bias. Large (int) among

age and gender groups may be related to question content: ambiguity, sensitivity, and

cultural appropriateness. What may be routinely asked in one country may be looked at

askance in another. Although there is a lack of space here to understand this interaction

between interviewer, respondent subclass, and item more clearly, it is necessary to

check second and third-order interaction levels to isolate the exact components of

interviewer bias.

This Moscow data supported Kish’s additive model of interviewer error, where

(int) is reduced across subclasses if there is no interaction between (int) and

subclass traits. If subclasses, such as age and sex, and interactions between subclasses,

are not taken into account, the interviewer bias may be overestimated by the total

(int). The effect of (int) on subclass interviewer variability decreases proportionately

to a decrease in the size of the subclass, therefore oversampling women in certain age

groups will increase the (int) for women and for those age groups.

The magnitude of the interviewer bias is related to the random assignment of

workloads as well as workload size. There were a total of 51 interviewers in the

264

Moscow Quality of Life Survey: 35% had a response rate of 80% and an average

workload, m or k, of m=n/a= 39.04 (where n= sample size= 1991, a= number of

interviewers= 51); and k=39.54, adjusted for different workload sizes per interviewer

(Kish, 1962). (int) is calculated with m or k. The mean workload, however, obscured

the fact that only a quarter of the interviewers completed nearly half (48%) the sample.

Figure 24 indicates the uneven distribution of respondent workloads among

interviewers: 25% of interviewers had large workloads of more than 41 respondents.

These factors increased interviewer effect, (int).

FIGURE 24: INTERVIEWER WORKLOAD, MOSCOW HRQOL SURVEY, SEPTEMBER, 1991

Table 5: Interviewer Workload (W) by N Interviewers

4

29

5 62

5

05

101520253035

W 1-20 W 21-30 W 31-40 W 41-60 W 61-80 W >80

In nonrandomized workloads, where the interviewer has wide scope in choosing

respondents, the overestimation of (int) due to selection bias may be substantial (Kish,

1962). Since large workloads affect between interviewer variance, maintaining equal

and optimally sized workloads, as well as random assignments, is important in

265

regulating the magnitude of (int). It is likely that several persistently high values of

(int) in the Moscow data were due to specific interviewers with large workloads or

uneven distributions of respondent subclasses.

Interviewers were responsible for randomly selecting respondent within

telephone households. Incorrect procedures resulted in a sample distribution which had

twice as many women 30-49 years old as the Moscow population. The sample also had

46% more women who were 50-64 years old and 45% more women who were 65

years old than in the general Moscow population. The size of respondent subclasses is

unlikely to be related to the distribution of telephones since the oversampling is based

on sex within age groups. A smaller (int) for men in the Moscow data is thus also due

to a smaller number of men in the sample, rather than to different attitudes of

interviewers toward certain questionnaire items and gender.

A ratio I was calculated of interviewers who were significantly different from

more than 80% of interviewers, as compared to the total number of interviewers who

were significantly different from each other at the 0.05 level. The ratio I indicated that

among items with significant (int), among age groups, interviewer outliers were

responsible for much of the variance. (int) , as the between interviewer variance, tends

toward zero when the variance between interviewers equals the variance within

interviewer workload for a specific item (dep var). (int) tends toward 1 when total

variance equals variance between interviewers. In this case, the maximum variance in

the item is accounted for by the variation between interviewers and not by the variation

within workload.

The largest between interviewer variance for specific items was related to a few

266

interviewers who differed significantly from 80% of interviewers, for example, for the

item nwork among respondents over 64 years old (Table 35).

In summary, the data reported in this study support the Kish additive model of

interviewer error as a useful indicator in a cross-cultural survey for isolating several

components of interviewer bias: workload size; number of interviewers; equal

distribution of workloads; random assignment of respondents; interviewer outliers, item

type, inadequate field training and supervision. Mean values for (int) were reduced

when respondent traits, such as age and sex, were controlled. Large values for (int)

within respondent subclasses were related to differential workloads among interviewers;

nonrandom selection of respondents within subclasses by interviewers; interaction

between question type, respondent subclass and interviewer; as well as outliers among

interviewers.

267

APPENDIX 8: Russian language questionnaire

268

APPENDIX 10: Alameda Physical Health ProfileDISABILITY-SEVERE

Q1: Here is a list of activities that people sometimes have trouble with: troublefeeding themselves, trouble dressing themselves, trouble moving around. Do you havetrouble doing any of these things?

1-yes2-no

Q2: Here are two more activities that people sometimes have trouble with:trouble climbing stairs and trouble getting outdoors. Do you have trouble doing either ofthese things?

1-yes2-no

Q3: Are you now unable to work because of some illness or injury?1-yes2-no

Q4: If yes, then for how long?_______________________

DISABILITY-LESSQ5: Have you had to change the kind of work you used to do, or had to cut

down on the number of hours you used to work because of some illness or injury, forthe past 6 months?

1-yes2-no

Q6: Have you had to cut down or stop any other activity you used to do becauseof some illness or injury, for the past 6 months?

1-yes2-no

CHRONIC CONDITIONS Q7:Here is a list of medical conditions that usually last for some time. Have

you had any of these conditions during the past 12 months?7.1 high blood pressure 1-yes 2-no7.2 heart trouble 1-yes 2-no7.3 stroke 1-yes 2-no7.4 chronic bronchitis 1-yes 2-no7.5 asthma 1-yes 2-no7.6 arthritis or rheumatism 1-yes 2-no7.7 epilepsy 1-yes 2-no7.8 cancer 1-yes 2-no7.9 tuberculosis 1-yes 2-no

269

7.10 stomach ulcer or duodenal ulcer 1-yes 2-no7.11 chronic gallbladder trouble 1-yes 2-no7.12 chronic liver trouble 1-yes 2-no7.13 hernia or rupture. 1-yes 2-no

Q8: Here is a list of physical impairments. Do you have any of these?8.1 missing hand/arm 1-yes 2-no8.2 missing foot/leg 1-yes 2-no8.3 trouble seeing (even with glasses) 1-yes 2-no8.4 trouble hearing (even with hearing aid)1-yes 2-no

Q9: Do you have any other medical condition, ailment, or impairment that hasnot been listed so far? Describe it here.

_____________________________________________

SYMPTOMATICQ11: Here is a list of physical ailments. Have you had any of these during

the past 12 months?11.1 frequent cramps in the legs 1-yes 2-no11.2 pain in the heart or tightness

or heaviness in the chest 1-yes 2-no11.3 trouble breathing or shortness of breath 1-yes 2-no11.4 swollen ankles 1-yes 2-no11.5 pains in the back or spine 1-yes 2-no11.6 repeated pain in the stomach 1-yes 2-no11.7 frequent headaches 1-yes 2-no11.8 constant coughing or frequent heavy chest colds 1-yes 2-no11.9 paralysis of any kind 1-yes 2-no11.10 stiffness, swelling, or aching in

any joint or muscles 1-yes 2-no11.11 getting very tired in a short time 1-yes 2-no

NO COMPLAINTS/PHYSICAL-ENERGY Q10: How often are you completely worn out at the end of the day?

10.1 almost every day 1-yes 2-no10.2 sometimes 1-yes 2-no10.3 rarely or never 1-yes 2-no

Q12: Would you say you have more energy or less energy than most peopleyour age?

1-more2-same3-less

270

Q13: How often do you have any trouble getting to sleep or staying asleep?1-almost every day2-sometimes3-rarely or never

Q14: When you have only 4 or 5 hours of sleep during the night how tireddo you feel the next day?

1-a great deal2-somewhat3-not at all

271

APPENDIX 11: Measures of HRQOL (Self-rated Health, Life Satisfaction , LifeHappiness) and Life Choices, Life Chances, and Civic Community

HRQOL:(SELF-RATED HEALTH) Q21: How would you describe your overall

general health?1-excellent2-good3-fair4-poor

(LIFE SATISFACTION) Q122: How often are you completely satisfied withyour life?

1-rarely or never2-sometimes3-often

(LIFE HAPPINESS) Q127: All in all, how happy are you these days?1-not so happy2-pretty happy3-very happy

(DOMAIN SATISFACTION) Q128: Considered everything, how satisfied areyou with your present job?

1-not satisfied2-somewhat satisfied3-very satisfied

LIFE CHOICEHEALTH PRACTICES(CHOICE1) Q26: How many hours of sleep do you usually get a night?

1-less than 7 hours2-7-8 hours3-9 or more hours

(CHOICE2) Q27: How often do you eat breakfast?1-almost every day2-sometimes3-rarely or never

(CHOICE3) Q28: How often do you eat in between your regular meals?1-almost every day2-sometimes3-rarely or never

(CHOICE4) Q29: Here is a list of active things that people do in their freetime. How often do you do any of these things?

272

1-morning exercises 1.1-almost every day 1.2-sometimes

1.3-rarely or never2-exercises at home 2.1-almost every day 2.2-sometimes

2.3-rarely or never3-sports 3.1-almost every day

3.2-sometimes 3.3-rarely or never

4-work on dacha 4.1-almost every day 4.2-sometimes 4.3-rarely or never

5-other ______________________

(CHOICE5) Q30: How many times did you drink wine, beer, or liquor in thepast month?

_______________________

(CHOICE6) Q31: How many drinks of alcohol did you have in the pastmonth?

_______________________(in grams)

(CHOICE7) Q32: If you smoke regularly, how much do you smoke in oneday?

_______________________

(CHOICE8) Q33: How tall are you?_______________________(in meters)

(CHOICE9) Q34: How much do you weigh?_______________________(in kilograms)

PREVENTION(DDS) Q15: When was the last time you went to a dentist just for a general

checkup even though your teeth were not giving you any trouble?1-within past 6 months2-within the past year3-over a year ago4-never

(MD) Q16: When was the last time you went to see a doctor just for ageneral checkup even though you were feeling well and had not been sick?

