Design

This is an ecological study of multiple groups of countries examining associations between economic variables and health outcomes. We will explore these associations between countries at the general level and also among countries with similar income levels. For this purpose, we will group countries according to their wealth using the World Bank (WB) income groups.

Selection of cases

Countries considered for the study were taken from the UN country members list.

The selection criteria for countries to be included was that information about economic and health variables was available. Health data from 2016 were chosen after determining that it was the most recent year with the largest number of countries with complete health data. For economic variables, GNI (gross national income) per capita corresponds to data from 2016. We accepted to include the most recent Gini coefficients between 2011 and 2016 due to the significant variability of the information available.

Variables

Independent variables

GNI Per Capita: The gross national income represents a country’s income and is used by the WB to estimate the size of the economies using the Atlas method. For this study, we obtained the GNI per capita from the World Development Indicators (WDI) by the WB¹⁷.

Gini Index: The Gini coefficient or index is a measure of statistical dispersion to describe the inequality of income of countries or determined demographic or economic groups¹⁸. We obtained Gini coefficients for each country from the Standardized World Income Inequality (SWIID) database in its 9th version¹⁹ and used income inequality data after taxes and payments. We used the SWIID data since it increases international and temporal comparability due to its inclusion of samples from different sources and years.

Dependent variables

Health variables: The prevalence of high BMI (overweight and obesity) is defined as the percentage of the population presenting a BMI ≥25 kg/m². The NCDs mortality rate represents the proportion of the population – in a given place and period – that dies due to NCDs. We used age-adjusted data from 2016 from the Global Health Observatory^20,21.

Statistical analysis

Descriptive statistics were used to describe the economic and health variables. Variables were analyzed for normality using descriptive statistics, P-P and Q-Q plots, and the Shapiro-Wilk test. The income variable GNI per capita was log-transformed.

Due to the non-normality of the prevalence of high BMI and the income variable – which persisted after the logarithmic transformation – and the relatively small sample, associations between economic and health indicators were carried out using a non-parametric test. Kendall’s tau rank correlation test was selected over Spearman’s coefficient due to a small sample.

We grouped countries by wealth, following the WB’s income groups. The groups are based on the GNI per capita. We selected this classification as it is one of the most commonly used in public policy and health. Data were analyzed using IBM SPSS version 28.

Descriptives

Of the 193 eligible countries, 153 (79%) met the inclusion criteria for the study.

Grouping of countries by income level resulted in four groups divided as follows: Low, 20 (13%); Lower Middle, 48 (31%); Upper Middle, 38 (25%); and High, 47 (31%). Ranges of income are shown in Table 1.

Preliminary analyses showed non-normality of the income variable (Skewness=1.92; Kurtosis=3.09; W=0.71, p<0.001). Log transformation was carried out. However, the variable did not reach the desired normality (Skewness=0.07; Kurtosis = -0.90; W=0.06, p=0.007). We kept the variable log-transformed but used non-parametric testing.

The Gini coefficient considers a normal distribution. Our analysis of normality showed that considering our sample of countries, normality persisted (Skewness=0.32, Kurtosis=0.49, W=0.98, p=0.05). Table 1 shows descriptive Log GNI per capita and Gini index results for each income level.

Regarding the health variables, the NCDs mortality rate showed a normal distribution (Skewness=0.14, Kurtosis= -0.64, W=0.98, p=0.055.), while the prevalence of high BMI did not (Skewness= -0.35, Kurtosis= -1.35, W=0.89, p<0.011).

Associations

General associations

The correlation analyses revealed associations between the mortality rate of NCDs and prevalence of high BMI with economic variables. Specifically, the NCDs mortality rate showed a moderate negative correlation with Log GNI per capita, (r[151]= -0.54, p<0.05). Conversely, a weak correlation was observed between the NCDs mortality rate and the Gini coefficient (r[151]=0.20, p<0.05). In terms of prevalence of high BMI, a moderate positive correlation was found with Log GNI per capita (r[151]=0.46, p<0.05). In contrast, a small negative correlation was observed with the Gini coefficient (r[151]= -0.17, p<0.05). Results are shown in Table 2.

Associations by income groups

The significance of the association between Log GNI per capita and the Gini coefficient was evident solely within the highest income group. Within the same group, Log GNI per capita exhibited a weak negative association with the Gini coefficient (r[45]= -0.20, p<0.05) and a moderate and negative correlation with the NCDs mortality rate (r[45]= -0.44, p<0.01). However, Log GNI per capita did not show a significant association with the prevalence of high BMI (r[45]= -0.03, p>0.05). Additionally, whithin the highest income group, the Gini coefficient showed a weak negative association with the prevalence of high BMI (r[45]= -0.27, p<0.05).

In the Upper Middle group, a single significant association was observed. The NCDs mortality rate showed a weak negative correlation with the Gini coefficient (r[38]= -0.23, p<0.05). Similarly, in the Lower Middle group, only one significant association was identified: a weak positive correlation between the prevalence of high BMI and Log GNI per capita (r[48]= -0.26, p<0.01).

