Socioeconomic status, health inequalities and non-communicable diseases: a systematic review

Aim A comprehensive approach to health highlights its close relationship with the social and economic conditions, physical environment and individual lifestyles. However, this relationship is not exempt from methodological problems that may bias the establishment of direct effects between the variables studied. Thus, further research is necessary to investigate the role of socioeconomic variables, their composition and distribution according to health status, particularly on non-communicable diseases. Subjects and methods To shed light on this field, here a systematic review is performed using PubMed, the Cochrane Library and Web of Science. A 7-year retrospective horizon was considered until 21 July 2017. Results Twenty-six papers were obtained from the database search. Additionally, results from “hand searching” were also included, where a wider horizon was considered. Five of the 26 studies analyzed used aggregated data compared to 21 using individual data. Eleven considered income as a study variable, while 17 analyzed the effect of income inequality on health status (2 of the studies considered both the absolute level and distribution of income). The most used indicator of inequality in the literature was the Gini index. Conclusion Although different types of analysis produce very different results concerning the role of health determinants, the general conclusion is that income distribution is related to health where it represents a measure of the differences in social class in the society. The effect of income inequality is to increase the gap between social classes or to widen differences in status.


Introduction
From a broad point of view, an individual's health is considered not only an absence of disease, but also a fundamental human right (WHO 1986). A comprehensive approach to health highlights its close relationship with social and economic conditions, the physical environment and individual lifestyles. According to the Commission on Social Determinants of Health, we can consider health inequalities to be the result of the cumulative impact of decades of exposure to health risks of those who live in socioeconomically less advantaged circumstances (WHO 2008).
If we focus on all the socioeconomic variables, the relationship between income (understood as a measure of socioeconomic status) and health is probably the most complicated (Fuchs 2004). The correlation coefficient, obtained from the crudest associations, can range from highly positive to slightly negative, depending on the context and the aggregation level. Even when the positive correlation is strong and stable, causal interpretations may include income influencing health, health influencing income and/or Bthird variables^affecting both indicators in the same direction and at the same time. For this reason, the gross domestic product (GDP) is related to some health outcomes indicators (Kanavos and Mossialos 1996). However, there are exceptions. For example, some Southern countries in the European Union that are relatively poor have a life expectancy indicator greater than that of the rich countries of Northern Europe. Also, we can observe that the USA, one of the world's richest countries in terms of GDP per capita, has infant mortality rates similar to those of poorer countries (Starfield 2000).
In addition, there is a large and growing body of literature in which the effects of income on health are examined because of the importance of these effects in the development of appropriate economic policies (Gravelle et al. 2002). Many studies have shown a negative association between income and mortality (Lutter and Morrall 1994;McCarron et al. 1994;Viscusi 1994;Singh and Siahpush 2002;Shaw et al. 2005;Pearce and Dorling 2006;Leyland et al. 2007;Ezzati et al. 2008; Thomas et al. 2010). These empirical findings suggest that individual health is a function of individual income-the absolute income hypothesis. In relation to income inequality, the relative income-health hypothesis suggests that income inequality has a detrimental effect on population health because it is an individual's relative rather than absolute income that is important for health (Marmot et al. 1991;Wilkinson 1997Wilkinson , 1998Wildman 2001Wildman , 2003López I Casasnovas and Rivera 2002;Gravelle et al. 2002;Eberstadt and Satel 2004). Income inequality may therefore be a health risk (Le Grand 1987;Wilkinson 1992Wilkinson , 1996. Thus, life expectancy and population mortality have been used as key indicators of economic and social development (Van Doorslaer and Koolman 2004;Cantarero et al. 2005).
Although previous empirical literature presents different interpretations of the evidence, most analyses report that the average health is worse in more unequal societies. However, this relationship is not perfect, since there are several determinants that can affect it. There is clear evidence indicating that a nonlinear, typically concave relationship between health and income at an individual level will generate an aggregate relationship in which average health will depend negatively on the degree of inequality in the income distribution (Duleep 1995;Wilkinson 1996;Mackenbach et al. 2005;Mackenbach 2012). Hence, income redistribution from the rich to disadvantaged groups may improve some health indicators (Kawachi and Kennedy 1999). Also, some authors have suggested the existence of conceptual difficulties in studying the relationship between income and individual health when aggregated data are used, because revenues have a diminishing marginal effect on health (Deaton and Muellbauer 1980). This is because if income inequality increases, it tends to reduce average health but improve the health of Bthe rich,^although this latter effect is less significant than the reduction in overall health.