1-past 6 months2-within the past year3-over a year ago4-never

273

UTILIZATION(ILLMD) Q19: How many times have you seen a doctor because you were not

feeling well in the last 30 days?_______________________

(HOSP) Q20: How many times have you been to a clinic or hospital becauseyou were not feeling well in the last 30 days?

_______________________

(REGMD) Q21A:Where do you receive your regular source of medical care inthe past year?

1-polyclinic2-occupational/job clinic3-medical cooperative4-private physician

(QUALMD) Q24: Where, in your opinion, is the best quality medical careprovided?

1-polyclinic2-occupational/job clinic3-medical cooperative4-private physician

(SATISFMD) Q25: How satisfied are you with the quality of health carewhich you have regularly received in the past year?

1-very satisfied2-satisfied3-somewhat satisfied4-not at all satisfied

ECONOMIC VALUES(PAYMD) Q22: If you could see any doctor which you yourself chose to

see, would you pay to do so, if necessary?1-yes2-no

(FREEMD) Q23: Do you think the government should provide all medicalcare free of charge?

1-yes2-no

CIVIC COMMUNITY

SOCIAL COHESION(NORM1) Q113:With everything in such a state of disorder, it's hard for a

274

person to know where he stands from one day to the next.1-true2-false

(NORM2) Q114:People were better off in the old days when everyone knewjust how he was expected to act.

1-true2-false

(NORM7) Q119:Everything changes so quickly these days that I often havetrouble deciding which are the right rules to follow.

1-true2-false

(MEANING3) Q115:What is lacking in the world today is the old kind offriendship that lasted for a lifetime.

(MEANING4) Q116:The trouble with the world today is that most peoplereally don't believe in anything.

1-true2-false

(MEANING5) Q117:I often feel that many things our parents stood for are justgoing to ruin before our very eyes.

1-true2-false

(MEANING6) Q118:With everything so uncertain these days, it seems asthough anything could happen.

1-true2-false

SOCIAL SUPPORT - MARITAL QUALITY(FAMILY1) Q104:Many women(men) feel that they are not as good wives

(husbands) as they would like to be. How often have you felt this way?1-never or rarely2-sometimes3-often

(FAMILY2) Q105:Does your wife (husband) give you as much understandingas you need?

1-no, not really2-yes,but not completely3-yes, completely

(FAMILY3) Q106:Does your wife (husband) show you as much affection as

275

you would like?

1-more than I like2-as much as I like3-less than I like

(FAMILY4) Q107:Even happily married couples sometimes have problemsgetting along with each other. How often does this happen to you?

1-rarely or never2-sometimes3-often

(FAMILY6) Q109:Do you ever regret your marriage?1-rarely or never2-sometimes3-often

(FAMILY7) Q110:Have you seriously considered separation or divorcerecently?

1-yes2-no

SOCIAL CAPITAL(INFORMAL-NFAM) Q35: How many close friends do you have (people

that you feel at ease with, can talk to about private matters, and can all on for help)?_______________________

(INFORMAL-NFRIENDS) Q37: How many relatives do you have thatyou feel close to?

_______________________

(INFORMAL-NCONTACT) Q38: How many of these friends or relativesdo you see at least once a month?

_______________________

(FORMAL-RELIGION1) Q39: Do you go to any religious services (church,synagogue, evangelical meeting, muslim, etc.)?

1-yes2-no

(FORMAL-RELIGION2) Q40: If yes, how often?_______________________

(FORMAL-HOBBIES1) Q41: Do you belong to any social or recreationalgroup?

1-yes

276

2-no

(FORMAL-HOBBIES2) Q42: If yes, which ones?_______________________

(FORMAL)Q43: Do you belong to any:43.1-labor union 1-yes 2-no43.2-commercial group 1-yes 2-no43.3-professional association? 1-yes 2-no

(FORMAL-CHILDGR) Q44: Do you belong to a group concerned withchildren (such as PTA, Boy Scout group)?

1-yes2-no

(FORMAL-CHARITY) Q45: Do you belong to any group concerned withcommunity betterment, or charity, or any other group?

1-yes2-no

LIFE CHANCES and DEMOGRAPHICS(AGE) Q48: How old are you?

_______________________

(SEX) Q49: What is your sex?1-male2-female

(EDUC) Q50: What is your last educational level?1-elementary2-incomplete secondary3-general secondary4-technical secondary5-incomplete higher (college)6-college degree

(MARITAL) Q51: Have you ever been married? If so, are you now:1-single2-married3-married, but living apart4-divorced5-divorced, but living together6-widowed

(NDIVORCE) Q52: How many times have you been divorced?_______________________

277

(NCHILD) Q53: How many children do you have in your family?_______________________

(OCCUP, PRESTIGE, EMPLOY) Q54: What is your occupation?1-student in technical institute2-student in university3-worker4-amed services5-research worker; college teacher6-engineer-technician; industrial manager7-service professional: physician, lawyer, teacher, etc8-creative professional: writer, journalist, artist, performing arts, etc.9-government worker, official10-social organization worker11-trade and retail worker12-worker in cooperative, joint-venture, real estate13-service worker: registered nurse, accountant laboratory technician, etc.14-temporarily unemployed15-housework16-pensioner

(URBAN AREA) Q55: In which raiyon of Moscow do you live?1-Babushkinski2-Baumanski3-Volgogradski4-Gagarinski5-Dzerzhinski6-Zhelezhno.7-Kalininski8-Kievski9-Kirovski10-Krasnopresnenskyi11-Krasnogvardeiskyi12-Kuibishevskyi13-Kuncevski14-Leningradskyi15-Leninskyi16-Lyublinskyi17-Moskvoretskyi18-Octiabrskyi19-Pervomaiskyi20-Perovskyi21-Proletarskyi22-Sverdlovskyi23-Sevastopolskyi

278

24-Sovetskyi25-Sokolnicheskyi26-Solncevskyi27-Taganskyi28-Timiriazevskyi29-Tushinskyi30-Frunzenskyi31-Horoshevskyi32-Cheremushkinskyi33-Zelenograd

(APT) Q56: What are your living accomodations?1-government apartment2-cooperative apartment3-room in a communal apartment4-place, room in a dormitory5-rent privately

(LANG) Q57: What language do you prefer to speak at home?_______________________

279

REFERENCES

Abel T. and Cockerham, W. C. (1993) Lifestyle or Lebensfuehrung? Criticalremarks on the mistranslation of Weber’s “Class, Status, Party”. The SociologicalQuarterly 34, 551-556.

Adams, P. F., and Benson, V. (1991) Current Estimates from the NationalHealth Interview Survey. National Center for Health Statistics. Vital Health Statistics10 (181), 112

Albrecht, G. (1993) The Social Experience of Disability, In Social Problems:Essay on Health & Medicine, ed. Calhoun & Ritzer, McGraw Hill, Inc. New Jersey,179-196.

Andreev, E., Scherbov, S., and Willekens, F. (1998) Population of Russia: Whatcan we expect in the future? World Development 26, 1939-1955.

Andrews, F. M. and Inglehart, R. F. (1979) The structure of subjective well-being in nine Western societies. Social Indicators Research 6, 73-90.

Andrews, F.M. and Crandall, R. (1976) The validity of measures of self-reportedwell-being. Social Indicators Research 3, 1-19.

Antonovsky, A. (1979) Health, Stress and Coping. Jossey-Bass, San Francisco.Antonovsky, A. (1993) Complexity, Conflict, Chaos, Coherence, Coercion, and

Civility. Social Science and Medicine 37, 969-981.Antonovsky, A. (1993) The structure and properties of the sense of coherence

scale. Social Science and Medicine 36, 725-733.Appels, A., Bosma, V., Grabauskas, A., Gostautas, A., Sturmans, F. (1996) Self-

rated health and mortality in a Lithuanian and a Dutch population. Social Science andMedicine 42, 681-689.

Armer, M. and Grimshaw, A. D. (1974) Comparative Social Research:Methodological Problems and Strategies. John Wiley and Sons, New York.

Becker, C. M. and Bloom, D. (1998) The demographic crisis in the FormerSoviet Union: Introduction. World Development 26, 1913-1919.

Becker, C. M. and Hemley, D.D. (1998) Demographic change in the FormerSoviet Union during the transition period. World Development 26, 1957-1975.

Bell, W. (1957) Anomie, social isolation, and the class structure. Sociometry 20,105-116.

Bellah, R. N. (1973) Editor. Emile Durkheim: On Morality and Society.University of Chicago Press, Chicago.

Belloc, N. B., Breslow, L., Hochstim, J. R. (1971) Measurement of PhysicalHealth in a General Population Survey. Am. J. Epi. 93, 328-336.

Benavides, F.G., Benach, J., Diez-Roux, A.V., Roman, C. (2000) How do typesof employment relate to health indicators? Findings from the second European surveyon working conditions. Journal of Epidemiology and Community Health 54 : 494-501.

Bennett, N. G., Bloom, D., and Ivanov, S.F. (1998) Demographic Implicationsof the Russian Mortality Crisis. World Development 26, 1921-1937.

Berkman, L.F. and Kawachi, I. Editors. (2000) Social Epidemiology. OxfordUniversity Press, New York.

Berkman, L.F., Glass, T. (2000) ‘Social integration, social networks, social

280

support, and health”. In Berkman, LF, Kawachi, I. Eds. Social Epidemiology. NewYork: Oxford University Press: 137-173.

Berkman, L.F., Glass, T., Brissette, I., and Seeman, T. (2000) From socialintegration to health: Durkheim in the new millennium. Social Science and Medicine51, 843-857.

Berkman, L. F. (1985) “The Relationship of Social Networks and Social Supportto Morbidity and Mortality”. In Social Support and Health (Ed. S. Cohen andS.L.Syme), Academic Press, NY: 241-260.

Berkman, L. F. and Breslow, L. (1983) Health and Ways of Living: TheAlameda County Study.: Oxford University Press, New York.

