"False"
Skip to content
printicon
Main menu hidden.

Old age, health and inequality: mortality from a life course perspective in Sweden 1986-2009

Research project During recent years, the question if income inequality in itself is detrimental for health has been a hot topic. In this proposed project, we will explore the interrelations between income inequality and health outcomes among the elderly Swedish population from a life course perspective.

In this project we study the relationship between income inequality and health among elderly 55+) in Sweden 1986-2009. Some previous studies at national level indicate that income inequality in itself is detrimental for health. The connection is less clear at lower geographical levels where it even can be an advantage to live in unequal neighbourhoods if the economic resources are stronger there. In this project we can distinguish if the possible relationship is explained by the individual income or if there is an independent effect of income inequality. We apply a life course perspective where we consider previous living conditions and the residential career of people.

Head of project

Sören Edvinsson
Professor emeritus
E-mail
Email

Project overview

Project period:

2013-01-01 2015-12-31

Funding

The Swedish Research Council, 2013-2015: SEK 3,000,000

Participating departments and units at Umeå University

Centre for Demographic and Ageing Research (CEDAR)

Project description

Old age, health and inequality – mortality from a life course perspective in Sweden 1986-2009

Purpose and aim

A multitude of studies demonstrate that socio-economic conditions have strong impact on health, even in economically advantaged welfare societies. In October 2011, the World Conference on Social Determinant of Health in Rio de Janeiro adopted the political
declaration on Social Determinants of Health to outline measures to close the inequality gaps and move knowledge into policy and practice (WHO 2012). Measuring the health gaps within and across countries and identifying social determinants that influence health gradients,
provide a basis to develop action against adverse conditions in order to alleviate such inequalities (Friel and Marmot 2011). One aspect of this is the potential health risk of living in regions and nations characterized by socio-economic inequality as suggested by Wilkinson
and Pickett (2009), a question that has led to a lively public and academic debate. They argue that inequality in society has an independent effect on health and adds to increasing health
differentials.
In this proposed project, we will explore the interrelations between income inequality and health outcomes among the elderly Swedish population. The data for our project is available at a high level of geographical resolution, which will allow us to disentangle the complex interplay between socio-economic inequality at various levels and health. More specifically, we will examine:
(i) how residence in equal or unequal areas, from a life course perspective, affects subsequent mortality;
(ii) the impact of different levels of area aggregation (counties, local labor market regions, municipalities, neighborhoods or urban/rural) on the association between area inequality
and health;
(iii) the effects of income inequality on health from different economic measures (i.e. Gini coefficient, absolute income, and area-level average income); and (iv) the interplay between individual level socio-economic and health determinants (such as role of migration) with contextual determinants at area level (such as residential segregation) for health outcomes.

