Patterns of SES Health Disparities Among Older Adults in Three Upper Middle- and Two High-Income Countries

Patterns of SES Health Disparities Among Older Adults in Three Upper Middle- and Two High-Income... Abstract Objectives To examine the socioeconomic status (SES) health gradient for obesity, diabetes, and hypertension within a diverse group of health outcomes and behaviors among older adults (60+) in upper middle-income countries benchmarked with high-income countries. Method We used data from three upper middle-income settings (Colombia-SABE-Bogotá, Mexico-SAGE, and South Africa-SAGE) and two high-income countries (England-ELSA and US-HRS) to estimate logistic regression models using age, gender, and education to predict health and health behaviors. Results The sharpest gradients appear in middle-income settings but follow expected patterns found in high-income countries for poor self-reported health, functionality, cognitive impairment, and depression. However, weaker gradients appear for obesity, hypertension, diabetes, and other chronic conditions in Colombia and Mexico and the gradient reverses in South Africa. Strong disparities exist in risky health behaviors and in early nutritional status in the middle-income settings. Discussion Rapid demographic and nutritional transitions, urbanization, poor early life conditions, social mobility, negative health behavior, and unique country circumstances provide a useful framework for understanding the SES health gradient in middle-income settings. In contrast with high-income countries, the increasing prevalence of obesity, an important risk factor for chronic conditions and other aspects of health, may ultimately change the SES gradient for diseases in the future. Diabetes, Health disparities, Hypertension, Middle-income countries, Obesity, Socioeconomic status Cross-national studies provide additional insight into the understanding of the socioeconomic status (SES) health gradient but there have been only a few cross-national studies of middle- and high-income countries that focus on the SES health gradient for older adults (Beltrán-Sánchez, Palloni, Riosmena, & Wong, 2016; Goldman, Turra, Rosero-Bixby, Weird, & Crimmins, 2011; Payne, Gómez-Olivé, Kahn, & Berkman, 2017). Although health is comprised of physical, mental, and social dimensions, few cross-national studies of middle- and high-income countries examine the SES gradient for older adults across a broad group of dimensions. The relationship between SES and health is complex (Davey Smith & Lynch, 2004), making international comparisons challenging. Most studies in the developed world show a strong positive relationship between poor health and lower SES status. However, one of the more intriguing questions facing social science researchers is the degree to which this relationship holds true in developing economies. Middle-income countries are particularly of interest given that they are experiencing large increases in diet-related conditions—obesity, diabetes, and heart disease—among a growing population of adults over the age of 60 (He, Goodkind, & Kowal, 2016; World Health Organization, 2000). Existing evidence from middle-income settings shows mixed results in regards to the SES health gradient, in particular for obesity (Dinsa, Goryakin, Fumagalli, & Suhrcke, 2012). One of the most comprehensive studies of a middle-income country to date showed reversals of the gradient for mortality, obesity, metabolic syndrome, and other diet-related health conditions in older adults; in contrast, other aspects of health followed an expected positive pattern (Rosero-Bixby & Dow, 2009). In other studies of older adults in middle-income countries, education, body mass index, and waist circumference were either not related (Goldman et al., 2011) or contrasting patterns appeared depending on rural/urban residence (Smith & Goldman, 2007). Demographic Regimes of the 20th Century and Nutritional Transitions in Middle-income Countries There are several reasons for cross-national differences observed in the SES gradient for health. Unequal societies tend to have a more negative impact on health (Wilkinson, 1996) and high levels of inequality mean that certain subgroups within a population remain at lower levels of development even though the country as a whole may prosper (López-Alonso, 2007). Limited social mobility and limited access to quality health care affect the magnitude of differences between SES groups (CHDS, 2008). Levels of country development may also affect health. Reversals in the gradient appear for obesity according to levels of development; at higher levels of development, there is a shift toward an expected positive gradient of low SES and obesity (Dinsa et al., 2012; Monteiro, Conde, Lu, & Popkin, 2004; Pampel, Denney, & Krueger, 2012). However, level of development alone does not completely explain the SES gradient given that, among middle-income countries, positive, negative and no associations appear between SES and health (Dinsa et al., 2012). The rapid demographic, epidemiological, and nutritional transitions from the 20th century to the present are relevant explanations for the mixed patterns found in the SES gradient in middle-income countries. Three aspects of these rapid transitions are notable. First, the demographic transitions of the 20th century characterized by mortality decline did not affect all countries equally in terms of the timing, pace, and reasons for the transitions. Even within countries, transitions have been uneven, particularly in rural areas where existing disparities between rural and urban are large and transitions have lagged behind urban areas. Rapid urbanization produced even greater differences between rural and urban areas (Flórez, Guataquí, Mendez, & Cote, 2016). The increase in obesity and chronic conditions among older adults is predominantly concentrated in urban areas (Carrillo et al., 2016; Ford, Patel, & Narayan, 2017; Unwin & Alberti, 2006) and suggests that the impact of rapid demographic and nutritional transitions in terms of health is most evident in urban areas. Weak associations between SES and health could thus reflect the stark contrast between rural and urban areas in their demographic and nutritional transitions. Urban areas (particularly large metropolitan areas) may be more similar to developed economies in SES health patterns whereas rural areas may produce a reversal of the gradient for obesity because lower SES groups still maintain a more traditional diet while higher SES do not (Smith & Goldman, 2007). On the other hand, higher inequality in urban areas in some developing economies (World Bank Group, 2017c) could produce sharp SES gradients when examining urban areas only. Second, older adults born in the mid-20th century in middle-income countries experienced rapid changes in nutrition later in life (FAO, 1946, 2010) with increased exposure to foods high in saturated fats amidst increased urbanization (Lloyd-Sherlock, Beard, Minicuci, Ebrahim, & Chatterji, 2014; Popkin, 2006; Schmidhuber & Shetty, 2005). The transition away from more traditional food to a diet with higher saturated fat content in developing economies increases the risk of poor health (Popkin, 2006; Popkin, Horton, & Kim, 2001) and could account for differences in the SES health gradient. Reversals of the gradient in middle-income countries could be a result of worse health behavior among higher SES groups. In more developed economies, higher SES groups tend to have access to healthier foods and obesity is more prevalent in lower SES groups (Dinsa et al., 2012). However, higher SES groups in some middle-income countries may be engaging in risky behavior by consuming foods higher in saturated fat content, suggesting a deteriorating health profile for those with higher SES who engage in risky health behaviors (Rosero-Bixby & Dow, 2009), even in settings with good health care coverage and quality health care. A third aspect of the transitions has yet to be thoroughly examined. That is, we know very little regarding the long-term consequences of the demographic regimes of the 20th century on adult health in developing economies. Mortality declined in the 20th century in many countries but the type of demographic regime most of interest is that characterized by rapid mortality decline without parallel improvement in standard of living. The dramatic decrease in mortality in the mid to late 20th century in developing economies, in particular among infants and children, produced cohorts characterized by increasing survivorship of poor early life conditions (nutrition and infectious diseases in particular). Improvements in early life health occurred largely in the context of