Berkman, L. F. and Orth-Gomer, K. (1996) Prevention of cardiovascularmorbidity and mortality: Role of social relations. In Behavioral Medicine Approaches toCardiovascular Disease Prevention, eds. K. Orth-Gomer and N. Schneiderman, pp. 51-67. Lawrence Erlbaum Associates, New Jersey.

Biryukov, P. P. (1998) Healthy City. City Government of Moscow. PressaPublishing, Moscow, Russia.

Blalock, H. M. (1982) Conceptualization and Measurement in the SocialSciences. Sage Publications, Beverly Hills, California.

Blau, P. (1980) Contexts, units, and properties in sociological analysis. InSociological Theory and Research, ed. H.M. Blalock, pp. 51-66. The Free Press, NewYork.

Blaxter, M. (1989) A comparison of measures of inequality in morbidity. InHealth Inequalities in European Countries, ed. J. Fox, pp. 199-230. Gower Publishing,Aldershot, UK.

Blaxter, M. (1990) Health and Lifestyles. Tavistock/Routledge, London.Blazer, Harley D. (1996) Russia’s Missing Middle Class. M.E.Sharpe, Armonk,

New York.Bobadilla, J., Costello, C.A., and Mitchell, F. (1997) Premature Death in the

New Independent States. Washington, D.C., National Academy Press.Bobak, M. and Marmot, M. (1996) East-West mortality divide and its potential

explanations: proposed research agenda. BMJ, 312, 421-425.Bobak, M., Pikhart, H., Hertzman, C., Rose, R., Marmot, M. (1998)

Socioeconomic Factors, Perceived Control and Self-Reported Health in Russia. ACross-sectional Survey, Social Science and Medicine 47, 269-279.

Bobak, M., Chenet L., Leon D., McKee, M. Shkolnikov, V. (1998) Patterns ofSmoking in Russia. Studies in Public Policy, Center for the Study of Public Policy, 299,27-36.

Bobak, M., Chenet, L., Hertzman, C., Leon, D., McKee, M., Marmot, M.,Pikhart, H., Rose, R., and Shkolnikov, V. (1998) Surveying the Health of Russians.Studies in Public Policy Number 299. Center for the Study of Public Policy, Universityof Strathclyde, Glasgow, Scotland.

Bodrova, V. (1995) Reproductive behaviour of Russia’s Population in theTransition Period. Berichte des Bundesinstituts fuer ostwissenschaftliche undinternationale Studien. Koeln, Germany.

Bosma, H., Marmot, M. G., Hemingway, H., Nicholson, A. C., Brunner, E., andStansfeld, S. A. (1997) Low job control and risk of coronary heart disease in Whitehall

281

II (prospective cohort) study. BMJ 314, 1-22.Bottomore, T. (1975) Competing paradigms in macrosociology. Annual Review

of Sociology 5, 191-201.Bourdieu, P. (1984) Distinction. Cambridge: Harvard University Press.Bourdieu, P. (1990) The Logic of Practice. Stanford: Stanford University Press.Branch, L. G. and Jetta, A. M. (1984) Personal Health Practices and Mortality

among the Elderly. AJPH, 74, 1126-1132.Breslow, L. (1972) A quantitative approach to the World Health Organization

definition of health: physical, mental, and social well-being. International Journal ofEpidemiology 1, 347- 355.

Brislin, R. W. (1986) The wording and translation of research instruments. InField Methods in Cross-Cultural Research, ed. W.J. Lonner and J.W. Berry, pp.137-164. Sage Publications, Beverly Hills, California.

Brock, B. M., Haefner, D. P., Noble, D. S. (1988) Alameda County Redux:replication in Michigan. Preventive Medicine 17, 483-495.

Bruce, A. (1989) An Illustrated Companion to the First World War . London,Michael Joseph Ltd., p. 86.

Bruhn, J.G. and Wolf, S. 1979. The Roseto Story. University of Oklahoma Press,Norman, OK

Brunswick, A.F., Messeri, P. (1986) Drugs, lifestyle, andhealth: a longitudinalstudy of urban black youth. American Journal of Public Health 76: 52-57.

Brunswick, A.F., Messeri, P. (1999) Life stage, substance use and health declinein a community cohort of urban African Americans. Journal of Addictive Diseases 18:53-71.

Brunswick, A.F., Messeri, P. (1983-1984) Causal factors in onset ofadolescents’ cigarette smoking: a prospective study of urban black youth. Advances inAlcohol Substance Abuse 3: 35-52.

Carlson, P. and Vågerö, D. (1998) The social pattern of heavy drinking inRussia during transition. EJPH 8, 280-285.

Carmines E. G. and Zeller, R. A. (1979) Reliability and Validity Assessment.Sage University Paper series on Quantitative Applications in the Social Sciences, 07-017. Sage Publications, Beverly Hills and London.

Carr-Hill, R. (1990) The measurement of inequities in health: lessons from theBritish experience. Social Science and Medicine 31, 393-404.

Centers for Disease Control (1991) Consensus set of health status indicators forthe general assessment of community health status - United States. MMWR 40, 449-463.

Centers for Disease Control (1992) Nutritional Needs Surveys Among theElderly-Russia and Armenia. MMWR 41, 809-811.

Centers for Disease Control (1992) Public Health Assessment-RussianFederation, 1992. MMWR 41,89-91.

Chamberlain, K. (1988) On the structure of subjective well-being. SocialIndicators Research 20, 581-604.

Chenet, L. et al., (1998) "Alcohol and Cardiovascular Mortality in Moscow:New Evidence of a Causal Association," Journal of Epidemiology and CommunityHealth, 52, 772-774.

Chenet, L. et al., (1998) "Deaths from Alcohol and Violence in Moscow: Socio-

282

Economic Determinants," European Journal of Population, 14, 19-37.Cloward, R. (1959) Illegitimate means, anomie, and deviant behavior. American

Sociological Review 24, 164-176.Cockerham, W. C. (1995) Medical Sociology. Prentice-Hall, Englewood Cliffs,

New Jersey.Cockerham, W. C., Abel, T., and Guenther, L. (1993) Max Weber, Formal

Rationality, and Health Lifestyles. The Sociological Quarterly 34, 413-435.Cockerham, W.C. (1997) The social determinants of the decline of life

expectancy in Russia and Eastern Europe. Journal of Health and Social Behavior 38:117-130.

Cohen, A. K. (1959) The Study of Social Disorganization and Deviant Behavior.In Sociology Today: Problems and Prospects, eds. R.K.Merton, L. Broom, and L.S.Cottrell, Jr. Harper Torchbooks, New York.

Cohen, S., Wills, T.A., (1985) Stress, social support, and the bufferinghypothesis. Psychological Bulletin 98: 310-357.

Coleman, J. S. (1986) Social Theory, Social Research, and a Theory of Action.AJS 91, 1309-1335.

Colhoun, H., Ben-Shlomo, Y., Dong, W., Bost, L. and Marmot, M. G. (1997)Ecological analysis of collectivity of alcohol consumption in England: importance ofaverage drinker. BMJ 314, 1164-1168.

Culyer, A. J. (1989) Commodities, characteristics of commodities,characteristics of people, utilities and the quality of life. In The Quality of Life:Perspectives and Policies, ed. S. Baldwin, C. Godfrey, and C. Propper. Routledge &Kegan Paul, London.

Dahlgren, G. and Whitehead, M. (1992) Policies and straegies to promoteequity in health. World Health Organization Regional Office for Europe, Copenhagen.

Dahrendorf, R. (1979) Life Chances. Chicago: University of Chicago Press.Davis, C. A. (1993) Eastern Europe and the Former USSR: An Overview.

RFE/RL Research Report 2, 31-34.Davis, C. A. (1993) The Former Soviet Union. RFE/RL Research Report 2, 35-

43.Davis, E. E., Fine-Davis, M., Meehan, G. (1982) Demographic determinants of

perceived well-being in eight European countries. Social Indicators Research 10, 341-358.

Deutscher, I. (1974) Asking Questions Cross-culturally. In Institutions and thePerson. Ed., H. S. Becker, B. Geer, D. Riesman, and R. S. Weiss, pp. 318-341.. AldinePublishing Co., Chicago, Illinois.

Diener, E. (1984) Subjectie Well-Being. Psychological Bulletin 95, 542-575.Diener, E., Suh, E. M., Lucas, R. E., Smith, H. L. (1999) Subjective Well-Being:

Three decades of progress. Psychological Bulletin, in press.Diez-Roux, A.V., F.J.Nieto, C.Muntaner, H.A.Tyroler, G.W.Comstock,

E.Shahar, L.S.Cooper, R.L.Watsom, and M.Szklo (1997) Neighborhood environmentsand coronary heart disease: a multilevel analysis, American Journal of Epidemiology,146, 48-63.

Diez-Roux, A.V., F.J.Nieto, L.Caulfield, H.A.Tyroler, R.L.Watson, M.Szklo,

283

(1999) Neighborhood differences in diet: the Atherosclerosis Risk in Communities(ARIC) Study. Journal of Epidemiology and Community Health 53: 55-63.

Diez-Roux, A.V., (1998) Bringing Context Back into Epidemiology: Variablesand Fallacies in Multilevel Analysis. American Journal of Public Health 88: 216-222.

Diez-Roux, A.V., (2002) Invited commentary: places, people, and health.American Journal of Epidemiology 155: 516-519.

Diez-Roux, A.V. (2001) Investigating neighborhood and area effects on health.American Journal of Public Health 91: 1783-1789.

Diez-Roux, A.V., Merkin, S.S., Arnett, D., Chambless, L., Massing, M., Nieto,F.J., Sorlie, P., Szklo, M., Tyroler, H.A., Watson, R.L. (2001) Neighborhood ofresidence and incidence of coronary heart disease. New England Journal of Medicine345 : 99-106.