Survey of the field
The idea that unequal societies have higher mortality and more health problems came in the forefront of the academic debate in the 1990’s when Wilkinson compared health in nations with different levels of inequality. He restricted his study to the wealthiest nations in the
world with the motivation that it is only at a certain level of economic development that we can expect to find any correlation between health and inequality. While the ideas of Wilkinson and his colleague Pickett has been widely spread, the research has also been
criticized, in particular their interpretation of what determines the effect of inequality on health (Mellor and Milyo 2003; Wagstaff and van Doorslaer 2000). In his editorial, Mackenbach (2002) argued that the evidence of the association between income inequality and health was slowly dissipating. Later studies have shown that this is not yet the case. Two scientific questions are still to be solved - whether unequal societies are less successful in health and, if this is the case, what determines these differences. Much of the earlier studies
were based on aggregate-level data. These studies could demonstrate interesting patterns of the relation, but they were restricted in the sense that causality was difficult to prove.
The evidence that health is worse in unequal societies is quite strong. The question is rather how to explain this. Some researchers point to the “concavity effect” (or absolute income effect) where the relation between absolute income and health is concave with less return in health advantages at higher income levels. This leads to better health in a society where income and wealth among the poorest are improved. Wilkinson and Pickett, as well as other researchers, argue for the “pollution effect” that refers to an independent contextual income inequality effect. They postulated that the independent effect of inequality on health in high-income countries, not the pure effects of economic and material conditions, can be explained by psychosocial conditions (Wilkinson and Pichett 2009). Relative deprivation, i.e. being in a lower social position, affects the psychosocial wellbeing and result in distress and could be
detrimental for health, as demonstrated in a large number of empirical studies (Kondo et al. 2009; Kondo 2012). Subramanian and Kawachi (2004) identify three main pathways of social inequality and health. First, the structural pathway means that economic inequality leads to residential segregation, which in its turn leads to negative health consequences. The second pathway emphasizes the role of social cohesion, sometimes conceptualized as social capital. The third is the policy pathway (also called a neomaterial pathway) where economic
inequality operates through general social and health-related policies.
Previous research provides so far an ambiguous picture, partly because the majority of researches are either based on cross-sectional data, fixed geographical units, or small study population resulting in lack of statistical power to detect any possible effects of inequality on health (Wagstaff and van Doorslaer 2002; Lynch 2004; Subramanian and Kawachi 2004; Wilkinson and Pickett 2007; Kondo 2011). In a meta-analysis, Kondo et al. (2009) find a modest adverse effect of income inequality on self-rated health and mortality. A recent crossnational study by Karlsson et al (2010) also supports these results. The association seems to be stronger in America than in Europe and on higher geographical levels (Lynch et al 2004).
Previous studies in Sweden have not been able to verify any effect of income inequality on health and mortality. Gerdtham and Johansson (2004) find – as expected - a clear effect of absolute income on all cause mortality in Swedish municipalities, but fail to find any effect of neither mean community income (relative income) nor income inequality. In a multilevel analysis, Henriksson et al (2006) find an association between income distribution and mortality in Swedish municipalities when analyzed at the individual level, but this
disappeared when applying a multilevel model. Other Swedish studies of relevance for this topic is Stjärne et al (2006) that study neighborhood effects on myocardial infarcts, Åberg and Lundberg (2007) who analyze the role of relative deprivation and social comparison on health, and Fritzell et al (2004) who find a clear curvi-linear function between self-rated health and income on micro-level data.

Project description

Theory
In their agenda for future research, Subramanian and Kawachi (2004) point to some important issues that need to be developed. This includes better ways of sorting out the determinants and
pathways of economic inequality to poor health, the need to analyze the effect on different geographical levels, and the importance to perform longitudinal studies in order to better handle causality and to develop multilevel models for this research. This planned project aims to consider these issues.
The theoretical background for Wilkinson’s and Pickett’s postulate is that socio-economic inequality has an independent effect of health and that this is caused by psychosocial conditions. As mentioned above, this has however been questioned. It could equally well be
explained by the concativy effect. In this project we can analyze this in more detail.
Furthermore, we apply a life course approach. Much of previous research has not been able to analyze the long-term effects of inequality, a problematic restriction as the effect does not
come immediately. Previous studies show that inequality has its strongest effects with a time lag up to 15 years (Subramanian and Kawachi 2004, p 86). By using a life course approach, we can control for changes in economic conditions. Another major issue in the study of health and social inequality is the risk of reverse causation. Health conditions influences social position and can be a component leading to residential segregation. The best way to consider this and to disentangle the possible causal pathways is to apply a longitudinal perspective, where migrations of the studied population can be followed. This is of special interest when having a focus on health in old age. Economic situation and social position among elderly is strongly determined by previous income and occupation. According to the cumulative advantage hypothesis (Dannefer 2003) differences increase throughout life and thus become even more apparent in old age. Other researchers, on the other hand, argue that differences become smaller due to selection effects and/or stronger impact of biology in old age.
Wilkinson and Pickett (2009) argue that the inequality effect is not equally visible in various types of health outcomes. Only diseases or causes of death that have a social gradient within a country were strongly related to inequality in society. According to them, the inequality effect is stronger in causes of death or diseases that are more sensitive to social position. Some diseases, such as breast cancer that does not have strong social gradient, are not influenced by inequality.
Another problem relate to what geographical level is investigated. The effect has usually been less clear at levels within nations. This causes Wilkinson and Pickett (2007) to argue that “these processes are structured primarily on national level). On the other hand, if the negative influence of inequality would be an effect of social comparison this ought to have impact also on lower levels. The different impact of socioeconomic inequality on increased risks for death and disease at different levels of aggregation is probably caused by different mechanisms. On a national level, but also on county or municipality level, political systems are important
through resource redistribution and measures favoring belonging and coherence in society.
Different health policies are important in this respect. On local or neighborhood level psychosocial factors may be weaker or even irrelevant. The neighborhood is not necessarily the unit where the social position is defined. Living in homogeneous low-income
neighborhoods could even add to the individual risks. Residents in poor homogenous and segregated areas are more likely to have weaker social networks and to lack crucial social and economic resources. We know from previous research that the outcomes of these kinds of
analyses are very sensitive to the delimitations of the spatial units of analysis (Fotheringham & Rogerson 1993). And as Fritzell (2005, p 295) has pointed out, “… it is plausible, to my mind, that the distribution of income at various levels of aggregation are inversely related to each other and thereby could also have opposite health effects. For example, if the context is a small area, i.e. a neighborhood, a low degree of income inequality could be a sign of high economic segregation which in turn might lead to higher income inequality at a higher level of aggregation.”