stagnant improvements in standard of living resulting in continued exposure to poor environmental conditions during childhood. Poor early life conditions not only affected childhood health but also may affect the health of these unique cohorts at older ages (Palloni & Souza, 2013). There is now ample evidence of the long-term consequences of poor childhood conditions for older adult health (Barker, 1998; Bateson & Gluckman, 2011; Crimmins & Finch, 2006; Elo & Preston, 1992; Gluckman & Hanson, 2005). The historical circumstances of the 20th century provide a pertinent explanation for the SES health gradient among older adults in developing economies (McEniry, 2014; Palloni & Souza, 2013). Poor early life nutrition can affect cognitive and non-cognitive skills that lead to poor educational attainment (Martorell, Stein, & Schroeder, 2001). Cohorts increasingly characterized by survivorship of poor early life conditions could result in a larger group of individuals in low SES groups with lower educational attainment and less opportunity for social mobility producing sharper SES health disparities (Palloni & Souza, 2013). Alternatively, for these unique cohorts, mismatches between early life and later nutritional environments, in combination with increased social mobility among lower SES groups and negative health behavior among higher SES groups, could explain weak SES gradients. Predictive physiological and metabolic adaptations to an undernourished environment in early life lead to mismatches later in life if there is more favorable nutrition-enriched environments at older ages, resulting in disease (Bateson & Gluckman, 2011; Gluckman & Hanson, 2005). Improved economic status and social mobility for lower SES groups born into poor nutritional environments could lead to changes in food consumption resulting in a more enriched diet higher in saturated fat content. These circumstances could produce mismatches, increasing the risk of obesity, diabetes, and heart disease. Weak gradients appear because there are fewer differences between middle and higher SES groups in terms of diet, especially if higher SES groups also engage in negative health behavior and consume a diet rich in saturated fat. Taken as a whole, the complexity of historical circumstances in developing economies in transition and the importance of early life conditions may produce conflicting patterns of the SES gradient for health—reversals, sharp or weak gradients—in contrast to what occurred in developed economies with different historical circumstances. The demographic regimes of high-income countries in the 20th century are characterized by earlier and more gradual mortality decline (early 1900s or before) attributed to both improvements in standard of living and public health (McEniry, 2014). These regimes produced the pervasive increase in the older adult population and in chronic conditions and the established pattern of low SES being associated with poor health. In contrast, these increases are a more recent phenomenon in middle-income countries, more evident in urban areas, and more characterized by survivorship of poor early life conditions. Our Study Our study expands upon other studies of the SES health gradient in middle-income countries in two important ways. First, we compare the SES gradient and health across developed and developing world settings in the context of the demographic regimes of the 20th century. One type of regime (urban settings in upper-middle income countries) is characterized by rapid mortality decline without parallel improvement in standard of living and which has resulted in relatively recent increases in the aging population and chronic conditions. The other type of regime (high-income countries) experienced an earlier, more gradual mortality decline, where increases in the older population and chronic conditions have been pervasive for quite some time. Although other authors have made cross-national comparisons of older adult health in middle and high-income countries (Goldman et al., 2011; Payne et al., 2017), there have been few studies that use the framework of the demographic regimes of the 20th century to examine the health of older adults born during those regimes. Second, we compare gradients for obesity, diabetes, and hypertension relative to several other health outcomes along with adult lifestyle and early life nutritional status. With a notable exception (Rosero-Bixby & Dow, 2009), there have been few studies which have examined the SES gradient with such a diversity of outcomes in middle-income countries. We select three upper middle-income countries (Colombia, Mexico, and South Africa) and two high-income countries (United States and England) with the purpose of comparing, where possible, SES gradients from the study of the upper middle-income country of Costa Rica (Rosero-Bixby & Dow, 2009). The selected middle-income countries experienced similar rapid demographic transitions in the mid-20th century, but are also similar in terms of present-day country-level national income. The United States and England are settings with similar demographic transitions and levels of income but where health results and health systems are quite different. Rapid mortality decline occurred in Colombia beginning in the mid-1930s (Flórez et al., 2016) and in Mexico and South Africa in the 1940s, although, in the case of South Africa, political and economic events may have had slowed improvements for segments of the population until much later (Beinart & Dubow, 1995; Riley, 2005; de Romo & de Pérez, 1998). Poor nutritional environments were pervasive for older adults born in these settings during the 1930s–1940s although there was rapid improvement in nutrition in the decades after, in particular for Colombia and Mexico (FAO, 1946, 2010). Both Colombia and Mexico experienced some degree of increased social mobility toward the end of the 20th and beginning of the 21st centuries (directly benefiting either older adults or their children), increasing purchasing power for families, particularly in urban areas (Franco, Leon, & Atria, 2007). The selected middle-income settings show similarities in per capita gross national income (World Bank Group, 2017b) and human development (United Nations Development Program, 2016). Nevertheless, there are country differences in life expectancy, health systems, inequality, and history in the selected settings. Life expectancy at age 60 is lowest in South Africa (World Health Organization, 2015) and the quality of its health system is rated much lower than either Colombia or Mexico (World Health Organization, 2000). Apartheid dominated the political and economic system in South Africa for much of the mid to late 20th century; the HIV epidemic greatly impacted population health (Hosegood & Timaeus, 2006). Inequality is much higher in South Africa than in Colombia and Mexico; South Africa is one of the most unequal countries in the world; inequality in the United States and England is much lower than all three middle-income countries (World Bank Group, 2017a). Large discrepancies exist between rural and urban areas in the middle-income countries. However, although there is greater concentration of wealth and resources in urban settings, better economic opportunities, and better access to health care than rural areas, inequality is higher in urban areas than in rural areas in Colombia, Mexico, and South Africa (Bhorat, van der Westhuizen, & Jacobs, 2009; World Bank Group, 2017c). Our focus is on the direction and magnitude of the SES gradient for obesity, diabetes, and hypertension relative to other health outcomes and lifestyle. We examine reversals in the gradient and the degree to which the gradient is sharper or weaker than expected especially for obesity, diabetes, and hypertension. We also examine the degree to which there are similar patterns across countries with other health outcomes and behaviors. We expect that the gradient will be consistently strong and positive in high-income countries and mixed in middle-income settings. We expect to observe sharper inequalities in the middle-income countries, in particular in South Africa. Using the framework of rapid demographic, epidemiological, and nutritional transitions, we identify the most plausible explanations for cross-national differences and similarities in the SES health gradient for the middle-income settings. Methods Data Data on older Colombian adults come from SABE-Bogotá. SABE Bogotá is a representative cross-sectional sample of 2,000 older adults, 60 years and older within the metropolitan area of Bogotá from 2012 and is based on the original SABE study carried out in seven major cities of Latin America and the