Diez-Roux, A.V. (2000) Multilevel analysis in public health research. AnnualReview of Public Health 21: 171-192.

Diez-Roux, A.V., Link, B.G., Northridge, M.E. (2000) A Multilevel analysis ofincome inequality and cardiovascular disease risk factors. Social Science and Medicine50: 673-687.

Diez-Roux, A.V. On genes, individuals, society, and epidemiology. (1998)American Journal of Epidemiology 148: 1027-1032.

Diez-Roux, A.V., Nieto, F.J. (1997) Epidemiology 8: 459-461.Diez-Roux, A.V., Nieto, F.J., Tyroler, H.A., Crum, L.D., Szklo, M. (1995)

Social Inequalities and atherosclerosis. The atherosclerosis risk in communities study.American Journal of Epidemiology 15: 960-972.

Diez-Roux, A.V., Benach, J., Tapia, J.A. (1994) Should policyrecommendations be excluded from epidemiologic research papaers? Epidemiology 5:637-638.

Digest of the Post-Soviet Press (1998) 50, 30, 4.Doeglas, D., Suurmeijer, T., Briancon, S., Moum, T., Krol, B., Bjelle, A.,

Sanderman, R., and Van den Heuvel, W. (1996) An international study on measuringsocial support: interactions and satisfaction. Social Science and Medicine 43, 1389-1397.

Dohrenwend, B. P., Levav, I., Shrout, P. E., Schwartz, S., Naveh, G., Link, B.G. Skodol, A., Stueve, A. (1992) Socioeconomic status and psychiatric disorders: thecausation-selection issue. Science 255, 946-952.

Donabedian A, Elinson J, Spitzer W, Tarlov A. (1987) Advances in HealthAssessment Conference Discussion Panel. Journal of Chronic Disease 40, (Suppl.1),183S-191S.

Duncan, C. Jones, K. Moon, G. (1993) Do Places Matter? A Multilevel Analysisof Regional Variations in Health-Related Behaviour in Britain, Social Science andMedicine 37, 725-733.

Duncan, C., Jones, K., Moon, G. (1996) Health-related behaviour in context: Amultilevel modelling approach. Social Science and Medicine 42, 817-830.

Duncan, C., Jones, K., Moon, G. (1998) Context, Composition, andHeterogeneity: Using multilevel models in health research. Social Science and Medicine46, 97-117.

Duncan, C., Jones, K., Moon, G. (1999) Smoking and deprivation: are there

284

neighborhood effects? Social Science and Medicine 48, 497-505.Duncan, O. D. (1975) Does money buy satisfaction? Social Indicators Research

2, 267-274.Durkheim, E. (1951) Suicide: A Study in Sociology. (Translated by John A.

Spaulding and George Simpson, Edited with an Introduction by George Simpson). TheFree Press, New York.

Durkheim, E. (1961) Moral Education: A Study in the Theory and Application ofthe Sociology of Education. (Translated by Everett K. Wilson and Herman Schnurer,Edited with a new Introduction by Everett K. Wilson). The Free Press, New York.

Durkheim, E. (1966) The Rules of Sociological Method. The Free Press, NewYork.

Easterlin, R. A. (1973) Does money buy happiness? The Public Interest 30, 3-10.

Easterlin, R. A. (1974) Does economic growth improve the human lot? Someempirical evidence. In P. A. David and M. W. Reder, ed., Nations and households ineconomic growth. Academic Press, New York.

Elder, J. W. (1973) Problems of cross-cultural methodology: Instrumentationand interviewing in India. In Comparative Social Research: Methodological Problemsand Strategies. Ed., M. Armer and A. D. Grimshaw, pp. 119-144, Wiley, New York.

Elder, J. W. (1976) Comparative Cross-national Methodology. Annual Review ofSociology 1 , 209-229.

Elinson, J. (1980) Medical Sociology: Theoretical Underdevelopment and SomeOpportunities. In Sociological Theory and Research: A Critical Appraisal. ed. H.M.Blalock, pp. 373-382. The Free Press, New York.

Ellman, M. (1994) The increase in death and disease under ‘katastroika’.Cambridge Journal of Economics, 18, 329-355.

Eyles, J. (1990) How significant are the spatial configurations of health caresystems? Social Science and Medicine 30, 157-164.

Farrelly, M.C., Healton, C.G., Davis, K.C., Messeri, P., Hersey, J.C., Haviland,M. (2002) Getting to the truth: evaluating national tobacco countermarketingcampaigns. American Journal of Public Health 92: 901-907.

Ferrie, J. E., Shipley, M. J., Marmot, M. G., Stansfeld, S., and Smith, G. D.(1998) The Health Effects of Major Organizational Change and Job Insecurity. SocialScience and Medicine 46, 243-254.

Feshbach, M. (1999) "Dead Souls" Atlantic Monthly 1, 26-27.Feshbach, M. and Friendly, Jr. A. (1992) Ecocide in the USSR, Basic Books,

New York.Feshbach, M. (1995) Ecological Disaster: Cleaning up the Hidden Legacy of the

Soviet Regime. The Twentieth Century Fund Press, New York.Feshbach, M. (1995) Editor-in-chief. Environmental and Health Atlas of Russia

, Paims Publishing House, Moscow.Fiscella, K. and Franks, P. (1997) Does psychological distress contribute to

racial and socioeconomic disparities in mortality? Social Science and Medicine 45,1805-1809.

Fiscella, K. and Franks, P. (1997) Poverty of income inequality as predictors ofmortality: longitudinal cohort study. BMJ 314, 1724-1728.

285

Flynn, B. C. and Dennis, L. I. (1995) Documenting the Urban Health Situation:Tools for Healthy Cities. WHO Regional Office for Europe, Copenhagen.

French, R. A. (1995) Plans, Pragmatism and People. UCL Press, London.Fordyce, M. W. (1988) A review of reearch on the happiness measuress: a sexty

second index of happiness and mental health. Social Indicators Research 20, 355-381.Freeman, J. and Butler, E. W. (1976), Some sources of interviewer variance in

surveys. Public Opinion Quarterly 40, 79-91.Frijters, P. and van Praag, B. M. S. (1995) Estimates of Poverty Ratios and

Equivalence Scales for Russia and parts of the Former USSR. Foundation for EconomicResearch, Faculty of Economics and Econometrics, University of Amsterdam,Tinbergen Institute, TI95-149.

Galea, S., Ahern, J., Tardiff, K., Leon, A.C., Vlahov, D. (2002) Drugs andfirearm deaths in New York City, 1990-1998. Journal of Urban Health 79: 70-86.

Galea, S., Ahern, J., Resnick, H., Kilpatrick, D., Bucuvalas, M., Gold, J.,Vlahov, D., (2002) New England Journal of Medicine 28: 982-987.

Gallup, G. H. (1976) Human needs and satisfactions: a global survey. PublicOpinion Quarterly 40, 459-467.

Garcia P. and McCarthy, M. (1995) Measuring Health: A step in thedevelopment of city health profiles. World Health Organization Regional Office forEurope, Copenhagen.

Giddens, A. (1971) A typology of suicide. In The Sociology of Suicide. (Ed.), A.Giddens, pp. 97-120. Frank Cass, London.

Giddens, A. (1973) The Class Structure of Advanced Societies. Barnes andNobles, New York.

Giddens, A. (1984) The Constitution of Society: Outline of the Theory ofStructuration. Polity Press, Cambridge.

Giddens, A. (1991) Modernity and Self-Identity. Stanford University Press,Stanford, California

Glatzer, W. and Zapf, W. (1984) Lebensqualitaet in der Bundesrepublik(Quality of Life in West Germany). Campus Verlag, Frankfurt/Main, New York.

Goldman, N., Korenman, S., Weinstein, R. (1995) Marital Status and Healthamong the Elderly. Social Science and Medicine 40, 1717-1730.

Goskomstat (1991) Statistical Yearbook of the Russian Federation, Moscow,Russia.

Goskomstat (1996) Statistical Yearbook of the Russian Federation, Moscow,Russia.

Goskomstat (1996a) Administrtivniye Okruga Goroda Moskvy, 1995. Moscow,1996: 179.

Goskomstat (1998) Gosudarstvenniyi Doklad: O sostoyannii okruzhauscheisredy Moskvy v 1997. Moscow, 1998: 117.

Gould, M. I. and Jones, K. (1996) Analyzing perceived limiting long-termillness using U.K. census microdata. Social Science and Medicine 42, 857-869.

Grand, A., Grosclaude, P., Bocquet, H., Pous, J., Albarede, J. L. (1988)Predictive Value of Life Events, Psychosocial Factors and Self-Rated Health onDisability in an Elderly Rural French Population, Social Science and Medicine 27,1337-1342.

286

Groves, R. M. and Magilavy, L. J. (1980) Estimates of interviewer variance intelephone surveys.Proceedings of Survey Research Methods Section, AmericanStatistical Association, 622-627.

Groves, R.M. and Magilavy, L. J. (1986) Measuring and Explaining InterviewerEffects in Centralized Telephone Surveys. Public Opinion Quarterly, 50, 251-266.

Guralnik, J. M. and Kaplan, G. A. (1989) Predictors of Healthy Aging:Prospective Evidence from the Alameda County Study. American Journal of PublicHealth 79,703-708.

Haan M., Kaplan, G. A., and Camacho, T. (1987) Poverty and healthprospective evidence from the Alameda County Study. American Journal ofEpidemiology 125, 989.

Haan, M.N., Kaplan, G.A., and Syme, S.L. (1989) Socioeconomic status andhealth: Old observations and new thoughts. In Pathways to Health: The Role of SocialFactors, edited by J.P. Bunker, D.S. Gomby, and B.H.Kehrer, The Henry J. KaiserFamily Foundation, Menlo Park, California: 76-135.

Haber, L. (1973) Some Parameters for Social Policy in Disability: A Cross-National Comparison, MMFQ Health and Society, Summer, 319-340.