Methods
A systematic literature search was performed in PubMed, the Cochrane Library and Web of Science (until 21 July 2017) to identify the most relevant published evidence regarding the relationship between income and health. In all databases, terms related to Bhealth,^Bincome^and Binequalities^were combined. See Appendix Table 4 for the search strategy/search terms used. The searches were confined to papers published in the English language since 2010 to limit the scope of this review to the most recent data and the state of the art. In other words, we considered this retrospective horizon, from the beginning of this century up to date, to be enough.

Results
After finding publications in the electronic searches, duplicate records were removed. The selection of papers was ultimately based on the following eligibility criterion: an applied study with a focus on one or more OECD countries (including the European Union and other developed countries). Additionally, the results of Bhand searching^were also included in the following pages, where a wider horizon was considered. Figure 1 is a diagram of the paper selection process.
The literature search located 381 publications in the databases under consideration, and 17 papers published between 2010 and 2017 were identified through Bhand searching.^A total of 11 duplicates were removed, resulting in 387 Bunique papers.^After screening the titles against the eligibility criteria, 105 papers were selected. Of these, 68 articles were excluded as they did not fit the previous criteria. So, a final set of 37 selected studies were taken into account in this review, and further papers were finally considered to have a robust overview. Table 1 focuses on the 26 papers found in the database search.
We also reviewed the works obtained from Bhand searching,^where the results are almost all based on economic criteria. Specifically, we can highlight that there is also much evidence for the effect of income on health status for different socioeconomic groups. References are listed at the end of this article, and we included references in journals and cited books. Table 1 focuses on the 26 papers obtained from the database search. The first group of studies explores the fact that people who live in areas of high inequalities tend to report themselves as having both an objective (having a shorter life expectancy and high adult mortality) and subjective health status and that this tendency increases over time (Allanson et al. 2010;Elgar 2010;Huijts et al. 2010;Idrovo et al. 2010;Islam et al. 2010;Karlsson et al. 2010;Oshio and Kobayashi 2010;Petrie et al. 2011). Moreover, the following group of studies developed various econometric approaches (multilevel regression, bivariate and count data or logit models) to consider geographic, socioeconomic and poverty-related issues (Chen and Crawford 2012;Hosseinpoor et al. 2012;Karlsdotter et al. 2012;Martinson 2012;Allanson and Petrie 2013;Ásgeirsdóttir and Ragnarsdóttir 2014;Wilson et al. 2017). The health concentration index and its corrections have been employed  [The correlations between the variables studied were statistically significant (p < 0.05) and displayed the desired tendency. Where LEB was lower, income inequality, political rights and civil liberties or ethnic fractionalization were greater.
In addition, LEB was higher when less corruption was perceived and where there is more generalized trust. The Gini coeffi cient was greater where the ethnic fractionalization was greater and lower where there was less perceived corruption. The Gini coeffi cient was negatively correlated with generalized trust and the CPI 2004, and it was positively correlated with ethnic fractionalization and the index of freedom. The correlations between the CPI and the index of freedom or ethnic fractionalization were negative, indicating that fractionalization and political rights and civil liberties were greater where less corruption was perceived. Ethnic fractionalization was positively correlated with the index of freedom and negatively correlated with generalized confidence. Finally, the index of freedom was negatively correlated with generalized trust] Islam et al. (2010Islam et al. ( ) 1980Islam et al. ( -1981Islam et al. ( , 1988Islam et al. ( -1989Islam et al. ( , 1996Islam et al. ( -1997  Ordered probit model There is evidence of a negative relationship between income inequality and individual health in high-income countries Oshio and Kobayashi (2010) 2001, 2004, 20092000, 2003 Japan ANOVA and ordered bivariate probit models Individuals who live in areas of high inequality tend to report themselves as both unhappy and unhealthy Petrie et al. (2011Petrie et al. ( ) 1999Petrie et al. ( -2004 Scotland, England and Wales Decomposition method to account explicitly for mortality in the longitudinal analysis of income-related health inequalities Accounting for deaths in the decomposition analysis shows that the relative health changes for both regions and genders between 1999 and 2004 were significantly regressive, such that initially poor people experienced a greater share of health losses compared to their initial state of health. [Income-related health mobility, expressed as a proportion of final health inequality, was 11.5% for CI U and 15.9% for CI, with the latter value being 35.3% larger than the former where this ratio is also given by the ratio of the corresponding P values] suggests