Caribbean. We select respondents 60 years and older living in urban areas in Mexico and South Africa using the WHO Study on Global Ageing and Adult Health Study (SAGE, 2007–2008) and in the United States and England using the Health and Retirement Study (HRS, 2000, 2006–2008) and the English Longitudinal Study on Ageing (ELSA, 2007–2008). For the United States, we obtained measured hypertension, diabetes, and performance measures first collected in 2006–2008. Race is an important distinguishing factor in South Africa; we focus on the two majority racial groups in the SAGE survey using their racial categories of South African blacks and colored. The final sample sizes were Bogotá (2,000), Mexico (1,286), South Africa (911), the United States (12,527), and England (6,183). The data form part of a NIA-funded project to harmonize data from major surveys of older adults or households (McEniry, 2015). The studies are representative of the older adult population either nationally or in major country provinces. All studies had good interviewer training, good questionnaire design, concern for data quality, and high response rates. All selected surveys collected measured blood pressure in addition to other dimensions of physical and mental health. Measures We use education for SES because education is a comparable measure of SES in older adults in cross-national studies (Goldman et al., 2011). However, the definition of educational categories can influence study results (Montez, Hummer, & Hayward, 2012). Because educational attainment differs between middle-income and high-income countries, we use a slightly different coding of education to reflect low, middle, and high educational groups. For the middle-income countries, we use three categories similar to educational categories used in other cross-national studies of middle-income countries (Goldman et al., 2011; Rosero-Bixby & Dow, 2009): incomplete primary, incomplete secondary, and complete secondary plus. For the United States and England, we use three categories defined by cross-national studies (Banks, Marmot, Oldfield, & Smith, 2006): low (0–11 years England, 0–12 years United States), middle (12–13 years England, 13–15 years United States), and high (more than 13 years England, more than 15 years United States). For health outcomes, we define obesity as a body mass index (BMI) of greater than or equal to 30. We define hypertension as systolic blood pressure rates greater than or equal to 140 mmHg, diastolic blood pressure rates greater than or equal to 90 mmHG, or taking medication to control hypertension. Diabetes is self-reported except for in the United States and England where we also include diabetes based on the biomarker HbA1c using a cutoff point of 6.5% or greater (of total hemoglobin), or taking diabetes medication (Yan et al., 2012). Although important in some settings, underestimation of diabetes may not be so problematic in some middle-income countries such as Colombia (Vargas-Uricoechea & Casas-Figueroa, 2015). Other chronic conditions are self-reported (heart disease, arthritis, respiratory disease, cancer, and stroke). We define poor self-reported health (SRH) based on a question asking respondents about their overall health. We use comparable and well-validated measures across countries for other dimensions of health, using established cutoff points where possible. Cognitive impairment is based on learning and immediate recall of words. For SABE-Bogotá, we defined impaired cognition as a short-term memory of 0–1 out of 3 words from a modified version of the Minimental State Exam. HRS, ELSA, and SAGE all had immediate recall of 10 words, with a cutoff point of less than 4 to define impaired cognition. Depression is defined in terms of major depressive symptoms. SABE-Bogotá uses the Geriatric Depression Scale (Yesavage et al., 1982–1983), which includes 15 items, and major depression is identified by scoring 10 points or higher on this scale. Depression in the SAGE surveys is based on the DSM IV criteria reviewed by psychiatrists. We define major depression as reporting at least one of two major symptom categories (mood and loss of interest), daily symptoms lasting at least 2 weeks, plus having at least four additional symptoms (insomnia, weight loss, psychomotor agitation, feelings of worthlessness, diminished ability to think, recurrent thoughts about death, or low energy). HRS and ELSA use the CES-D (Radloff, 1977) scale based on eight items. We define a score of 4 or greater as depressed. Difficulty with functionality is harmonized across countries and based on difficulty with at least one ADL (bathing, dressing, transferring, or toilet use). We define trouble walking (an indicator of frailty) in terms of the highest quintile of time to perform a normal walking test and weak grip strength in terms of a dynamometer reading of less than or equal to 20 kg. For adult lifestyle or risky health behavior (smoking, exercise, and diet), we define smoking as those who currently smoke, no rigorous exercise as those who do not perform consistent and weekly rigorous exercise, and poor diet as those who consume less than two servings of fruits and vegetables daily. We use the first quartile of height as a marker of poor early life nutritional status. Analyses For each country and health outcome and behavior, we estimate logistic regression models using age, gender, and educational groups with the reference group as the middle educational group. We control for race in South African models. To put our analyses of South African blacks and colored into better perspective, we estimate models for the entire population of South African older adults living in urban areas including the white and Asian populations and we estimate models for U.S. blacks. Comparing similar racial/ethnic groups may help better understand health disparities; national context is important and environmental factors differ across countries (Agyemang et al., 2010). Because there may be differences in cohorts in terms of diet, we estimate models with interactions between age and education for obesity, hypertension, and diabetes as a very broad dietary evaluation of cohorts. We summarize results with graphs showing patterns of the gradient across all countries along with joint SES significance levels. We estimate predicted probabilities and calculate the ratio of the predicted probabilities between low and high education groups as a measure of the magnitude of the gradient (Beltrán-Sánchez et al., 2016). To reduce the amount of cluttering in tables we show only the odds ratio with its significance level. The 95% confidence intervals for models are available upon request. Missing values were more notable in the middle-income countries (see Supplementary Notes for an explanation regarding missing values). Results Sample Characteristics and Health Profiles The selected samples are all similar in terms of a majority female composition (56%–61%) with the exception of Mexico where only 51% are female (Table 1). Age patterns are similar across all countries but there are noted differences in education. The large metropolitan area of Bogotá has a higher percent (26%) of respondents with at least a secondary educational level as compared with urban Mexican and South African respondents (10% and 11%). The United States and England have similar percent in the lowest educational group (63% and 60%). Table 1. Basic Sample Characteristics Colombia Bogotá Mexico Urban S Africa Urban England United States Female (%) 57 51 61 56 59 Age group (%)  60–64 35 24 34 23 24  65–69 25 22 31 22 22  70–74 17 17 14 18 20  75–79 12 19 9 16 17  80+ 12 18 12 20 17 Education (%)  Low 39 57 56 60 63  Middle 35 34 34 15 18  High 26 10 11 26 18 Colombia Bogotá Mexico Urban S Africa Urban England United States Female (%) 57 51 61 56 59 Age group (%)  60–64 35 24 34 23 24  65–69 25 22 31 22 22  70–74 17 17 14 18 20  75–79 12 19 9 16 17  80+ 12 18 12 20 17 Education (%)  Low 39 57 56 60 63  Middle 35 34 34 15 18  High 26 10 11 26 18 Source. SABE Bogotá; SAGE-Mexico, urban dwellers; SAGE-South Africa, urban dwellers, African blacks and colored; ELSA-England; HRS-US. Note: Weighted; numbers do not always sum to 100 due to rounding. View Large Table 1. Basic Sample Characteristics Colombia Bogotá Mexico Urban S Africa Urban England United States Female (%) 57 51 61 56 59 Age group (%)  60–64 35 24 34 23 24  65–69 25 22 31 22 22  70–74 17 17 14 18 20  75–79 12 19 9 16 17  80+ 12 18 12 20 17 Education (%)  Low 39 57 56 60 63  Middle 35 34 34 15 18  High 26 10 11 26 18 Colombia Bogotá Mexico Urban S Africa Urban England United States Female (%) 57 51 61 56 59 Age group (%)  60–64 35 24 34 23 24  65–69 25 22 31 22 22  70–74 17 17 14 