The correlation analyses revealed no significant associations between the health outcomes and the economic variables in the Low income group, as shown in Table 3.

Methodological considerations

Several studies^{1,3,11,12,22,23} have discussed the relationship of income and inequality with health variables in determined groups accounting for wealth or development, namely ‘high income’, ‘developing’, or ‘the 30 wealthier’, among other classifications. However, much of the results will be determined by the grouping criteria and the variables considered to describe ‘wealth’ and ‘development’.

To determine how rich or developed a country is, researchers and economists often rely on national indicators such as the GNI in its per capita variation to get an idea of the income of its citizens. The WB income classification used in this study is the most frequent tool to group countries across research and public policy²⁴. Nevertheless, some authors have criticized such classifications. Nielsen²⁵ has argued that the WB, International Monetary Fund, and United Nations Development Program (UNDP) categorizations lack clarity in their ‘underlying rationale’ and that there is a need for a new classification system based on transparent methodology instead of ad hoc rules and criteria. Furthermore, Kaplan²⁶ argues that the WB classification provides a distorted picture of reality, ignoring key aspects of development such as inequality, human development, social exclusion, and governance, which undoubtedly affect health indicators.

Perhaps one of the most relevant issues regarding the use of GNI per capita alone is the lack of a component that accounts for the dispersion of wealth within a country – especially relevant for Latin America and Asia. As we did in this study, several authors have proposed to include the Gini Index when exploring relationships between development and health^12,27. However, even when inequality is considered, there are concerns that income is not a good indicator enough, and some have proposed that a multidimensional approach should be proffered, such as the Human Development Index by the UNDP²². Lynch et al.²⁸ indicate that some authors have also used arbitrary cut-off points to suggest associations between income or development and health.

Literature indicates that even well-established and widely used instruments, such as the division proposed by the WB, present essential imperfections. We believe that a new and different method to describe the development of nations is required for its use in health research.

In addition to the variables used to construct economic indicators, which influence when and how wealth explains diverse health outcomes among countries, we identify methodological issues commonly found in ecological studies. These issues can also impact the statistical associations between economic indicators and health outcomes, contributing in part to inconclusive results. Studies on disparities and population health often look for associations with countries stratified by wealth. By having a small number of cases (countries), this method leads to small groups of countries with similar income indicators, reducing the data’s variability and diminishing the power of associations observable at a general level.

Characteristics of the health variables

As Linden and Ray¹⁶ stated, our results also show that the wealthiest countries are less inequitable and have better health outcomes. The prevalence of high BMI was the exception, being higher in the wealthiest nations. As reported by the Global Burden of Disease Study 2016⁴, this NCDs risk factor increases with socioeconomic development. Despite this fact, as stated by Bennett et al.³, those with the highest risks of death from NCDs – based on the NCDs mortality rate – were the low- and middle-income countries (LMICs).

Prevalence of high BMI

High BMI is a remarkably complex variable to analyze. It presents a distribution pattern marked by socioeconomic and cultural covariates, and may even be the opposite between countries of different levels of development^8,22.

In developing countries, inhabitants face food shortages and a predominance of manual labor. Low caloric intake and high energy expenditure would explain the low prevalence of being overweight or obese in these nations. The absence of significant associations with the economic variables, at least by the WB grouping, may be due to the division according to income (developed in the previous section) in which the narrow range in the variable of GNI per capita (<US$1005) is the lowest within the four income levels. The middle-income countries show a relation between high BMI and income much less clearly. Industrialization, urbanization, and nutritional transition are commonly postulated as drivers of being overweight or obese with increasing wealth in the country. This was also observed at the grouped level, where prevalence of high BMI increases exponentially along with economic development.

Among the groups divided by income according to the WB, the range of the GNI per capita variable is increasing, allowing for more significant income differences between countries; and, presumably, greater class differences also within each. Despite the poorest countries being the most inequitable, the highest income group showed the most significant income differences. Concordantly, it was the only income group where the Gini index and GNI per capita showed a significant association. As the level of development increased, the demarcation of social classes is accompanied by signs of distinction²⁹, in which obesity is increasingly associated with poverty and intertwines with other sociocultural variables such as gender and education level. Although the measurement of the latter variables is beyond the scope of this study, it is essential not to forget the aforementioned high BMI paradoxical distribution pattern in order to avoid an ecological fallacy. Concordantly, our results showed that high BMI was significantly associated with the Gini index only in the high-income group, where the income range is far broader.

Social, economic, and cultural variables can act as protective or risk factors for a prevalence of high BMI, depending on the level of socio-economic development of the country and the subgroup to which people belong. This could explain that, at the grouped level, no significant association was observed between high BMI and economic variables. It demonstrates the importance of analyzing the prevalence of high BMI within specific population subgroups, defined by the social, economic, and cultural variables above. Our study coincides with the review by McLaren²², where no significant associations between income and high BMI in the wealthiest countries were found. Other economic indicators should be considered in the study of the association between socioeconomic status (SES) and prevalence of high BMI in wealthier nations. The author identifies other SES indicators (education level, living area, or occupation) associated with a high BMI. This could be due to the ability of these indicators to translate distal effects of inequity in nations where income inequality is lower.