that a 1% increase in income inequality would increase infant mortality by 0.47%. For an increase in the Gini index, a 43% increase in income inequality would correspond to 14% (1.43^0.373) increase in child mortality for men and 16% (1.43^0.424) increase for women. However, since the mortality rate at ages 1-14 is lower for women than for men (in our data, on average 25% lower), the absolute effect is greater for men. At ages 15-49, the male mortality coefficient for the Gini index is still positive (0.285) and highly significant (p < 0.10) However, for women aged 15-49 the coefficient size is 40% smaller than for men (0.171) and significant only at p < 0.10. For the GDP per head predictor, for both men and women, the coefficient -0.30 at ages 1-14 suggests that a doubling of GDP per head would decrease child mortality by 30%] Chauvel and Leist (2015)  Jutz (2015) 2008-2009 42 European countries Two-step hierarchical estimation approach. Income inequality has more impact on health inequalities than do social policies.
[The relation between social protection expenditures and health inequalities could not be confirmed. As expected, SPE and health inequalities are negatively related (−0.25), but the relation does not reach significance. The Gini index was positively related to health inequalities (0.39, p < 0.05), i.e., higher income inequality was linked to higher health inequalities] in recent studies (Siegel et al. 2014;Vallejo-Torres et al. 2014;Siegel et al. 2015). As described in Table 1 and Fig.  2, the relative income-health hypothesis was also analyzed by Hu et al. (2015), who conclude that in European countries income inequality does not have an independent effect on mortality. Moreover, it seems logical that there would be a difference between rich and poor countries in how income distribution would affect health status (Waldmann 1992;Rodgers 1979;Deaton 2001b). Wilkinson and Pickett (2006) performed a review of the literature analyzing the association between the income distribution and health of the population. The general conclusion is that income distribution is related to health where it represents a measure of the differences in social class in the society.
Among the studies conducting their analysis on individual data, Ettner (1996) estimates the effects of income on a set of individual health proxies. The results show a strong positive effect of income on health. Dahl et al. (2006) analyzed the degree to which contextual income inequality affects the health in Norwegian regions; the results differ from previous studies in suggesting that in Norway a comparatively egalitarian income distribution interferes with the emergence of regional-level income inequality effects on mortality. From another point of view, a study developed in Spain (Regidor et al. 2014) showed that inequality in the distribution of provincial income declined during the 4 decades covered by the study. More recently, Pickett and Wilkinson (2015) conducted a new review of the literature on the subject, exploring the causal role of wider income differences in health. The authors highlight that the effect of income inequality is to increase the gap between social classes or to widen differences in socioeconomic status.
As Table 2 shows in relation to health indicators, measuring the health status of a population is problematic because there is no complete and comparable health index among countries or regions.
Data aggregation, used in numerous studies examining the health status of the population in different countries and its relationship to the level of income, can also present problems from a methodological point of view. The first problem is the availability of comparable data for long periods of time. The observations are often measures at the national or regional level, in contrast to individual panel data for which there are many observations of cross-sectional measurements at very few points in time. Therefore, the problems differ depending on the observation unit adopted: the individual or an aggregated geographical area.
Causality of the variables that are considered in the analysis of the relationship between income and health is another methodological aspect that is particularly relevant (Fuchs 2004). Although numerous studies indicate a positive relationship  Lillard et al. (2015Lillard et al. ( ) 1913Lillard et al. ( -2009Lillard et al. ( , 1984Lillard et al. ( -2009 US Ordered probit models Exposure to income inequality in early life is related to worse health in later life Quon and McGrath (2015) 1994-1995

Canada
Multilevel modeling Income inequality is associated with injuries, general physical symptoms, and limiting conditions, but not associated with most adolescent health outcomes/behaviors. Income inequality has a moderating effect on family socioeconomic status for limiting conditions Rambotti (2015) 1999 US (plus international comparisons) Bivariate and cross-sectional associations. Poverty has a significant and adverse effect Kim (2016) 1980s and 1990s US Linear probability models Linkages were identified between state-level spending on welfare and education and lower individual risks of dying, particularly from coronary heart diseases. López et al. (2016) 1998-2011 US Multivariable linear and Poisson regressions Income inequality is independently associated with higher health care expenditures and more health care use Wilson et al. (2017) 2000-2011