18 20  75–79 12 19 9 16 17  80+ 12 18 12 20 17 Education (%)  Low 39 57 56 60 63  Middle 35 34 34 15 18  High 26 10 11 26 18 Source. SABE Bogotá; SAGE-Mexico, urban dwellers; SAGE-South Africa, urban dwellers, African blacks and colored; ELSA-England; HRS-US. Note: Weighted; numbers do not always sum to 100 due to rounding. View Large Large differences exist in health outcomes and risk factors across countries (Table 2). We observe particularly higher percentages with obesity, hypertension, and trouble walking in South Africa (48%, 86%, and 29%), heart disease in the United States and England (25% and 29%), and arthritis in the United States (61%). For overall health, we find high percentages in poor SRH in Bogotá (48%), cognitive impairment in Mexico (40%), and weak grip strength in Bogotá and urban Mexico (48% and 46%). For risky behavior, we find a high percent of older adults do no regular strenuous exercise in the middle-income countries (88%–90%) or in England (86%), and eat less than two servings of fruit or vegetables in Bogotá and urban Mexico (33% and 35%). On average, female respondents in middle-income country settings are between 3–9 (7–13) cm shorter than their English (United States) counterparts. Likewise, male respondents are 9 (14) cm shorter than their English (United States) counterparts. Table 2. Overall Prevalence of Health/Risk Factors Colombia Bogotá Mexico Urban S Africa Urban England United States Overall health (%) Poor self-reported health 48 14 17 32 26 Difficulty with functionality 6 25 25 25 16 Trouble walking 21 19 29 21 19 Weak grip strength 48 46 27 29 28 Cognitive impairment 10 40 18 15 16 Depression 7 8 1 16 15 Chronic conditions (%) Hypertension (measured) 76 73 86 54 73 Diabetes 18 21 12 9 22 Heart disease 5 10 10 29 25 Stroke 4 6 5 7 7 Respiratory disease 15 7 4 8 10 Arthritis 28 13 30 42 61 Cancer 6 9 14 Adult risk factors (%) Obesity 26 30 48 28 21 Current smoker 8 21 23 13 12 No vigorous exercise 90 89 88 86 57 Less than 2 fr/veg daily 33 35 21 Early life nutritional status Height (cm)  Female 150 (6) 149 (7) 155 (11) 158 (7) 162 (7)  Male 163 (7) 163 (7) 163 (10) 172 (7) 177 (7) Colombia Bogotá Mexico Urban S Africa Urban England United States Overall health (%) Poor self-reported health 48 14 17 32 26 Difficulty with functionality 6 25 25 25 16 Trouble walking 21 19 29 21 19 Weak grip strength 48 46 27 29 28 Cognitive impairment 10 40 18 15 16 Depression 7 8 1 16 15 Chronic conditions (%) Hypertension (measured) 76 73 86 54 73 Diabetes 18 21 12 9 22 Heart disease 5 10 10 29 25 Stroke 4 6 5 7 7 Respiratory disease 15 7 4 8 10 Arthritis 28 13 30 42 61 Cancer 6 9 14 Adult risk factors (%) Obesity 26 30 48 28 21 Current smoker 8 21 23 13 12 No vigorous exercise 90 89 88 86 57 Less than 2 fr/veg daily 33 35 21 Early life nutritional status Height (cm)  Female 150 (6) 149 (7) 155 (11) 158 (7) 162 (7)  Male 163 (7) 163 (7) 163 (10) 172 (7) 177 (7) Source. SABE Bogotá; SAGE-Mexico, urban dwellers; SAGE-South Africa, urban dwellers, African blacks and colored; ELSA-England; HRS-US. Note: Early life nutritional status shows averages with standard deviations. Weighted. Prevalence for diabetes in England and United States based on biomarkers. View Large Table 2. Overall Prevalence of Health/Risk Factors Colombia Bogotá Mexico Urban S Africa Urban England United States Overall health (%) Poor self-reported health 48 14 17 32 26 Difficulty with functionality 6 25 25 25 16 Trouble walking 21 19 29 21 19 Weak grip strength 48 46 27 29 28 Cognitive impairment 10 40 18 15 16 Depression 7 8 1 16 15 Chronic conditions (%) Hypertension (measured) 76 73 86 54 73 Diabetes 18 21 12 9 22 Heart disease 5 10 10 29 25 Stroke 4 6 5 7 7 Respiratory disease 15 7 4 8 10 Arthritis 28 13 30 42 61 Cancer 6 9 14 Adult risk factors (%) Obesity 26 30 48 28 21 Current smoker 8 21 23 13 12 No vigorous exercise 90 89 88 86 57 Less than 2 fr/veg daily 33 35 21 Early life nutritional status Height (cm)  Female 150 (6) 149 (7) 155 (11) 158 (7) 162 (7)  Male 163 (7) 163 (7) 163 (10) 172 (7) 177 (7) Colombia Bogotá Mexico Urban S Africa Urban England United States Overall health (%) Poor self-reported health 48 14 17 32 26 Difficulty with functionality 6 25 25 25 16 Trouble walking 21 19 29 21 19 Weak grip strength 48 46 27 29 28 Cognitive impairment 10 40 18 15 16 Depression 7 8 1 16 15 Chronic conditions (%) Hypertension (measured) 76 73 86 54 73 Diabetes 18 21 12 9 22 Heart disease 5 10 10 29 25 Stroke 4 6 5 7 7 Respiratory disease 15 7 4 8 10 Arthritis 28 13 30 42 61 Cancer 6 9 14 Adult risk factors (%) Obesity 26 30 48 28 21 Current smoker 8 21 23 13 12 No vigorous exercise 90 89 88 86 57 Less than 2 fr/veg daily 33 35 21 Early life nutritional status Height (cm)  Female 150 (6) 149 (7) 155 (11) 158 (7) 162 (7)  Male 163 (7) 163 (7) 163 (10) 172 (7) 177 (7) Source. SABE Bogotá; SAGE-Mexico, urban dwellers; SAGE-South Africa, urban dwellers, African blacks and colored; ELSA-England; HRS-US. Note: Early life nutritional status shows averages with standard deviations. Weighted. Prevalence for diabetes in England and United States based on biomarkers. View Large Health Outcomes The effects of education on several health outcomes show expected consistent and significant patterns of lowest educational group being most at risk of being in poor health in the United States and England (Table 3). In Colombia, strong gradients appear in all variables except trouble walking and grip strength, and in Mexico, strong gradients appear with the exception of grip strength. The gradients in South Africa are significant for poor SRH, functionality, and trouble walking which shows a reversal in the gradient. The total urban South African population of older adults shows similar patterns with the blacks and colored; U.S. blacks tend to display sharper disparities than South African blacks/colored (Supplementary Table S1). Table 3. Effects of Education on Selected Health Outcomes Colombia Bogotá Mexico Urban S Africa Urban England United States Poor SRH L 1.46*** 1.87** 1.55* 1.96*** 2.14*** H 0.41*** 0.59 0.58 0.74** 0.68*** Difficulty with Functionality L 1.68** 1.88*** 0.71 1.47*** 1.50*** H 0.77 1.08 0.23*** 0.91 0.99 Trouble Walking L 1.22 1.85** 0.36*** 2.05*** 2.06*** H 0.79 0.62 1.43 0.83 1.03 Weak Grip Strength L 1.29 0.94 1.01 1.53*** 1.31** H 1.06 1.52 0.46 0.85 1.03 Cognitive Impairment L 1.86* 1.87*** 1.33 2.48*** 2.17*** H 0.77 0.78 0.46 0.90 0.67** Depression L 1.51* 2.04** 1.10 1.79*** 1.91*** H 0.38** 0.79 1.13 0.98 0.76** Colombia Bogotá Mexico Urban S Africa Urban England United States Poor SRH L 1.46*** 1.87** 1.55* 1.96*** 2.14*** H 0.41*** 0.59 0.58 0.74** 0.68*** Difficulty with Functionality L 1.68** 1.88*** 0.71 1.47*** 1.50*** H 0.77 1.08 0.23*** 0.91 0.99 Trouble Walking L 1.22 1.85** 0.36*** 2.05*** 2.06*** H 0.79 0.62 1.43 0.83 1.03 Weak Grip Strength L 1.29 0.94 1.01 1.53*** 1.31** H 1.06 1.52 0.46 0.85 1.03 Cognitive Impairment L 1.86* 1.87*** 1.33 2.48*** 2.17*** H 0.77 0.78 0.46 0.90 0.67** Depression L 1.51* 2.04** 1.10 1.79*** 1.91*** H 0.38** 0.79 1.13 0.98 0.76** Source. SABE Bogotá; SAGE-Mexico, urban dwellers; SAGE-South Africa, urban dwellers, African blacks and colored; ELSA-England; HRS-US. Note: Shown are odds ratios by education group (low/high). Reference group is middle educational group. Models significantly differently from null model at p < .001 for England and the United States; models also significant in other countries with some exceptions (see note in Supplementary Material). SRH = Self-reported health. *p < .05; **p < .01; ***p < .001. View Large Table 3. Effects of Education on Selected Health Outcomes Colombia Bogotá Mexico Urban S Africa Urban England United States Poor SRH L 1.46*** 1.87** 1.55* 1.96*** 2.14*** H 0.41*** 0.59 0.58 0.74** 0.68*** Difficulty with Functionality L 1.68** 1.88*** 0.71 1.47*** 1.50*** H 0.77 1.08 0.23*** 0.91 0.99 Trouble Walking L 1.22 1.85** 0.36*** 2.05*** 2.06*** H 0.79 0.62 1.43 0.83 1.03 Weak Grip Strength L 1.29 0.94 1.01 1.53*** 1.31** H 1.06 1.52 0.46 0.85 1.03 Cognitive Impairment L 1.86* 1.87*** 1.33 2.48*** 2.17*** H 0.77 0.78 0.46 0.90 0.67** Depression L 1.51* 2.04** 1.10 1.79*** 1.91*** H 0.38** 0.79 1.13 0.98 0.76** Colombia Bogotá Mexico Urban S Africa Urban England United States Poor SRH L 1.46*** 1.87** 1.55* 1.96*** 2.14*** H 0.41*** 0.59 0.58 0.74** 0.68*** Difficulty with Functionality L 1.68** 1.88*** 0.71 1.47*** 1.50*** H 0.77 1.08 0.23*** 0.91 0.99 Trouble Walking L 1.22 1.85** 0.36*** 2.05*** 2.06*** H 0.79 0.62 1.43 0.83 1.03 Weak Grip Strength L 1.29 0.94 1.01 1.53*** 1.31** H 1.06 1.52 0.46 0.85 1.03 Cognitive Impairment L 1.86* 1.87*** 1.33 2.48*** 2.17*** H 0.77 0.78 0.46 0.90 0.67** Depression L 1.51* 2.04** 