Hamilton,. E. L. (1993) Social Areas Under State Socialism: The Example ofMoscow. Doctoral Dissertation, Department of Geography, Columbia University.

Hancock, T. and Duhl, L. (1986) Promoting Health in the Urban Context. WHOHealthy Cities Papers No.1. FADL, Copenhagen.

Hansen, M. H., Hurwitz, W. N., and Bershad, M. A. (1961) Measurement errorsin censuses and surveys. Bulletin of the ISI, 38, 351-374.

Harkin, A. M., Anderson, P., Goos, C. (1997) Smoking, drinking, and drugtaking in the European region. WHO Regional Office for Europe, Copenhagen.

Harris, K. (1997) Social stress and trauma: synthesis and spatial analysis. SocialScience and Medicine 45, 1251-1264.

Healton, C., Messeri, P., Reynolds, J., Wolfe, C., Stokes, C., Ross, J., Fliont, K.,Robb, W., Farrelly, M. (2000) Tobacco use among middle and high school students –United States, 1999. MMWR Morbidity Weekly Reports 49: 49-53.

Henkel, R. A. (1976) Tests of Significance. Sage University Paper series onQuantitative Applications in the Social Sciences, 07-004. Sage Publications, BeverlyHills and London.

Hertzman, C. (1995) Environment and health in Central and Eastern Europe.World Bank, Washington D.C.

Herzlich, C. and Pierret, J. (1987) Illness and self in society. Translated byElborg Forster. John Hopkins University Press, Baltimore, Maryland.

Hochstim, J. R., Athanasopoulos, D. A., larkins, J. H. (1968) Poverty area underthe microscope. AJPH 58, 1815-1827.

Hochstim, J. R. (1970) “Health and ways of living – The Alameda County,California, Population Laboratory. In The Community as a Epidemiologic Laboratory ,ed., Kessler, I.I. and Levin, M.L. p.149, Johns Hopkins University Press, Baltimore,Md.

House, J. S. and Robbins, C. (1983) Age, psychological stress and health. InAging and Society: Selected Reviews of Recent Research, ed. M.W.Riley, B.B.Hess, andK.Bond, Erlbaum, Hillsdale, New Jersey.

287

House, J.S., Landis, K. R., and Umbertson, D. (1988) Social relationships andhealth. Science 214, 540-545.

Idler, E. L. (1979) Definitions of Health and Illness and Medical Sociology.Social Science and Medicine 13A , 723-731.

Idler, E. L. (1992) Self-assessed health and mortality: a review of studies. InInternational Review of Health Psychology, vol. I. ed., M.S.Leventhal and M.Johnson,John Wiley, New York.

Idler, E. L., Angel, R. J. (1990) Self-Rated Health and Mortality in theNHANES I Epidemiologic Follow-Up Study. American Journal of Public Health 80,446-452.

Idler, E. L., Yael, B. (1997) Self-rated health and mortality: a review of twenty-seven community studies. Journal of Health and Social Behavior 38, 21-37.

Illsley, R. (1990) Comparative review of sources, methodology, and knowledge.Social Science and Medicine 31, 229-236.

Illsley, R. (1990) Comparative review of sources, methodology, and knowledge.Social Science and Medicine 31, 229-236.

Inglehart, R. (1977) The Silent Revolution: Political Change among WesternPublics. Prineton University Press, Prineton, New Jersey.

Inglehart, R. (1990) Cultural Shift in Advanced Industrial Societies. PrincetonUniversity Press, Princeton, New Jersey.

Inglehart, R. (1997) Modernization and Postmodernization: Cultural, Economicand Political Change in 41 Societies. Princeton University Press, Princeton.

International Monetary Fund (1999) International Financial Statistics, 51, 1.Istomin, V. (1998) Healthy City Movement: Health of Muscovites.

Memorandum, Moscow, Russia.Istomin, V. (1997) Editor. My Moscow: A healthy person in a healthy city.

Pressa Publishing, Moscow, Russia.Jette, A. M. (1994) How Measurement Techniques Influence Estimates of

Disability in Older Populations, Social Science and Medicine 38, 937-942.Jones, A., Connor, W. D., Powell, D. E. Editors. (1991) Soviet Social Problems,

Westview Press, Boulder, Colorado.Jones, K. and Moon, G. (1987) Health, Disease, and Society: A Critical Medical

Geography. Routledge & Kegan Paul, London and New York.Kaasik, T., Andersson, R., and Hoerte, L. (1998) The effects of Political and

economic transitions on health and safety in Estonia: an Estonian-Swedish comparativestudy. Social Science and Medicine 47, 1589-1599.

Kaplan, G.A. (1996) People and Places: Contrasting Perspectives on theAssociation between Social Class and Health. International Journal of Health Services26:507-519.

Kaplan, G. A. and Keil, J. E. (1993) Socioeconomic factors and cardiovasculardisease: a review of the literature. Circulation 88, 1973-1976.

Kaplan, G. A. Pamuk, E. R., Lynch, J. W., Cohen, R. D., Balfour, J. L. (1996)Inequality in income and mortality in the United States: analysis of mortality andpotential pathways. BMJ 312, 999-1003.

Kawachi, I. And Berkman, L.F. (2001) Social Ties and Mental Health. Journalof Urban Health : 458-467.

288

Kawachi, I. and Kennedy, B. P. (1997) Health and social cohesion: why careabout income inequality? BMJ 314, 1037-1040.

Kawachi, I. and Kennedy, B. P. (1997) The Relationship of income inequality tomortality: does the choice of indicator matter? Social Science and Medicine 45, 1121-1127.

Kawachi, I., Kennedy, B.P., Lochner, K., and Prothrow-Stith, D. (1997) Socialcapital, income inequality, and mortality. AJPH 87, 1491-1498.

Kennedy, B. P., Kawachi, I., Brainerd, E. (1998) The Role of Social capital inthe Russian Mortality Crisis. World Development Report, 26, 2029-2043.

Kish, L. (1962) Studies of Interviewer Variance for Attitudinal Variables.Journal of the American Statistical Association, 57, 91-115.

Kish, L. (1987) Statistical Design for Research. New York: Wiley.Klugman, J. (1997) Editor. Poverty in Russia: Public Policy and Private

Responses. The World Bank, Washington, D.C.Kohn, M. L. (1989) Cross-National Research in Sociology. Sage Publications,

Newbury Park, California.Kohn, M.L. (1987) Cross-national Research as an analytic strategy. American

Sociological Review 52, 12, 713-731.Konrad, A. E. (1998) On the state of the environment of Moscow. Review of the

state report of 1996. Ecology of the capital is of special interest of the Moscow CityGovernment, Moscow, Russia.

Kreft, I. and De Leeuw, J. (1998) Introducing Multilvel Modeling. SagePublications, London.

Kruiderink, A. (1999) Forword, In Transition 1999. UN Development Program,Geneva.

Kunst, A. E., Groenhof, F., Mackenbach, J. P., EU Working Group onSocioeconomic Inequalities in Health (1998) Mortality by Occupational Class AmongMen 30-64 years in 11 European Countries. Social Science and Medicine 46, 1459-1476.

Kunst, A. E., Mackenbach, J. P. (1994) The Size of Mortality DifferencesAssociated with Educational Level in Nine Industrialized Countries. AJPH, 84, 932-937.

Kunst, A. E., Geurts, J. J. M., van den Berg, J. (1995) International variation insocioeconomic inequalities in self-reported health. Journal of Epidemiology andCommunity Health, 49, 117-123.

Kunst, A. E., Groenhof, F., Mackenbach, J. P., and the EU Working Group onSocioeconomic Inequalities in Health (1998) Mortality by occupational class amongmen 30-64 years in 11 European countries. Social Science and Medicine 46, 1459-1476.

Langford, I. H. and Bentham, B. (1996) Regional variations in mortality rates inEngland and Wales: an analysis using multi-level modeling. Social Science andMedicine 42, 897-908.

Lasker, J. N., Egolf, B. P., and Wolf, S. (1994) Community Social Change andMortality. Social Science and Medicine 39, 53-62.

Lavrakas, Paul J. (1987) Telephone Survey Methods: Sampling, Selection, andSupervision. Applied Social Research Methods Series, vol. 7, Sage Publications,Newbury Park, California.

289

LeClere, F. B., Rogers, R. G., Peters, K. (1998) Neighborhood social contextand racial differences in women’s heart disease mortality. JHSB 39, 91-107.

Lee, L. A. (1995) Results of the Moscow longitudinal Survey: Housing andEconomic Characteristics. The Urban Institute, USAID Shelter Cooperation Program,Moscow, Russia (mimeo).

Lee, L. A. and Romanik, C. T. (1995) The Moscow Longitudinal HouseholdSurvey 1992-1995. The Urban Institute, USAID Shelter Cooperation Program, Moscow,Russia (mimeo).

LeGrand, J. (1982) The Strategy of Equality: Redistribution and the SocialServices. George Allen & Unwin, London.

LeGrand, J. (1987) Equity, health, and health care. In Three Essays on Equity.Welfare State Project: Discussion Paper 23. London School of Economics and PoliticalScience, London.

Lemert, C. (1993) Editor. Social Theory: The multicultural and classic readings.Westview Press, Boulder, Colorado.

Levada, Y. Democratic Disorder and Russian Public opinion Trends in VCIOMSurveys, 1991-1995. Studies in Public Policy, 255, Center for the Study of PublicPolicy, Glasgow, Scotland.

Lilienfeld, A. M. (1976) Foundations of Epidemiology. Oxford UniversityPress, New York.

Link, B. G. and Phelan, J. (1995) Social conditions as fundamental causes ofdisease. JHSB, extra issue, 80-95.

Litva, A. and Eyles, J. (1994) Health or healthy: why people are not sick in asouthern Ontarian town. Social Science and Medicine 39, 1083-1091.

Lomas, J. (1998) Social Capital and Health: Implications for Public Health andEpidemiology. Social Science and Medicine 47, 1181-1188.