Data
A unique combination of longitudinal socio-economic and health data, the Linnaeus database which is now available at Umeå University (Malmberg et al 2010), allows for assessment of the impact of socio-economic inequality on health at regional or local level. Previous studies have often relied on cross-sectional or aggregated data (see Lynch et al. 2004; Subramanian and Kawachi 2004; Wilkinson and Pickett 2007), which limit the potential to fully understand the underlying mechanism between inequality and health. This phenomenon needs to be analyzed carefully and thoroughly, and hypotheses about the interplay between various determinants and mechanisms leading to health impact yet to be tested and falsified. We can overcome some of the problems in previous studies with our research approach because we
have access to appropriate data and will use methods for analysing the issue properly. The data are on individual level, and its longitudinal nature makes it possible to follow individuals over time. The longitudinal nature and the hierarchical geographical structure of the data require a multilevel approach in order to be analysed correctly. Furthermore the population studied is large serving as a sufficient base for estimating the models.
To test the research questions, we will utilize the Linnaeus database, an anonymized longitudinal database maintained at the Centre for Population Studies at Umeå University (Malmberg et al. 2010). The Linnaeus database consists of register data from the Statistics Sweden’s LISA database (which is longitudinal integration database for health insurance and labour market studies) for the period 1986-2009, population censuses 1960, 1970 and 1980. It also consist the Cause of Death Registry and Patient Registry from national Board of Health and Welfare for the period 1990-2006. It contains rich information about socio-economic conditions for each individual and also place of residence on one-hundred meter squares.
Using digitalized information of place of residence for all residents in Sweden for the period 1986 – 2006, we can aggregate data to any spatial unit and characterize any such unit by for instance level of income or education. In this study, we include all residents in Sweden aged 55 or older during the period 1986-2009.
Measures of health outcome include information about mortality by death cause and derive from the Cause of Death Registry. It also includes hospitalization data that derive from the Patient Registry and include information about hospitalization by diagnosis. These data are available for the period 1990-2006 and cover all residents in Sweden who have been hospitalized or have died during the period of investigation.