1.10 1.79*** 1.91*** H 0.38** 0.79 1.13 0.98 0.76** Source. SABE Bogotá; SAGE-Mexico, urban dwellers; SAGE-South Africa, urban dwellers, African blacks and colored; ELSA-England; HRS-US. Note: Shown are odds ratios by education group (low/high). Reference group is middle educational group. Models significantly differently from null model at p < .001 for England and the United States; models also significant in other countries with some exceptions (see note in Supplementary Material). SRH = Self-reported health. *p < .05; **p < .01; ***p < .001. View Large Graphical depictions of the gradients by ordering the regression odds ratios summarize these initial general impressions for Colombia, Mexico, South Africa, United States (Figure 1), and England, which is similar to the United States (Supplementary Figure 1). Although the order of the odds ratios differs across countries, the general patterns for these aspects of health are similar in Colombia, Mexico, the United States, and England especially for cognitive impairment, depression, functionality, and poor SRH. South Africa shows fewer significant associations between education and health/adult lifestyle and more variation in the SES patterns. Figure 1. View largeDownload slide Cross national patterns of SES health disparities. SES = Socioeconomic status; see Supplementary Figure 2. Figure 1. View largeDownload slide Cross national patterns of SES health disparities. SES = Socioeconomic status; see Supplementary Figure 2. Predicted probabilities show the sharpest gradients in the middle-income countries in functionality, cognitive impairment, and depression (Colombia), and SRH, trouble walking (Mexico) and weak grip strength (South Africa) (Figure 2, Supplementary Table S3). The probability of poor health for the lowest SES group is between two to nearly four times that of the probability of poor health for those in the highest SES group for these aspects of health in the middle-income countries. Figure 2. View largeDownload slide Ratio of predicted probabilities for low/high SES. SES = Socioeconomic status; see Supplementary Figure 3. Figure 2. View largeDownload slide Ratio of predicted probabilities for low/high SES. SES = Socioeconomic status; see Supplementary Figure 3. Chronic Conditions and Obesity The middle-income settings show a weaker pattern for the effects of obesity, diabetes, and hypertension with the exception of obesity and diabetes in South Africa where a reversal of the gradient appears showing a strong likelihood of being obese among high SES and protective effects of being obese and diabetic for low SES (Table 4). Models for obesity with all urban South African older adults show similar results for low SES but the significant effect in high SES disappears; U.S. blacks tend to display sharper disparities than South African blacks, although the gradients for obesity and diabetes are flat for U.S. blacks (Supplementary Table S1). High SES shows protective effects on diabetes (Colombia) and hypertension and obesity (Mexico). In contrast, strong gradients in England and the United States appear in the expected direction. Predicted probabilities (Figure 2, Supplementary Table S3) show weaker gradients for hypertension, diabetes, and obesity across all countries, although the gradient is sharpest for diabetes in Colombia and Mexico and sharpest for obesity in the United States and England. The predicted probability of diabetes for the lowest SES group is almost 1.5 times the predicted probability of diabetes for the highest SES group in Colombia and Mexico; similarly for obesity in the United States and England (Figure 2, Supplementary Table S3). Models with interactions for age and education produced no consistent significant interactions across countries and are not shown. Table 4. Effects of Education on Hypertension, Diabetes, Obesity, Adult Risk Factors, Childhood Conditions Colombia Bogotá Mexico Urban S Africa Urban England United States Hypertension L 1.09 0.96 1.14 1.15 1.36*** H 0.86 0.57* 1.59 0.82* 0.94 Diabetes L 1.03 1.05 0.64* 1.40* 1.38*** H 0.69* 0.64 1.21 0.96 0.93 Obesity L 0.92 0.77 0.48*** 1.54*** 1.38*** H 0.85 0.55* 2.72*** 0.93 0.82* Smoker L 1.10 1.10 1.86*** 1.89*** 1.33*** H 0.79 0.82 0.45 0.73* 0.57*** No Vigorous Exercise L 1.78* 0.64* 0.78 1.61*** 1.36*** H 0.65* 0.43** 0.92 0.77* 0.95 LT 2 Fr/Veg Daily L 1.64*** 1.42** 2.29*** H 0.49*** 0.81 0.35* Low Height L 1.80*** 1.79*** 1.16 1.70*** 1.49*** H 0.77 0.78 0.92 0.83 0.85 Colombia Bogotá Mexico Urban S Africa Urban England United States Hypertension L 1.09 0.96 1.14 1.15 1.36*** H 0.86 0.57* 1.59 0.82* 0.94 Diabetes L 1.03 1.05 0.64* 1.40* 1.38*** H 0.69* 0.64 1.21 0.96 0.93 Obesity L 0.92 0.77 0.48*** 1.54*** 1.38*** H 0.85 0.55* 2.72*** 0.93 0.82* Smoker L 1.10 1.10 1.86*** 1.89*** 1.33*** H 0.79 0.82 0.45 0.73* 0.57*** No Vigorous Exercise L 1.78* 0.64* 0.78 1.61*** 1.36*** H 0.65* 0.43** 0.92 0.77* 0.95 LT 2 Fr/Veg Daily L 1.64*** 1.42** 2.29*** H 0.49*** 0.81 0.35* Low Height L 1.80*** 1.79*** 1.16 1.70*** 1.49*** H 0.77 0.78 0.92 0.83 0.85 Source. SABE Bogotá; SAGE-Mexico, urban dwellers; SAGE-South Africa, urban dwellers, African blacks and colored; ELSA-England; HRS-US. Note: Shown are odds ratios by education group (low/high). Reference group is middle educational group. Models significantly differently from null model at p < .001 for England and the United States; models also significant in other countries with some exceptions (see note in Supplementary Material). Models for diabetes for England and the United States use diabetes based on biomarkers; models using self-reported diabetes produced similar patterns and are not shown in table. *p < .05; **p < .01; ***p < .001. View Large Table 4. Effects of Education on Hypertension, Diabetes, Obesity, Adult Risk Factors, Childhood Conditions Colombia Bogotá Mexico Urban S Africa Urban England United States Hypertension L 1.09 0.96 1.14 1.15 1.36*** H 0.86 0.57* 1.59 0.82* 0.94 Diabetes L 1.03 1.05 0.64* 1.40* 1.38*** H 0.69* 0.64 1.21 0.96 0.93 Obesity L 0.92 0.77 0.48*** 1.54*** 1.38*** H 0.85 0.55* 2.72*** 0.93 0.82* Smoker L 1.10 1.10 1.86*** 1.89*** 1.33*** H 0.79 0.82 0.45 0.73* 0.57*** No Vigorous Exercise L 1.78* 0.64* 0.78 1.61*** 1.36*** H 0.65* 0.43** 0.92 0.77* 0.95 LT 2 Fr/Veg Daily L 1.64*** 1.42** 2.29*** H 0.49*** 0.81 0.35* Low Height L 1.80*** 1.79*** 1.16 1.70*** 1.49*** H 0.77 0.78 0.92 0.83 0.85 Colombia Bogotá Mexico Urban S Africa Urban England United States Hypertension L 1.09 0.96 1.14 1.15 1.36*** H 0.86 0.57* 1.59 0.82* 0.94 Diabetes L 1.03 1.05 0.64* 1.40* 1.38*** H 0.69* 0.64 1.21 0.96 0.93 Obesity L 0.92 0.77 0.48*** 1.54*** 1.38*** H 0.85 0.55* 2.72*** 0.93 0.82* Smoker L 1.10 1.10 1.86*** 1.89*** 1.33*** H 0.79 0.82 0.45 0.73* 0.57*** No Vigorous Exercise L 1.78* 0.64* 0.78 1.61*** 1.36*** H 0.65* 0.43** 0.92 0.77* 0.95 LT 2 Fr/Veg Daily L 1.64*** 1.42** 2.29*** H 0.49*** 0.81 0.35* Low Height L 1.80*** 1.79*** 1.16 1.70*** 1.49*** H 0.77 0.78 0.92 0.83 0.85 Source. SABE Bogotá; SAGE-Mexico, urban dwellers; SAGE-South Africa, urban dwellers, African blacks and colored; ELSA-England; HRS-US. Note: Shown are odds ratios by education group (low/high). Reference group is middle educational group. Models significantly differently from null model at p < .001 for England and the United States; models also significant in other countries with some exceptions (see note in Supplementary Material). Models for diabetes for England and the United States use diabetes based on biomarkers; models using self-reported diabetes produced similar patterns and are not shown in table. *p < .05; **p < .01; ***p < .001. View Large No significant gradients appear in other chronic conditions (heart disease, stroke, arthritis, or respiratory disease) in Colombia, Mexico, or South Africa with the exception of respiratory disease in Colombia. There are strong gradients for these chronic conditions in the United States and England and a reversal of the gradient appears for cancer in Colombia, England, and the United States (Supplementary Table S2). South African urban residents and blacks/colored show similar patterns; the gradient for U.S. blacks is mostly flat with the exception of arthritis (Supplementary Table S2). Predicted probabilities (Supplementary Table S3) show the sharpest gradients appear mostly in the middle-income countries with the exception of arthritis and respiratory disease. Colombia shows the sharpest gradient for stroke and cancer and South Africa the sharpest gradient for heart disease. The predicted probability of stroke for the lowest SES group is nearly three times that of the highest SES group in Colombia. The predicted probability of heart disease for the lowest SES group in