Lynch, J. (2000) Income inequality and health: expanding the debate. SocialScience and Medicine 51, 1001-1005.

Lynch, J. W., Kaplan, G. A., Salonen, J. T. Why do poor people behave poorly?Variation in Adult health behaviours and psychosocial characteristics by stages of thesocioeconomic lifecourse. Social Science and Medicine 44, 809-819.

Macintyre, S. (1997) The Black report and beyond: what are the issues? SocialScience and Medicine 44, 723-745.

Macintyre, S., Hunt, K., Sweeting, H. (1996) Gender Differences in Health: AreThings Really as Simple as They Seem? Social Science and Medicine 42, 617-624.

Mackenabch, J. P. (1994) Socioeconomic inequality in health in theNetherlands: impact of a five year research programme. British Medical Journal 309,1487-1491.

Mackenbach, J. P., Kunst, A. E. (1997) Measuring the Magnitude ofSocioeconomic Inequalities in Health: An Overview of Available Measures Illustratedwith Two Examples from Europe. Social Science and Medicine 44, 757-771.

Madison B. (1988) The Soviet Pension System and Social Security for the Aged.In State and Welfare USA/USSR: Contemporary Policy and Practice. ed., G.W. Lapidusand G. E. Swanson, pp. 143-212, Institute of International Studies, University ofCalifornia, California.

Malmstrom, M., Sundquist, J., Johansson, S.E. (1999) Neighborhood

290

Environment and Self-Reported Health Status: A Multilevel Analysis, AmericanJournal of Public Health 89: 1181-1186.

Marmot, M. G. (1998) Improvement of social environment to improve health.The Lancet 351, 57- 60.

Marmot, M. G. et al. (1991) Health Inequalities among British Civil Servants:The Whitehall P Studies. Lancet, 337, 1387-1393.

Marmot, M. G., Smith, G. D., Stansfeld, S., Patel, C., North, F., Head, J., White,I., Brunner, E., and Feeney, A. (1991) Health inequalities among British civil servants:the Whitehall II study. Lancet 337, 1387-1393.

Marmot, M. and Wilkinson, R.G. (1999) Social Determinants of Health. Oxford:Oxford University Press.

Marmot, M., Siegrist, J., Theorell, T., and Feeney, M. (1999) “Health and thepsychosocial environment at work”. In M. Marmot and R.G.Wilkinson, SocialDeterminants of Health, Oxford: Oxford University Press: 105-113.

Matteson, D.W., Jeffrey, A.B., and Marshall, J.R. (1998) Infant mortality: Amultilevel analysis of individual and community risk factors. Social Science andMedicine 47, 1841-1854.

Maximova, T. M., Kakorina, E.P., Korolkova, T.A., Ismailova, D.I., Lushkina,N.P. (1997) Illness and Medical Treadtment of Disability. N.A. Semashko Scientific-Research Instituteof Social Hygiene, Economics, and Health Administration, Moscow,Russia.

McAuley, A. (1994) Social welfare in transition: What happened in Russia.World Bank, Research paper no.6, Transition Economics Division, Policy ResearchDivision, Washington, D.C.

McDowell, I. and Newell, C. (1987) Measuring Health: A Guide to RatingScales and Questionnaires. Oxford University Press, New York.

McIsaac, S. J. and Wilkinson, R. G. (1997) Income disribution and cause-specific mortality. AJPH 7, 45-53.

McKee, M., Bobak, M., Rose, R., Shkolnikov, V., Chenet, L. and Leon, S.(1998) Pattern of smoking in Russia. Tobacco Control, 7: 22-26.

McKeehan, I. V. (1992) Editor. National Conflicts On The Eve Of The Coup.Russian Sociology 31.

McKeehan, I. V. (1993a) Editor. Nationalistic Stereotypes of Russians AmongStudents and Instructors in Higher Education. Russian Education and Society 35.

McKeehan, I. V. (1993b) Editor. Students' Attitudes Towards Socialism,Nationalism, and Perestroika. Russian Education and Society 35.

McKeehan, I. V. and Campbell, R. (1992) "Russian Health Survey", Letter tothe Editor. Moscow News 27, July 5-12, 13.

McKeehan, I. V. (1991) Quality of Life Indicators in Comparative US-USSRContext, Institute for the Study of the USA and Canada, USSR Academy of Sciences,Moscow, Russia, mimeo.

McKeehan, I. V. (1994) Meta-Analysis of Soviet Survey Research Methods.American Statistical Association: Proceedings on Survey Methods, vol. II, pp. 1172-1177, Illinois, USA.

McKeehan, I. V. (1995a) Quality of Life Among Moscow Elderly.Sotsiologicheskiye Issledovaniya, 3, 109-114.

291

McKeehan, I. V. (1995b) "Planning of National Primary Health Care andPrevention Programs: The First Health Insurance Law of Russia, 1991-1993". InEugene Gallagher and Jan Subedi, ed. Global Perspectives on Health Care. Prentice-Hall, Englewood Cliffs, New Jersey, 174-197.

McKeehan, I. V. (1995c) Measurement of Interviewer Error in a MoscowTelephone Survey. American Statistical Association and Royal Statistical Association:Proceedings of the International Conference on Survey Measurement and ProcessQuality, pp. 46-51, Bristol, United Kingdom.

McKeehan, I.V. (1998) Health Inequality among the Moscow Elderly: A Genderhealth Profile from the Moscow Lifestyle and Health-related Quality of Life Survey.Proceedings of the International Congress for Global Health Progress, pp. 102,UNESCO, Paris.

McKeehan, I.V., Campbell, R., and Tumanov, S.V. (1993) Lifestyles and Habitsinfluencing the health of Muscovites before the implementation of the Health InsuranceLaw of Russia of 1991-1993. Sotsiologicheskiye Issledovaniya 3, 45-49.

McKeon, R. (1941) The Basic Works of Aristotle. Random House, New York.Mechanic D. (1975) The Comparative Study of Health Care Delivery Systems.

Annual Review of Sociology 1, 43-65.Mechanic, D. (1995) Sociological Dimensions of Illness, Social Science and

Medicine 41, 1207-1216.Mechanic, D. and Volkart, E. H. Stress, Illness behavior and the sick role.

(1961) American Sociological Review 26, 51.Messeri, P., Workman, S., Saunders, C., Francis, C. (1997) The application of

meta-analysis in assessing racial differences in the effects of antihypertensivemedication. Journal of the National Medical Association 89: 477-485.

Mezentseva, E., and Rimachevskaya, N. M. (1990) "The Soviet Country Profile:Health of the USSR Population in the 70s and 80s - An Approach to a ComprehensiveAnalysis," Social Science and Medicine 31, 871.

Michalos, A. C. (1980) Satisfaction and happiness. Social Indicators Research10, 385-422.

Michalos, A. C. (1983) Satisfaction and happiness in a rural northern resourcecommunity. Social Indicators Research 13, 225-252.

Mikhalev, V. (1996) Social Security in Russia under Economic Transformation,Europe-Asia Studies, 48, 5-25.

Mirowsly, J. and Ross, C. E. (1989) Social Causes of Psychological Distress,Aldine de Gruyter, New York.

Molinari, et al., (1998) Social Science and Medicine 47, 1113-1120.Moon, G. (1990) Conceptions of space and community in British health policy.

Social Science and Medicine 30, 165-171.Mooney, G. H. (1983) Equity in health care: confronting the confusion. Effective

Health Care 1, 179-185.Mooney, G. H. (1987) What does equity in health mean? World Health

Statistical Quarterly 40, 150-162.Mooney, G. H. and McGuire, A. (1987) Distributive justice with special

reference to geographical inequality and health care. In Health and Economics, ed. A.Williams. Macmillan, London.

292

Moore, W. (1999) Russia's new poor - a daily struggle for survival. The RussiaJournal, 8, (www.russiajournal.com).

Morris, R. and Carstairs, V. (1991) Which deprivation? A comparison ofselected deprivation indices. Journal of Public Health Medicine 13, 318-326.

Mosgorstat, Dept of Public Health. (1991) Moscow City Council, Bureau ofMedical Statistics, 21 March 1991. Mean Moscow Indicators of the MedicalDepartment of Mosgorispolkom 1989-1990. Moscow, Russia.

Moskomstat (1989, 1994, 1995, 1996, 1997, 1998) Administrative Districts ofthe City of Moscow. Moscow City Committee for Government Statistics, Moscow,Russia.

Moskomstat (1990) Moskva v Tsifrakh. Financy I Statistika, Moscow, Russia.Moskovskiy Komsomolets (28 July 1999) "Aids Infection Cases Become 12

Times More Frequent in Moscow and Its Environs".Mossey, J. M., Shapiro, E. (1982) Self -rated health: a predictor of mortality

among the elderly. American Journal of Public Health 72, 800-808.Nagi, S. Z. (1991) Disability concepts revisited: Implications for prevention. In

Disability in America, ed. Institute of Medicine, National Academy Press, WashingtonD.C.

Notzon, F. C., Komarov, Y. M., Ermakov, S. P., et al., (1998) Causes ofdeclining life expectancy in Russia. JAMA 279, 793-800.

O’Muircheartaigh, C. A. (1976) Respone errors in an attudinal sample survey.Quality and Quantity, 10, 97-115.

O’Muircheartaigh, C. A. (1979) Response effects. In C.A.O’Muircheartaigh andC.Payne (Eds.), The Analysis of Survey Data: Exploring Data Structure, New York:Wiley.

O’Reilly, P. (1989) Methodological issues in social support and social networkresearch. Social Science and Medicine 26, 863-873.

Palosuo, H., Uutela, A., Zhuravleva, I., Lakomova, N. (1998) Social patterningof ill health in Helsinki and Moscow. Social Science and Medicine 46, 1121-1136.