Method
The study will be based on a sample of the total Swedish population. For the selected sample we will produce and analyse life course data based on incomes and migration during the studied period as well as other relevant variables such as marital status. One of the difficult problems with proving the impact of income inequality is to distinguish between the possible ways that can lead to this effect. As mentioned above, absolute income can give the same results of a correspondence between inequality and higher mortality. Wagstaff and van Doorslaer (2000) have identified several different meanings and also suggested methods on how to distinguish the different meanings.
From the individual data we can measure the level of income distribution (using Gini-index) for any geographical area at the different levels with annual data for the period 1986 – 2006
and on parish (and higher) level from 1960 (data for every ten year). In the analyses, we will use both disposable income and income from employment to measure the equality and as well as the individual income, family income and average income in the geographical area. We
will also assess the individual’s relative income. We also plan to use time in education as an alternative measure of inequality. Using digitalized information of place of residence during the study period, we can aggregate data to any spatial unit and characterize any such unit by level of income or education. Taking advantages of the longitudinal database, we will map individuals’ residential mobility and migration, and their exposure to different neighborhoods,
local labor market areas and counties with various income inequality and income levels, and distinguishing between urban and rural regions. In this way, we can evaluate the effect of different time lags on our outcomes. As alternative measures of equality to Gini coefficient, we will also use the proportion of the total earnings in the geographical area that derives from income of the poorest quartile. In this way we examine the impact on health of income among the poor, compared to the general equality (measured by the Gini coefficient). The effect of inequality can be different for different economic subsets, making models that incorporate this interesting to test.
In our analysis, we will investigate the associations between the degree of equality (measured by Gini coefficient) in the residential areas where the person have lived and the health outcome, including all causes mortality, CVD mortality, cancer mortality and hospitalization with CVD or cancer diagnosis. By comparing the different causes of death, we can evaluate possible differences in how they are related to inequality. In the analyses we will control for the effect of various potential confounders at individual level including income, education, sex, marital status, etc. We will also use aggregated contextual data, which allows us to identify and separate contextual effects from those stemming from individual characteristics.
This structure of the nested data, with variables ranging from individual level to regional level, calls for multilevel analysis to understand the interaction of factors operating at different levels. We will apply multilevel survival analysis to model the association between income inequalities in subsequent all-cause, CVD and cancer mortality (Yang et al.). We will try different ways of area categorization and assess the variation in mortality that is
attributable to income inequality at different geographical levels. We will conduct structural equation modeling to assess the interplay between latent (construct) and observable (indicator) variables on income inequality and their impact on mortality. Ecological studies, that dominated much of the early studies, have the problem that the
association between inequality and health can either be causes by income inequality in itself or by the composition of the population. For our analyses we believe a multilevel survival model is the most adequate method (Subramanian and Kawachi 2004). With this approach,
survival is modeled controlling for individual specific variables as well as variables related to the geographical context at different levels. The study variables, different measures of inequality, are included in the models to assess their effect on mortality. This way the effect of variables measured at different geographical levels are estimated and compared while handling the heterogeneity between geographical units. In the literature on the impact of conditions in the residential areas on health, it is usually assumed that the effect comes from a long-term exposure. Many previous studies have,
however, been based on data with only one observation or a short time of exposure, often using cross-sectional design. The fact that composition in the geographical areas may have changed recently and that many residents may previously have lived in very different kinds of regions is previously not taken into consideration, often as a result of in inadequate data.
The associations between health outcome and the degree of equality may be mediated through various mechanisms and could merely be due to selection effects. For instance, people who experience living in an unequal region as a stressful situation, with negative health effects,
may move out from that region and end up in a more equal region. Since we have longitudinal information about previous residence and mobility trajectories we will control for the influence of having lived in regions and residential areas with different socio-economic
characteristics and also take into consideration the effects of self-selection. Due to the richness of our data we have good opportunities to control for various confounders and also to
analyze the risks of death and hospitalization over a long time period.
Since our data include annual information about place of residence on one hundred meters square for all individual we have the opportunity to consider the effects of both long-term and short term exposure to a residential area, changing composition in the area over time and the
time of exposure to different kinds of areas also for movers. And since we have longitudinal information about health outcome, we also have the possibility to analyze the long-term and the short-term effects on health outcome.
For the analyses, we will map individuals’ residential careers and their exposure to local labor market areas, municipalities and neighborhoods by income inequality and income level (inequality by education level will be used as an alternative measure). Since we believe the effect on health from the socio-economic environment may be different by levels of aggregation and by degree of urbanization, we will explore the association between health outcome and socio-economic conditions on different levels of aggregation and in urban and rural regions.

Working plan
The plan is to initiate the project in January 2013, and during the spring of 2013 we will start by organizing the data to conduct our various empirical analyses. Initially we map the socioeconomic differentiation in Swedish local labor market regions, municipalities and in neighborhoods (using different delimitations). We will also do the calculations of income equality. By using descriptive statistics and linear regression, we will identify place-specific characteristics of geographical areas (local labor market regions, municipalities, and neighborhoods) by degree of equality. Since the outcome of this first analysis is in itself an interesting result, we plan to publish the results from this first analysis during the fall 2013.
From the fall 2013 and during 2014 we will conduct analyses of associations between health outcome and socio-economic characteristics on the individual as well as the contextual level
to assess the impact of using different levels of area aggregation, including the impact of income equality. However, the plan for 2014 is also to summarize our first results and to present them in conferences and publications. For 2015 we plan to do additional analyses and to write articles with the aim to publish our results in international peer-reviewed journals.
We have planned for at least three articles, whereof one will focus on the effect of migration for how economic inequality is structured and what implications this has for the relation between economic inequality and health. Another will have as its main topic the effect of different aggregation levels. This second article is based on our assumptions that there are different pathways at different levels. The third discusses the cumulative advantage theory in relation to the health of the elderly and its relation to inequality. We will compare the effects of inequality in different parts of the life span, starting with those in the older part of the working age and continuing with different age groups of the retired population.
The scientific results from our project will be published in international peer-reviewed journals within the fields of population, social science, health and ageing research. We will also participate in international conferences and discuss our results in meeting with
collaborates in our international research network, for instance with our American colleagues at RAND Cooperation, University of Southern California and National Institute of Ageing.
The group has also well established contacts with contact with the county council (landstinget) in Västerbotten county and will present our results also to people working in the health sector as well as to the general public.