NCDs mortality rate

The WHO highlights four groups of diseases within the NCDs that account for over 80% of deaths: cardiovascular diseases, cancer, chronic respiratory diseases, and diabetes³⁰. The probability of dying from any of these diseases shows significant heterogeneity within each income group; for example, in the lowest income group, this probability is 15% in Ethiopia, while in Afghanistan it is over 30%. Similarly, in the highest income group, the likelihood is under 10% in countries such as Israel, Japan, or Italy, and over 20% in Latvia or Lithuania³⁰. The search for significant relationships between economic indicators and NCDs mortality is particularly complex when carried out on many diseases that only share non-transmissibility but differ widely in their etiology, development, and consequences.

Although our results show a significant association between economic indicators and the mortality rate from NCDs at a general level, we believe this is due to the general relationship between health and economic development, and not to particular conditions of these diseases. A good approximation could be the four main groups proposed by the WHO, which would show an association not only with economic indicators but also with geographical ones.

Income versus inequality

We agree with Pickett and Wilkinson²⁷ that larger income differences increase social inequality and, therefore, differences between social classes. The neo-material theory would not be sufficient to explain the differences in health among rich countries, whereas the social integration theory could become relevant. Nevertheless, there are few rich countries. Wilkinson and Pickett¹¹ propose that while nations join the group of developed economies – GNI per capita equal to or US$ >25000 – additional income increases become less relevant, and only through more equality health outcomes would improve.

We believe it is essential to reflect on this point since only 26 countries (17%) have reached this threshold. Many studies that support the social integration theory – and that show a positive correlation between health results and lower inequality – have been done precisely on high-income or European countries^8,31,32.

Similar to Lynch et al.²⁸ who also recognize the negative psychosocial consequences of income inequality, we argue that the interpretation of links between income inequality and health must begin from the structural causes of disparities and not only focus on the perceptions of those disparities. In these wealthy countries, perceptions of inequality could also become more relevant precisely because the structural causes of inequalities are not significant.

Inequality necessarily implies an essential share of poverty. We believe that inequality matters because a lower economic capacity involves low investment in human capital, education, healthcare services, and other social infrastructures. It is not due to inequality itself but to the share of the population that does not have the necessary economic resources and is systematically excluded from essential social services.

Income inequality is a manifestation of neo-material conditions that affect health. Associations would depend on the level and distribution of other social resources. Low-income populations living in high-income countries with high levels of inequality enjoy healthcare services, education, and nutrition above their counterparts in the low- and middle-income countries¹⁰. Furthermore, although gradients in health exist even in high-income countries with high levels of inequality, their public services are less concentrated in the wealthier population than in the lower income countries with high inequality.

It may seem paradoxical that it is in the countries with the highest income where the relationship between inequality and health results is more significant than in the lowest income countries. As we stated above, this might be due to the limitations of the grouped analysis present in ecological studies, particularly associated with the broader income range present in the high-income category, allowing a more significant association between inequality and income. Additionally, the literature shows that high-income countries do not necessarily share similar welfare states, which would also impact the effect of inequity within these countries. Chung and Muntaner³³ identified that the difference between high-income countries and health indicators such as infant mortality and low birth weight were associated with a comprehensive welfare state. This has also been described by Navarro et al.³⁴, who examined health indicators from OECD countries for 50 years and their relationship with political, economic, and social variables.

The considerable amplitude of the GNI per capita among high-income countries (US$ 20000–82000) means significant differences in development exist among them. This could produce a gradient, not only social but also a health one. Therefore, the intervention of the welfare state to neutralize the health gradient among high-income countries could have an exceptionally high impact.

Precisely, rich countries with the worst health results are those whose welfare state has significant access barriers to health based on ethnicity, race, and nationality. Among Gulf countries – Saudi Arabia, Qatar, and Bahrein, among others – some healthcare services are restricted to nationals. In contrast, others are designated for foreigners, in the circumstances where labor conditions for immigrants have been described as ‘modern slavey’³⁵. Saudi Arabia presents significant differences in healthcare access due to the concentration and development of high complexity services in rich and urban zones, segregating the rural and disadvantaged populations.

Something similar happens in the United States, where a healthcare system increases social inequalities, including health results. While income inequalities have increased, chronic health conditions follow a predictable increase in prevalence as income decreases³⁶.

We believe that the difference in health results – both for rich and for lower income countries – is due to the fact that a generous welfare state allows for better healthcare accessible to all, regardless of income or location. The social integration and neo-material theories can coexist, but we must remember that most countries are not rich. Inequality is always crucial – for rich and poor – but a generous welfare state requires, first and foremost, material resources before these can be distributed.

Limitations

An inherent limitation of ecological studies is that the results obtained at the aggregated level may not necessarily represent what occurs at the individual level. However, although it has been suggested that individual studies should be prioritized when assessing the effect of income distribution on health indicators, we consider that ecological studies can also be valuable as a first step in providing guidance for global policymaking.