Palosuo, H., Zhuravleva, I., Uutela, A., Lakomova, N., Shilova, L. (1995)Perceived Health, Health-related Habits and Attitudes in Helsinki and Moscow: AComparative Study of Adult Populations in 1991. Publications of the National PublicHealth Institute, Helsinki.

Patrick, D. L. and Bergner, M. (1990) Measurement of Health Status in the1990s. Annual Review of Public Health 11, 165-183.

Pereira, J. (1990) Comparative review of sources, methodology, and knowledge.Social Science and Medicine 31, 413-420.

Peter, R. and Siegrist, J. (1997) Chronic Work Stress, Sickness Absence, andHypertension in Middle Managers: General or Specific Sociological Explanations?Social Science and Medicine 45, 1111-1120.

Phillips, M., Feachem, R.G.A., Murray, C.J.L., Over, M., Kjellstrom, T. (1993)Adult Health: A Legitimate Concern for Developing Countries, AJPH 83, 1527-1536.

Pokrovsky, A. (1 July 1999) RIA Novosti, Moskva (www.mos.ru ).Popkin, B. M., Zohoori, N., Baturin, A. (1996) The Nutritional Status of the

Elderly in Russia, 1992 through 1994, AJPH, 86, 355-360.Preston, S.H. (1977) Mortality Trends. Annual Review of Sociology 3, 163-178.

293

Public Health Committee of the Mayorate of St. Petersburg, The St. PetersburgCentre of State Sanitary-Epidemiological Surveillance, Institute of Medical and SocialProblems and Management, WHO Healthy Cities Project, and Nottingham School ofPublic Health. (1995) St. Petersburg: The Health of the City. St. Petersburg, Russia.

Puska, P., Matilainen, T., Jousilahti, P., Korhonen, H., Vartiainen, E.,Pokusajeva, S., Moisejeva, N., Uhanov, M., Kallio, I., Artemjev, A. (1993)Cardiovascular risk factors in the Republic of Karelia, Russia, and in North Karelia,Finland. Int. J. Epidemiol. 22, 1048-1055.

Putnam, R. D., Leonardi, R., and Nanetti, R. Y. (1993) Making democracywork: civic traditions in modern Italy. Princeton University Press, Princeton, NewJersey.

Report to the Moscow City Government. (1999) On the state of the environmentin the city of Moscow, 1997. Mosgomstat, Moscow, Russia. (www.mos.ru).

Rice, N., Carr-Hill, R., Dixon, P., Sutton, M. (1998) The influence ofhouseholds on drinking behaviour: A multilevel analysis. Social Science and Medicine46, 971-979.

Rimashevskaya, N. M. (1997) Poverty trends in Russia: A Russian perspective.In Poverty in Russia: Public Policy and Private Responses, ed. J. Klugman, pp. 119-132. The World Bank, Washington, D.C.

Robert, S. A. (1998) Community-level socioeconomic status effects on adulthealth. Journal of Health and Social Behavior 39, 18-37.

Roberts, R.F., Kaplan, G.A., and Camacho, T.C. (1990) Psychological distressand mortality: Evidence from the Alameda County Study. Social Science and Medicine31, 527-536.

Robine, J. M., Ritchie, K. (1991) Healthy Life Expectancy: evaluation of globalindicator of change in population health. BMJ 302, 457-460.

Rose, G. (1985) Sick individuals and sick populations. International Journal ofEpidemiology 14, 32-38.

Rose, G. (1995) The Strategy of Preventive Medicine. Oxford University Press,Oxford, UK.

Rose, R. (1994) Getting by without government: everyday life in a stressfulsociety. Studies in Public Policy Number 227. Center for the Study of Public Policy,University of Strathclyde, Glasgow, Scotland.

Rose, R. (1998a) Getting things done in an anti-modern society: social capitalnetworks in Russia. Studies in Public Policy Number 304. Center for the Study ofPublic Policy, University of Strathclyde, Glasgow, Scotland.

Rose, R. (1998b) Gettings things done with social capital: New RussiaBarometer VII. Studies in Public Policy Number 303. Center for the Study of PublicPolicy, University of Strathclyde, Glasgow, Scotland.

Rose, R. (1999) What does social capital add to individual welfare? Anempirical analysis of Russia. Studies in Public Policy Number 318. Center for the Studyof Public Policy, University of Strathclyde, Glasgow, Scotland.

Rose, R. and Tikhomirov, E. (1995) Trends in the new Russia barometer, 1992-1995. Studies in Public Policy, 256, Center for the Study of Public Policy, Glasgow,Scotland.

Russian Longitudinal Monitoring Survey (1999) (www.cpc.unc.edu/ projects/

294

rlms/ links.html).Schneider, M. (1975) The quality of life in large American cities: objective and

subjective indicators. Social Indicators Research 1, 495-510.Schuessler, K. F. and Fisher, G. A. (1985) Quality of life research and

sociology. Annual Review of Sociology 11, 129-149.Schwartz, S. (1994) The fallacy of the ecological fallacy: the potential misuse of

a concept and the consequences. AJPH 84, 819-824.Schwartz, S., Diez-Roux, A.V., Diez-Roux, R., (2001) Commentary: causes of

incidence and causes of cases – a Durkheimian perspective on Rose. InternationalJournal of Epidemiology 30: 435-439.

Seeman, M. and Lewis, S. (1995) Powerlessness, Health, and Mortality: ALongitudinal Study of Older Men and Mature Women. Social Science and Medicine41, 517-525.

Seeman, T. and Berkman, L.F. (1988) Structural Characteristics of SocialNetworks and their relationship with Social Support in the Elderly: Who ProvidesSupport. Soc. Sci. Med., 26, 737-749.

Sen, A. (1981) Public action and the quality of life in developing countries.Oxford Bulletin of Economics and Statistics 43, 287-319.

Shin, D. C. and Johnson, D. M. (1978) Avowed happiness as an overallassessment of the quality of life. Social Indicators Research 5, 475-492.

Shkolnikov, V. (1996) Harvard, 32.Shkolnikov, V., Meslé, F., and Vallin, J. (1995) "La Crise Sanitaire en Russie:

Tendences récentes de l&#8217;espérance de vie et des causes de décès de 1970 à1993," Population 4-5, 907-943.

Shkolnikov, V. M. and Meslé, F. (1996) The Russian epidemiological crisis asmirrored by mortality trends. In Russia’s Demographic Crisis, ed. J. DaVanzo, pp. 113-160. National Academy of Sciences, Washington D.C., USA.

Shkolnikov, V. M., Cornia, G. A., Leon, D. A., Meslé, F. (1998) Causes of theRussian mortality crisis: Evidence and interpretations, World Development, 26, 1995-2011.

Shkolnikov, V. M., Leon, D. A., Adamets, S., Andreev, E., and Deev, A., (1998)Educational level and adult mortality in Russia: An analysis of routine data 1979 to1994, Social Science and Medicine 47, 357-369.

Siegrist, J. (1987) Impaired quality of life as a risk factor in cardiovasculardisease. Journal of Chronic Disease 40, 457-458.

Siegrist, J. (1989) Steps towards explaining social differentials in morbidity: thecase of West Germany. In Health Inequalities in European Countries, ed. J. Fox, pp.353-371. Gower Publishing, Aldershot, UK.

Siegrist, J. (1993) Sense of coherence and sociology of emotions. Social Scienceand Medicine 37, 978-979.

Siegrist, J. (1995) Social differentials in chronic disease: What can sociologicalknowledge offer to explain and possibly reduce them? Social Science and Medicine 41,1603-1605.

Siegrist, J. (1996) Adverse health effects of high-effort/low-reward conditions.Journal of Occupational Health Psychology 1, 2-41.

Siegrist, J. (1996) Soziale krisen und gesundheit [Social crises and health].

295

Hogrefe, Goettingen, Germany.Siegrist, J. (1997) Chronic work stress is associated with atherogenic lipids and

elevated fibrinogen in middle-aged men. Journal of Internal Medicine 242, 149-156.Siegrist, J. and Matschinger, H. (1989) Restricted status control and

cardiovascuflar risk. In A. Steptoe and A. Appels, eds., Stress, personal control, andhealth, pp. 65-82. Wiley, Chichester, England.

Siegrist, J., Peter, R., Cremer, R., and Seidel, D. (1997) Chronic work stress isassociated with atherogenic lipids and elevated fibrinogen in middle-aged men. Journalof Internal Medicine 242, 149-156.

Siegrist, J., Peter, R., Junge, A., Cremer, P. and Seidel, D. (1990) Low statuscontrol, high effort at work and ischemic heart disease: Prospective evidence from blue-collar men. European Heart Journal 13 (Suppl. D), 89-95.

Smelser, N. J. (1996) Social Science and Social Problems: The next century.International Sociology 11, 275-290.

Smith, D. M. (1989) Urban Inequality under Socialism. Cambridge UniversityPress, New York .

Smith, D. M. (1994) Geography and Social Justice, Blackwell, Oxford, UK.Snijders, T.A.B. and Bosker, R.J. (1999) Multilevel Analysis: An introduction to

basic and advanced multilevel modeling. Sage Publication, Beverly Hills, California.Social Security Administration (1995) Social Security Programs Throughout the

World. Washington, D.C.: Office of Research and Statistics.Sokolov, V. M. (1998) Moscuvites discussion ecology problems of the capital.

Pulse 4, Moscow, Russia.Spitzer, W. (1987) State of Science 1986: quality of life and functional status as

target variables for research, Journal of Chronic Disease, 40, 465-471.Srole, L. (1956) Social integration and certain corollaries: an exploratory study.