South Africa is nearly two times that of the highest SES group. Being in the lowest SES group reduces the probability of reporting cancer by about 50% as compared to being in the highest SES group in Colombia. The predicted probability of respiratory disease (arthritis) for the lowest SES group in England is about two (1.3) times more than that of the highest SES group. Risky Adult Behavior and Early Life Nutritional Status We find strong gradients for smoking in South Africa, England, and the United States (Table 4). Strong gradients also appear for exercise in Colombia, England, and the United States but a reversal in low SES in Mexico and no significant gradient in South Africa. Strong gradients appear for consumption of fruits and vegetables across all middle-income settings. With the exception of South Africa, strong gradients appear for low height across all countries. South African urban residents follow similar patterns as South African urban blacks and colored; U.S. blacks show a sharp gradient for exercise but flatter gradients for smoking and height (Supplementary Table S1). Predicted probabilities show a weak gradient for exercise across all countries with the exception of the United States. The sharpest gradients for smoking and consumption of fruits and vegetables appear in South Africa and for low height in Colombia (Supplementary Table S3). The predicted probability of being a current smoker (poor diet) for the lowest SES group is more than three (almost 5) times that of the highest SES group in South Africa. The predicted probability of short height for the lowest SES group is over two times that of the highest SES group in Colombia. Discussion We examined the SES health gradient using education for an extensive group of health outcomes and risky adult behavior in three middle-income urban settings, benchmarking with England and the United States within the framework of the demographic and nutritional transitions of the 20th century. High-income countries consistently showed strong gradients for health in the expected direction (low SES predicts poor health) whereas mixed patterns appeared in middle-income countries. The sharpest gradients appeared mostly in middle-income countries but followed expected patterns found in the United States and England especially in health outcomes such as poor SRH, functionality, trouble walking, cognitive impairment, and depression; however, weaker gradients appeared for obesity, hypertension, diabetes, and other chronic conditions in Colombia and Mexico with reversals observed in South Africa. Reversal of the gradient appeared for cancer in Bogotá, England, and the United States. Strong disparities existed in risky health behavior related to diet and early life nutritional status in the middle-income settings. The results contribute to expanding knowledge regarding cross-national comparisons of the SES health gradient in middle- and high-income settings. There have been few cross-national studies that examine the SES gradient across such a diverse group of health outcomes and risky health behaviors. The selection of urban populations in Colombia, Mexico, and South Africa is relevant given the increased prevalence of obesity and chronic conditions among older adults in the context of rapid demographic and nutritional transitions, increased urbanization, and high inequality. We do not claim that our selection of middle-income countries is representative of other upper middle-income countries. However, the selected countries are comparable in terms their demographic, epidemiological, and nutritional transitions and level of national income. The comparison with United States and England provides the perspective of well-established patterns from developed countries, which experienced much earlier growth of the older adult population and associated chronic conditions. We highlight four main points in regards to the SES gradient and health in middle-income countries. First, high level of inequality due to rapid urbanization is a relevant explanation for the sharpness of the gradient found in some health outcomes. Consistent with other studies in middle-income countries (Rosero-Bixby & Dow, 2009) and high-income countries (Banks et al., 2006), our study showed similar patterns in the expected direction for health such as poor SRH, cognition, depression, functionality, and trouble walking. However, the sharpest gradients appear in urban settings in middle-income countries. Sharper gradients are not surprising given that there are high levels of inequality found in urban settings in middle-income countries (World Bank Group, 2017a). Rapid urbanization accompanied the rapid demographic and nutritional transitions affecting middle-income countries (Flórez et al., 2016). Second, the combination of rapid transitions, poor early life nutritional environment, social mobility, and an enriched nutritional environment later in life provides a pertinent explanation for the weaker gradients for obesity, diabetes, and hypertension in Colombia and Mexico. The sharp disparities in poor early life nutritional status and current diet suggest a history of poor nutrition among low SES groups. The argument for the switch in the gradient in health conditions such as obesity according to country-level gross national income (Monteiro et al., 2004) does not adequately explain the patterns encountered in obesity in the middle-income countries in our study. The weaker gradient for obesity (particularly in Colombia and Mexico) and the differences noted in the gradient (particularly in South Africa) suggest multiple and opposing factors. The historical circumstances of cohorts born in the 20th century and characterized by increasing survivorship of poor early life conditions (Palloni & Souza, 2013) combined with uneven development affecting different SES groups and changes in food consumption is a plausible explanation for the weaker gradients. Both Colombia and Mexico experienced increased social mobility in the late 20th and early 21st century (Franco et al., 2007), potentially resulting in changes in food consumption characterized by higher consumption of saturated fats and processed foods and producing higher risk of obesity, diabetes, and hypertension. Older adults from cohorts of the 20th century with increased social mobility may be experiencing a mismatch in nutritional environment later in life (Bateson & Gluckman, 2011; Gluckman & Hanson, 2005) leading to higher risk of disease. In contrast, those in the lowest SES may not yet be experiencing large nutritional changes and not everyone in the highest SES may be better consumers of healthy food and value healthy behavior. Thus, even though the determinants of obesity, diabetes, and hypertension are different between groups (higher SES due to adult lifestyle and middle SES due to early life plus adult lifestyle), weaker gradients appear between middle and high SES groups. The weaker gradients could partially reflect a weaker association among educational attainment in the selected middle-income settings. However, improved access to education among lower SES groups due to public policy is probably not a major confounding factor in explaining the gradient in settings such as Colombia and Mexico because strong gradients appear with other health outcomes such as cognitive impairment, depression, functionality, and poor SRH. On the other hand, weaker gradients could reflect the long-term negative consequences of economic inequality unique to segments of the South African population. Educational attainment for some segments of the South African population may not have resulted in increased wealth. The coverage and quality of the health care system in countries such as Colombia, noted to have a good health system (World Health Organization, 2000), could provide an explanation for the weaker gradients. Better health observed in several health outcomes among Colombians due to a good health care system and better health across all SES groups would be an expected product of more equalized access to quality health care. However, similar results in the gradients appear for Mexico with a much lower rated health system. Additionally, studies from Costa Rica, which has a quality health system, showed reversals in the gradient for obesity and diet-related health conditions (Rosero-Bixby & Dow, 2009). Thus, better health care may not be the prime explanation for the observed weaker gradients. Underestimation of health conditions due to self-reported health could partially explain the reasons for weaker gradients in chronic conditions in the middle-income countries (Vellakkal et al., 2015). However, in some settings, underestimation of chronic