US
Logistic regressions. Racial disparities in health outcomes exist. Race/ethnic disparities are not merely a result of income between health and income, few of them analyze the causality of this association. The stability of income inequality over time in most countries makes this causality difficult to test (Babones 2008). This author points out that although there exist a Bstrong, consistent and statistically significant correlation between national income inequality and population health,^there is also evidence indicating that this correlation is causal. McKeown (2009) describes changing patterns of population age distributions, mortality, fertility, life expectancy and causes of death. However, within countries, differences in living standards establish a social order in the population. Among the papers that raise the issue of income distribution as a possible reason for inequalities in individuals' health, that published by Deaton (1999) is notable. Furthermore, Deaton and Paxson (2001) develop a similar analysis to examine the relationship between income inequality and mortality. The results show that neither the trends in the level of income nor the inequalities in income explain the adjusted mortality rates by age. Besides, Wagstaff and Van Doorslaer (2000) review a large body of literature on the effects of income inequality on population health. The literature review shows that the individual level studies considered to be relevant provide strong support for the absolute income hypothesis, no support for the relative income hypothesis and little or no support for the income inequality hypothesis. In relation to this, we can think about the countries of Eastern Europe, where, despite their egalitarian distribution of income, there are high mortality rates. Contoyannis and Foster (1999) found that it is absolute income that has a significant effect on health, not relative income. Along the same lines, we can point to the paper of Van Doorslaer et al. (1997), whose results support the idea that health inequalities cannot be definitively attributed to income inequalities.  Table 2 Health indicators, data aggregation and causality of variables (income and health)

Health indicators
There is no complete and comparable health index for all countries. The indicators commonly used are mortality rates (infant and adult) and life expectancy. However, these indicators are not sensitive to improvements in quality of life (Parkin et al. 1987). Data at the individual level are recommended (Wagstaff and Van Doorslaer 2000) Data aggregation This presents problems from a methodological point of view. The availability of comparable data for long periods of time is a problem, and individual conditions of linearity are required, while the evidence suggests that relationships in this regard are configured in a nonlinear way (Preston 1975;Rodgers 1979;Duleep 1995;Ettner 1996;Deaton 2001aDeaton , 2001bGravelle et al. 2002;Mackenbach et al. 2005) Causality of the variables Population health would also help explain differences in income levels among individuals and countries. The effect could bias results and make any inferences about the structural effect of income on health difficult (Fuchs 1974(Fuchs , 2004Ettner 1996) Also, as Table 3 describes, the trajectories of social mobility over the life course (U-shaped) and the variations in patterns of social mobility mean that it is very important to study inequalities in health and socioeconomic status because they are present early in life (Currie and Madrian 1999;Bengtsson and Mineau 2009;Almond and Currie 2011;Currie and Almond 2011;De Ree and Alessie 2011;Lundborg et al. 2014).
Furthermore, it would be interesting to take into account the existing links between parental socioeconomic status (measured by education, income or labor status) and child health and therefore between the health of a child today and his or her health and status in the future (its derived results in education, income and/or adult occupation) (Currie and Madrian 1999;Aizer and Currie 2014;Fletcher 2014;Flores and Kalwij 2014;Flores et al. 2015).