American Sociological Review 21, 709-716.Stassen, M. A. and Staats, S. R. (1988) Hope and happiness: a comparison of

some discrepancies. Social Indicators Research 20, 45-58.State Report on the state of the environment of Moscow in 1992 (1993)

Moscow, Russia (www.mos.ru).State Report on the state of the environment of Moscow in 1996 (1997)

Moscow, Russia (www.mos.ru).Strumpel, B. (1974) Subjective Elements of Well-Being. Organization for

Economic Co-operation and Development, Paris.Sullivan, (1989) Sullivan’s method for calculating life expectancy free from

disability. World Health Statistics Quarterly 42, 148.Sullivan J. L. and Feldman, S. (1979) Reliability and Validity. Sage University

Paper series on Quantitative Applications in the Social Sciences, 07-015. SagePublications, Beverly Hills and London.

Susser, M. (1994) The Logic in Ecological: I. The Logic of Design. AJPH 84,825-829.

Susser, M. (1994) The Logic in Ecological: II. The Logic of Design. AJPH 84,830-835.

Susser, M. 1973. Causal Thinking in the Health Sciences: Concepts andStrategies of Epidemiology. Oxford University Press, New York.

296

Theorell, T. (1992) The psychosocial environment, stress, and coronary heartdisease. In M. Marmot and P. Elliott (Eds.), Coronary heart disease epidemiology, pp.256-273. Oxford University Press, Oxford, England.

Thoits, P. (1995) Stress, Coping, and Social Support Processes: Where Are We?What Next? JHSB, extra issue, 53-79.

Thornberry, O. T., Wilson, R. W., Golden, P. (1986) Health Promotion Data forthe 1990 Objectives, Estimates from the National Health Interview Survey of HealthPromotion and Disease Prevention, United States, 1985. Advance Data From Vital andHealth Statistics. US Dept of Health and Human Services; Washington, DC.

Traugott, M. (1978) Editor. Emile Durkheim On Institutional Analysis. TheUniversity of Chicago Press, Chicago.

Travis, R. (1993) The MOS Alienation scale: an alternative to Srole’s Anomiascale. Social Indicators Research 28, 71-91.

Treml, V. G. (1982) Alcohol in the USSR: A Statistical Study . Duke UniversityPress, Durham, North Carolina.

Tucker, C. (1983), Interviewer Effects in Telephone Surveys. Public OpinionQuarterly, 47, 84-95.

Tulchinsky, T. H., Varavikova, E. A. (1995) Addressing the epidemiologicaltransition in the Former Soviet Union: Strategies for health system and public healthreform in Russia. AJPH, 86, 313-320.

U.N. Monthly Bulletin of Statistics (1999) 53, 10.U.N. Population Division (1999) World Population Prospects: The 1998

Revision, U.N. Population Division, New York.U.N. Population Division (1983) Model Life Tables for Developing Countries,

United Nations, New York.VanDevanter, N., Parikh, N.S., Cohall, R.M., Merzel, C., Faber, N., Litwak, E.,

Gonzales, V., Kahn-Krieger, S., Messeri, P., Weinberg, G., Greenberg, J. (1999)Factors influencing participation in weekly support groups among women completingHIV/STD intervention program. Women Health 30: 15-34.

Veenhoven R. (1990) Inequality in happiness: inequality in countries comparedbetween countries. Paper presented as the 12th World Congress of Sociology, July,1990, Madrid, Spain.

Veenhoven R. (1991) Is Happiness Relative? Social Indicators Research 24, 1-34.

Veenhoven, R. (1987) Cultural bias in ratings of perceived life quality. SocialIndicators Research 19, 329-334.

Veenhoven, R. (1994) Is happiness a trait? Social Indicators Research 32, 101-160.

Veenhoven, R. (1995) The cross-national pattern of happiness: test ofpredictions implied in three theories of happiness. Social Indicators Research 34, 33-68.

Veenhoven, R. (1996) Developments in satisfaction research. Social IndicatorsResearch 37, 1-46.

Vella, Venanzio. (1997) “Health and Nutritional Aspects of Well-Being”. InPoverty in Russia: Public Policy and Private Responses, ed., J. Klugman, pp. 91-118.Economic Development Institute of the World Bank, EDI Development Studies, TheWorld Bank, Washington D.C.

297

Verbrugge, L. (1990) The iceberg of disability. In The Legacy of Longevity.Health & Health Care in Later Life, ed., S. Stahl, Sage, Pennsylvania.

Verbrugge, L.M. and Jette A.M. (1994) The disablement process. Social Scienceand Medicine 38, 1.

Verheij, R.A. (1996) Explaining urban-rural variations in health: a review ofinteractions between individual and environment. Social Science and Medicine 42, 923-935.

Vishnevsky, A.G. (1995) Naselenie Rossii, 1995. Center of Demography andEcology of People. Russian Academy of Sciences, Moscow, Russia.

Vlahov, D., Galea, S., Resnick, H., Ahern, J., Boscarino, J.A., Bucuvalas, M.,Gold, J., Kilpatrick. D. (2002) Increased use of cigarettes, alcohol, and marijuanaamong Manhattan, New York, residents after the September 11th terrorist attacks.American Journal of Epidemiology 155: 988-996.

Vlahov, D., Galea, S., Frankel, D. (2002) New York City 2001: reaction andresponse. Journal of Urban Health 79: 2-5.

Walberg, P., McKee, M., Shkolnikov, V., Chenet, L., Leon, D.A. (1998)Economic change, crime, and mortality crisis in Russia: regional analysis. BMJ, 317,312-318.

Walt, G. (1998) Globalisation of international health. Lancet 351, 434-437.Ware, Jr., J. (1987) Standards for validating health measures: definition and

content. Journal of Chronic Diseases 40, 473-480.Watson, P. (1995) Explaining risingmortality among men in Eastern Europe.

Social Science and Medicine 41, 923-934.Watson, P. (1998) Health difference in Eastern Europe: preliminary findings

from the Nowa Huta study. Social Science and Medicine 46, 549-558.Weber, Max. (1947) The Theory of Social and Economic Oganizations. Edited

with an Inroduction by Talcott Parsons). The Free Press, New York.Weber, Max. (1949) The Methodology of the Social Sciences (Tranlated and

Edited by Edward A. Shils and Henry A. Finch) The Free Press, New York.Webster, P. and Price, C. (1997) Healthy Cities Indicators: analysis of data from

cities across Europe. World Health Organization Regional Office for Europe,Copenhagen.

Webster, P. Review of the “city health Profiles” produced by WHO-HealthyCities- do they present information on health and its determinants and what are theirperceived benefits? Journal of Epidemiology and Community Health 53: 125-127.

Werner, A. and Campbell, D. T. (1970) Translating, working throughinterpreters, and the problem of decentering. In A Handbook of Method in CulturalAnthropology. Ed., R. Naroll and R. Cohen, pp. 398-420, The Natural History Press,Garden City, New York.

Werner, O., and Campbell, D.T. (1970) Translating, Working ThroughInterpreters, and the Problems of Decentering. In A Handbook of Method in CulturalAnthropology, ed. R. Naroll and R. Cohen, pp. 398-420. The Natural History Press,New York.

West, P. (1991) Rethinking Health Selection. Social Science and Medicine 3,374-383.

White, S. (1996) Russia Goes dry: Alcohol, State, and Society. Cambridge

298

University Press, Cambridge, England.Whitehead, M. (1990) The concepts and principles of equity and health. World

Organization Regional Office for Europe, Copenhagen.Whitehead, M. (1990) The Concepts and principles of equity and health. WHO,

Regional Office for Europe, Copenhagen, 5.Whyte, W. F. and Braun, R. R. (1978) On Language and Culture. In Institutions

and the Person. Ed., H. S. Becker, B. Geer, D. Riesman, and R. S. Weiss, pp. 119-138.Aldine Publishing Co., Chicago, Illinois.

Wilkinson, R. G. and Marmot, M. (Editors) (1998) The Solid Facts: SocialDeterminants of Health. WHO, Geneva.

Wilkinson, R. G. (1997) Health inequalities: relative or absolute materialstandards? BMJ 314, 591-595.

Wilkinson, R. G., Kawachi, I. And Kennedy, B.P. (1998) Mortality, the socialenvironment, crime, and violence. Sociology of Health and Illness 20: 578-597

Wilkinson, Richard G. (1996) Unhealthy Societies: The Afflictions of Inequality.Routledge, New York, New York.

Willis, C. L. (1982) Durkheim’s concept of anomie – some observations.Sociological Inquiry 52, 106-113.

World Bank (1994) Russia: Social Protection during transition and beyond.Volume II: Annexes, Human Resources Division, Country Departments III, Europe andCentral Asia Region, Report No.11748-RU, Washington, D.C.

World Bank (1998) World Development Report 1998/99 Oxford UniversityPress, New York, pp. 190-193.

World Health Organization (1985) Targets for health for all. WHO RegionalOffice for Europe, [European Health for All Series No.1], Copenhagen.

World Health Organization (1994) Concern for Europe’s Tomorrow. WHORegional Publications, European Series, No. 53, Copenhagen.

World Health Organization (1998) Health in Europe 1997. Regional Committee48th Session, Copenhagen.

World Health Statistics Annual (1996) Copenhagen.World Health Statistics Annual (1993) Copenhagen.Zdavookhraneniye Rossiyskoy Federatsii 1999, (14 August 1999) no. 1 pp. 3-

18; translated as "State Report on Public Health in 1997" in U.S. Foreign BroadcastInformation Service (FBIS), FBIS-SOV-1999-0405, 105-127.

Zdravookhraneniye Rossiyskoy Federatsii (May-June 1998) no. 3.Zohoori, N., Mroz, T.A., Popkin, B., Glinskaya, E., Lokshin, M., and Mancini,

D. (1998) Monitoring the economic transition in the Russian Federation and itsimplications for the demographic crisis – the Russian Longitudinal Monitoring Survey.World Development 26, 1977-1993.

Zubova L.G., Kovaleva, N.V., and Mitiaeva, L.I. (1992) The problems ofquality of life in the eyes of the population. In Nationalism and National Attitudes onthe Eve of the Coup, ed. McKeehan, I.V., Sociological Research 31 (6), 71-90.


Top Related