Project group
Professor Sören Edvinsson has his background in history specialized in historical demography and is affiliated to the Centre for Population Studies (CPS) and the Demographic Data Base, Umeå University. He has a long experience of working with demographic register data in relation to health and mortality research. He contributes with a demographic competence and knowledge of social stratification and health from a long-term perspective. He will also be strongly involved in the organization of the data set. Professor Gunnar Malmberg works at the Department of Social and Economic Geography, Umeå University, and is specialized in the field of population geography and has a long experience of analyzing spatial register data. His main responsibility will thus be onissues related to spatial analysis. Dr Erling Lundevaller is statistician with his expertise in spatial statistics and has a long experience of working with register data. He is lecturer at the Department of Statistics, Umeå University. His main responsibility will be to organize the data, plan the statistical analysis and perform this analysis. Associate Professor Nawi Ng is a physician who works as epidemiologist at Umeå Centre for Global Health Research, Umeå University. His research focuses on health inequality in older adult population in Africa and Asia, and chronic disease risk factor
surveillance in Asia and in Sweden. He contributes with his epidemiological knowledge and experience of methods in epidemiological research. All participants are involved in the
analysis and writing.
All project participants are connected to the research program Ageing and Living Conditions, Umeå University, which is an interdisciplinary research environment with researchers from the social science and medical faculties. The group has an extensive international collaboration with researchers in the field of ageing, for example from the American National Institute of Aging (NIA). Edvinsson and Malmberg are promoted professors without research in their positions.
We apply five months per year during three years for Erling Häggström Lundevaller, four for Sören Edvinsson, three for Nawi Ng and one month for Gunnar Malmberg.

Significance
Our planned project is of special interest in that it investigates the issue in a country that is considered as one of the most equal nations in the world. Even so, there are large differences within the country and preliminary results indicate that the effects of inequality on health. The project focuses on the prevalence and causes of death and disease in society in particular in relation to class. In this approach we consider the psychosocial aspects of health, although not ignoring the material aspects of class and socioeconomic inequality. The project has also strong implications for the understanding determinants of health and wellbeing among the
elderly.
Despite a rapidly increasing life expectancy, there are still large health differences in Sweden between regions and social classes. The future ageing population will put heavier demand on health care and the costs will increase. Research has shown that health needs to be understood and analyzed in its social context. We thus need to study how social differentiation in society influences our contemporary health situation and to analyze the possible implications if social
and economic inequality will increase in society. We can disentangle the connections between migration, residence, social inequality and health. The project will also increase our knowledge about the spatial aspects of health, for example differences between urban and
rural settings and regional patterns of health and social inequality.

Preliminary results
In our first preliminary analyses, we explore the effect on all cause mortality for the age group 65-74 years from economic inequality on municipal level in Sweden 2006. We find a large distribution of levels of economic inequality on household level between Swedish
municipalities. Our results indicate the existence of a substantial, yet not huge, association between inequality and mortality in Sweden in a multilevel analysis. For example, an increase of 0.1 unit in Gini coefficient corresponds to about a fourth of the increased death risk for unmarried compared to married persons (Edvinsson et al. 2012).

International collaboration
Included in the research environment at Centre for Population Studies are scholars from various disciplines focusing on the interrelations between health and socio-economic conditions in the ageing population. Apart from the researchers in this project the research
group also included other experts on Demography (e.g. professor Anders Brändström) Epidemiology (e.g. professor Lars Weinehall), Statistics (e.g. professors Göran Broström and Xavier de Luna). The group also works together with international researchers engaged in
similar research, for instance researchers from National Institute of Ageing (Washington D.C.) the University of Michigan, University of Southern California and RAND cooperation.

Ethical considerations
The project is planned within the Ageing and Living Conditions programme at CPS, Umeå university. This programme has access to the so called Linnaeus database containing all Swedish citizens 1986-2009. The data on individuals are fully anonymized and the work is performed within the strict regulations set up by the CPS in order to guarantee secure handling of individual data. The research is approved by the Regional Ethical Committee at Umeå University (DNR 07-142Ö).
The regional ethical committee at Umeå University has approved research on the relations between socio-economic conditions on health outcome by use of the data presented above (DNR 07-142Ö).
Latest update: 2018-08-20