Discussion
In this article we analyze the literature that studies the determinants of health with special attention to the relationship between socioeconomic status and health status. The socioeconomic status will be approached through different indicators, mainly income. To do so, we first discuss relevant articles in this field, which are among the most cited by literature, and then focus on a systematic literature review of recent years. In the revised literature, 5 of the 26 studies analyzed use aggregated data (19%) compared to 21 using individual data (81%). Most of them analyze the effect of income inequality on health status (17) in comparison with the 11 studies that consider income as a main variable of the study (2 of the studies consider both the absolute level of income and the distribution of income). The indicator of inequality most used in the literature is the Gini index.
The revised literature shows that people who live in areas of high inequalities tend to have a shorter life expectancy and high adult mortality and that this tendency increases over time. Among the studies that conduct their analysis on individual data, the results show a strong positive effect of income on health. The effect is particularly relevant in areas of high inequalities, and its influence can be observed from different socioeconomic measures (education, income, labor status).
Findings vary according to the type of study if the individual age is considered. In this sense, we find articles that show that individuals are statistically more likely to report poorer health if they were more unequally distributed during the first years of their lives than at an advanced age. However, other studies find that the magnitude of health inequalities is not consistent across age groups. For the income level, most of the results find that the major driver of the disequalizing effects of mortality is the positive association between old age and poverty.
There is also an interest in solving the apparent paradox that income appears to be related to health within countries but not between them. The explanation relies on the fact that in developed countries, which have already achieved a certain standard of living, increases in per capita GDP have little effect on the levels of health because of the epidemiological transition (understood under the fourth proposition by Omran (1971Omran ( , 1982: BThe shifts in health and disease patterns that characterize the epidemiologic transition are closely associated with the demographic and socioeconomic transition that constitute the modernization complex^), as in addition to epidemiological changes or changes in health conditions, the health transition also incorporates related social changes as a health care transition, as has been shown, for example, by Karlsson et al. (2010), Petrie et al. (2011), Hosseinpoor et al. (2012 and Siegel et al. (2014).
However, population health would also help to explain differences in income levels between individuals and between countries. The importance of investment in health has been re-emphasized by the theories of human capital. Improvements in health diminish productivity losses caused by disease in the workforce, reducing disability, weakness and the number of days off work. Also, they increase assistance to schools and the learning capacity of Alternative specifications should be used for the model, or long panels should be used to follow the same individuals over a period of time, as their age could help to understand the impact of health on socioeconomic status and to predict future health and the expenditure required school children. One could also point to the decline of family disruption and other undesirable social issues as well as the reduction of negative externalities, for example, in the case of caring for the sick. The effects of productivity gains in workers are particularly great for countries with a low level of development. Poor people have a higher risk of illness, and their income depends exclusively on their physical work. Investment in health would therefore be a productive investment, since it would increase income. It would be an important part of development and would help to reduce the income gap between rich and poor countries. Testing this relationship may lead to inconsistencies because of the causality between the two variables. This reverse causality could bias the results and make it difficult to draw inferences about the structural effect of income on health.
Finally, there are some limitations to this review we should consider. First, the literature search was limited to the main (three) databases. Future systematic reviews could also include other relevant sources. Second, the findings were not weighted for sample size.

Conclusion
The published health economics literature on socioeconomic status, health and non-communicable diseases is characterized by many papers showing the complexity of those relationships. Improving this information is crucial if we are to capture the value of socioeconomic measures fully and to discover the most relevant determinants of health and noncommunicable diseases.
From the literature analysis, we can conclude that income inequality was associated with worse average health. These results remain practically coincident regardless of the health indicator considered. The main conclusion of the studies analyzing the temporal evolution of both variables is that income inequalities in health increase over time to the detriment of the economically disadvantaged.
What is true is that different types of analysis produce very different results on the role of health determinants. Thus, the individual conception of health provides a different framework of research from a social analysis. The differences are relevant when the results are presented in terms of effectiveness in health policies and welfare (Wildman 2003). Although the determinants of health identified in individual studies are important variables in an aggregate analysis, there are specific factors that affect social groups. In this sense, for example, a better health status derived from a greater level of education may be the result of an education variable directly influencing the individual's health or may be because of an improvement in social class due to a better education.
Finallly, further research is necessary to investigate the role of income level, its composition and its distribution in health status and the labor market. To help with this, perhaps we can highlight the greater potential of individual studies, with the new databases available, for analyzing hypotheses about a more detailed relationship among socioeconomic status, health and non-communicable diseases.
Authors' contribution All authors developed the idea for the systematic review and contributed to the concept and design. All authors contributed to the writing of the manuscript and read and approved the final manuscript. The materials have not been submitted to other journals.FundingThe paper is part of the research involved in the project BFRESHER-Foresight and Modelling for European Health Policy and Regulation^funded by the European Commission through the Horizon 2020 Framework under grant agreement no. 643,576. The sole responsibility for the content of this project lies with the authors. It does not necessarily reflect the opinion of the European Union. The European Commission is not responsible for any use that may be made of the information contained therein.

Compliance with ethical standards
Ethics approval and consent to participate Not applicable.
Availability of data and material Not applicable.
Competing interest The authors declare that they have no competing interest.
Open Access This article is distributed under the terms of the Creative Comm ons Attribution 4.0 International License (http:// creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.