conditions, such as diabetes, produces reliable estimates of disease (Banks et al., 2006; Goldman, Lin, Weinstein, & Lin, 2003) and estimates of prevalence based on self-reported diabetes and diabetes based on biomarkers are very similar in Colombia (Vargas-Uricoechea & Casas-Figueroa, 2015). Access to quality health care can explain underestimation of chronic conditions. However, urban areas, which are more prone to better resources and health care, make it more likely that individuals are aware of their health, which suggests that underreporting may be less problematic in our study. The reversal of the gradient for cancer in Colombia, England, and the United States points to the role of health education creating awareness of cancer risks in terms of early detection and diagnosis. Educational awareness provides a likely explanation for the reversal in developed world economies (Kinsey, Jemal, Liff, Ward, & Thun, 2008). In that sense, the large metropolitan city of Bogotá exhibits a pattern similar to high-income countries although the sharper gradient indicates higher inequality in health education. A third main point resolves around unique country histories. National context and differences in environmental factors matter in terms of interpreting cross national health patterns (Agyemang et al., 2010; Hebert et al., 1998). Explanations for the SES gradient are particularly complex in South Africa given its political and economic history (see Supplementary Material). South Africa is one of the most unequal societies in the world and income inequality is highest among South African blacks and colored (Bhorat et al., 2009; World Bank Group, 2017a). Our study showed a very high prevalence of obesity and reversals of the gradient for obesity, diabetes and trouble walking in South Africa similar to other studies (Alaba & Chola, 2014); obesity is a risk factor for diabetes and difficulty with functionality (Kuh & Ben-Shlomo, 2004). Costa Rica also showed a reversal in the SES gradient for obesity (Rosero-Bixby & Dow, 2009), but the likelihood of being obese among high SES groups is much greater among South African blacks/colored. The sharp SES disparity in obesity and higher prevalence of obesity in South Africa could thus indicate the added impact of difficult economic conditions experienced by the cohort of African black and colored populations born before the 1950s. As a result, even in a middle-income country, circumstances for a subgroup of the population may be similar to that of low-income countries, where the poorest experience food shortages and the richest have access to excess food and live a more sedentary lifestyle leading to the reversal of the gradient (Dinsa et al., 2012). The sharp gradient for consumption of fruits and vegetables in South Africa—double that of Colombia and Mexico—suggests the merit of this argument. The legacy of South Africa’s unique historical circumstances also provides explanations for the nonsignificance of the gradient in some aspects of health. Education may not be the best indicator of SES for older adult African blacks and colored born in the 20th century (Beinart & Dubow, 1995). Even though some cross-national studies suggest that education alone is a good indicator of SES for older adults (Goldman et al., 2011), multiple measures of SES may be more appropriate (Banks et al., 2006; Rosero-Bixby & Dow, 2009). Although our classification of education is very similar to that used by SAGE researchers (Vellakkal et al., 2015), and we follow other studies in classifying education into three categories (Banks et al., 2006; Rosero-Bixby & Dow, 2009), it may be that this grouping does not sufficiently distinguish inequality in South Africa. Admittedly, the sample size is smaller for urban South African blacks and colored but we observe similar patterns when we examine both urban and rural residents (results available upon request). Thus, sample size alone cannot explain the patterns observed. Explanations for the gradient in some aspects of health may not be unique to South Africa as other SAGE middle-income countries, for example, also show no education effects for depression (Anand & Phil, 2014). Although higher educational levels tend to engage in moderate or rigorous exercise (Malambo, Kengne, Lambert, De Villiers, & Puoane, 2016), it may that our measure of exercise (rigorous exercise) does not adequately distinguish between SES groups for older adults. A final main point is in regard to obesity. While sharp disparities in risk factors, such as smoking in South Africa, carry risks for health, it is not yet clear the degree to which the increasing prevalence of obesity will affect the SES gradient in middle-income countries. Obesity is an important risk factor for chronic conditions (Kuh & Ben-Shlomo, 2004) and other aspects of health such as cognition, depression, functionality. The large increase in obesity is a relatively recent phenomenon among older adults in middle-income countries. It may be too early to determine the impact of increasing obesity on the SES gradient for other aspects of health. The increasing prevalence of obesity in these populations ultimately may produce mixed patterns in the SES gradient in these other aspects of health. There are limitations to our study. First, cross-national comparisons require comparable measures of health and SES, which is not always possible in surveys of older adults. Having comparable measures for household wealth and health prevention, while enriching our study, was not possible given differences among surveys. There were no biomarkers for consumption of foods high in saturated fat in the middle-income settings. Second, cross-national comparisons are not precise due to country differences, which are often not measurable in surveys of older adults. Third, the meaning of educational attainment differs across countries signaling caution when making comparisons. Fourth, we used the legal definition of urban dwellers in the SAGE survey for Mexico and South Africa, making a direct comparison with older adults living in the capital city of Bogotá, Colombia not feasible. In spite of these limitations, rapid demographic and nutritional transitions, poor early life conditions, social mobility, an enriched nutritional environment later in life, negative health behavior, urbanization, and unique country circumstances provide a useful framework for understanding the SES gradient for health in middle-income countries. Given the importance of obesity as a risk factor, monitoring changes in the SES health gradient in related aspects of health in middle-income countries continues to be relevant. Supplementary Material Supplementary data is available at The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences online. Funding This work was supported by the Fulbright Commission in Colombia, the Colombian Department of Science, Technology, and Innovation (Colciencias), and the Aging Institute at the Pontificia Universidad Javeriana, Bogotá, within the framework of the Fulbright U.S. Scholar program, Cohort 2016–17. University of Wisconsin-Madison researchers are supported by National Institute of Health core grants to the Center for Demography and Ecology, University of Wisconsin (P2C HD047873); and the Center for Demography of Health and Aging, University of Wisconsin (P30 AG017266). Dr. Samper-Ternent was supported by the Sealy Center on Aging at the University of Texas Medical Branch in Galveston, Claude D. Pepper OAIC (P30-AG024832) and The Mexican Health and Aging Study (R01AG018016) from grants of the National Institutes of Health/National Institute on Aging. Conflict of Interest None reported. Acknowledgments The authors are particularly grateful to Robert M. Hauser, Alberto Palloni, and Emily Agree without whom this project would not have been possible. The authors thank the anonymous reviewers for their comments. Many thanks to Sarah Moen for editing the final version of the paper. Author contributions: M. McEniry provided background theories, conducted statistical analysis, and wrote the paper. R. Samper-Ternent, C. E. Flórez, R. Pardo contributed to significant discussions regarding Colombian data, measures, results, and revising the paper. C. Cano reviewed the paper and provided valuable comments. References Agyemang , C. , Kunst , A. , Bhopal , R. , Zaninotto , P. , Unwin , N. , Nazroo , J. ,… Stronks , K . ( 2010 ). A cross-national comparative study of blood pressure and hypertension between English and Dutch South-Asian– and African-origin populations: The role of national context . American Journal of Hypertension , 23 , 639 – 648 . doi: 10.1038/ajh.2010.39 Google Scholar CrossRef Search ADS PubMed Alaba , O. , & Chola , L . ( 2014 ). Socioeconomic inequalities in adult obesity prevalence in South Africa: A decomposition analysis . International Journal of Environmental Research and Public Health 11 , 3387 – 3406 . doi: 10.3390/ijerph110303387 Google Scholar CrossRef Search ADS PubMed Anand, A., & Phil, M. (2014). Understanding depression among older adults in six low-middle income countries using WHO-SAGE survey. Behavioral Health, 1, 1–11. Banks , J. , Marmot , M. , Oldfield , Z. , & Smith , J. P . ( 2006 ). Disease and disadvantage in the United States and in England . The Journal of the American Medical Association , 295 , 2037 – 2045 . doi: 10.1001/jama.295.17.2037 Google Scholar CrossRef Search ADS PubMed Barker , D. J. P . ( 1998 ). Mothers, babies and health in later life ( 2nd ed .). Edinburgh : Churchill Livingstone . Bateson , P. , & Gluckman , P . ( 2011 ). Plasticity, robustness, development and evolution . Cambridge : Cambridge University Press . Google Scholar CrossRef Search ADS Beinart , W. , & Dubow , S . (Eds.). ( 1995 ). Segregation and apartheid in twentieth-century South Africa . London : Routledge . Beltrán-Sánchez , H. , Palloni , A. , Riosmena , F. , & Wong , R . ( 2016 ). SES Gradients among Mexicans in the US and in Mexico: A new twist to the hispanic paradox ? Demography , 53 , 1555 – 1581 . doi: 10.1007/s13524-016-0508-4 Google Scholar CrossRef Search ADS PubMed Bhorat , H. , van der Westhuizen , C. , & Jacobs , T . ( 2009 ). Income and non-income inequality in post-apartheid South Africa: What are the drivers and possible policy interventions? Pretoria, South Africa : Trade & Industrial Policy Strategies . Development Policy Research Unit (DPRU) Working Paper No. 09/138. doi:10.2139/ssrn.1474271 Carrillo , R. , Barnabe-Ortiz , A. , Pillay , T. D. , Gilman , R. H. , Sanchez , J. F. , Poterico , J. ,… Miranda , J. J . ( 2016 ). Obesity risk in rural, urban and rural-to-urban migrants: Prospective results of the PERU MIGRANT study . International Journal of Obesity , 40 , 181 – 185 . doi: 10.1038/ijo.2015.140 Google Scholar CrossRef Search ADS PubMed CHDS . ( 2008 ). Closing the gap in a generation: health equity through action on the social determinants of health . Final report of the commission on social determinants of health . Geneva : World Health Organization . PubMed PubMed Crimmins , E. M. , & Finch , C. E . ( 2006 ). Infection, inflammation, height, and longevity . Proceedings of the National Academy of Sciences of the United States of America , 103 , 498 – 503 . doi: 10.1073/pnas.0501470103 Google Scholar CrossRef Search ADS PubMed Davey Smith , G. , & Lynch , J . ( 2004 ). Life course approaches to socioeconomic differentials in health . In G. D. Smith , & J. Lynch (Eds.), A life course approach to chronic disease epidemiology (pp. 77 – 115 ). Oxford : Oxford University Press . Dinsa , G. D. , Goryakin , Y. , Fumagalli , E. , & Suhrcke , M . ( 2012 ). Obesity and socioeconomic status in developing countries: A systematic review . Obesity Reviews , 13 , 1067 – 1079 . doi: 10.1111/j.1467-789X.2012.01017.x Google Scholar CrossRef Search ADS PubMed Elo , I. T. , & Preston , S. H . ( 1992 ). Effects of early-life conditions on adult mortality: a review . Population Index , 58 , 186 – 212 . doi:10.2307/3644718 Google Scholar CrossRef Search ADS PubMed FAO . ( 1946 ). World Food Survey . Washington, DC : United Nations . FAO . ( 2010 , 2017). Food balance sheets . Retrieved from http://faostat.fao.org/site/368/default.aspx#ancor Flórez, C. E., Guataquí, J. C., Mendez, R., & Cote, H. (2016). Las transformaciones sociodemográfics en Colombia Durante el Siglo XX y principios del Siglo XXI, Working paper. Bogotá, Colombia: Department of Economics, Universidad del Rosario. Ford , N. D. , Patel , S. A. , & Narayan , K. M. V . ( 2017 ). Obesity in low- and middle-income countries: Burden, drivers, and emerging challenges . Annual Review of Public Health , 38 , 145 – 164 . doi: 10.1146/annurev-publhealth-031816-044604 Google Scholar CrossRef Search ADS PubMed Franco , R. , Leon , A. , & Atria , R . ( 2007 ). Stratification and social mobility in Latin America: Structural transformations from a quarter of a century . Santiago, Chile : ECLAC and LOM Editions . Gluckman , P. , & Hanson , M . ( 2005 ). The fetal matrix: Evolution, development and disease . Cambridge : Cambridge University Press . Goldman , N. , Lin , I. F. , Weinstein , M. , & Lin , Y. H . ( 2003 ). Evaluating the quality of self-reports of hypertension and diabetes . The Journal of Clinical Epidemiology , 56 , 148 – 154 . doi: 10.1016/S0895-4356(02)00580-2 Google Scholar CrossRef Search ADS PubMed Goldman , N. , Turra , C. M. , Rosero-Bixby , L. , Weird , D. , & Crimmins , E . ( 2011 ). Do biological measures mediate the relationship between education and health: A comparative study . Social Science & Medicine , 72 , 307 – 315 . doi: 10.1016/j.socscimed.2010.11.004 Google Scholar CrossRef Search ADS He , W. , Goodkind , D. , & Kowal , P . ( 2016 ). An aging world: 2015 . U.S. Census Bureau, International Population Reports, P95/16-1. Washington, DC : U.S. Government Publishing Office . Hebert , J. R. , Hurley , T. G. , Olendzki , B. C. , Teas , J. , Ma , Y. , & Hampl , J. S . ( 1998 ). Nutritional and socioeconomic factors in relation to prostate cancer mortality: A cross-national study . Journal of the National Cancer Institute , 90 , 1637 – 1647 . doi: 10.1093/jnci/90.21.1637 Google Scholar CrossRef Search ADS PubMed Hosegood , V. , & Timaeus , I. M . ( 2006 ). HIV/AIDS and older people in South Africa . In B. Cohen , & J. Menken (Eds.), Aging in Sub-Saharan Africa: Recommendation for furthering research (pp. 250 – 275 ). Washington, DC : The National Academies Press . Kinsey , T. , Jemal , A. , Liff , J. , Ward , E. , & Thun , M . ( 2008 ). Secular trends in mortality from common cancers in the United States by educational attainment, 1993–2001 . Journal of the National Cancer Institute , 100 , 1003 – 1012 . doi: 10.1093/jnci/djn207 Google Scholar CrossRef Search ADS PubMed Kuh , D. , & Ben-Shlomo , Y . (Eds.). ( 2004 ). A life course approach to chronic disease epidemiology . New York : Oxford University Press . doi:10.1093/acprof:oso/9780198578154.001.0001 Google Scholar CrossRef Search ADS Lloyd-Sherlock , P. , Beard , J. , Minicuci , N. , Ebrahim , S. , & Chatterji , S . ( 2014 ). Hypertension among older adults in low- and middle-income countries: prevalence, awareness and control . The International Journal of Epidemiology , 43 , 116 – 128 . doi: 10.1093/ije/dyt215 Google Scholar CrossRef Search ADS PubMed López-Alonso , M . ( 2007 ). Growth with inequality: Living standards in Mexico, 1850–1950 . The Journal of Latin American Studies , 39 , 81 – 105 . doi: 10.1017/S0022216X06002045 Google Scholar CrossRef Search ADS Malambo , P. , Kengne , A. P. , Lambert , E. V. , De Villiers , A. , & Puoane , T . ( 2016 ). Prevalence and socio-demographic correlates of physical activity levels among South African adults in Cape Town and Mount Frere communities in 2008–2009 . Archives of Public Health , 74 , 1–9. doi: 10.1186/s13690-016-0167-3 Martorell, R., Stein, A. D., & Schroeder, D. G. (2001). Early nutrition and later adiposity. The Journal of Nutrition, 131, 874S–880S. doi:10.1093/jn/131.3.874S McEniry , M. C . ( 2014 ). Early life conditions and rapid demographic changes in the developing world: Consequences for older adult health . Dordrecht, Netherlands: Springer Science+Business Media. doi:10.1007/078-94-007-6979-3 Google Scholar CrossRef Search ADS McEniry, M. C. (2015). Research on Early Life and Aging Trends and Effects (RELATE): A Cross-National Study. Ann Arbor, MI: Inter-university Consortium for Political and Social Research (ICPSR34241-v2). doi:10.3886/ICPSR34241.v2 Monteiro , C. A. , Conde , W. L. , Lu , B. , & Popkin , B. M . ( 2004 ). Obesity and inequities in health in the developing world . International Journal of Obesity and Related Metabolic Disorders , 28 , 1181 – 1186 . doi: 10.1038/sj.ijo.0802716 Google Scholar CrossRef Search ADS PubMed Montez , J. K. , Hummer , R. A. , & Hayward , M. D . ( 2012 ). Educational attainment and adult mortality in the United States: A systematic analysis of functional form . Demography , 49 , 315 – 336 . doi: 10.1007/s13524-011-0082-8 Google Scholar CrossRef Search ADS PubMed Palloni , A. , & Souza , L . ( 2013 ). The fragility of the future and the tug of the past: Longevity in Latin America and the Caribbean . Demographic Research , 29 , 543 – 578 . doi: 10.4054/DemRes.2013.29.21 Google Scholar CrossRef Search ADS PubMed Pampel , F. C. , Denney , J. T. , & Krueger , P. M . ( 2012 ). Obesity, SES, and economic development: a test of the reversal hypothesis . Social Science & Medicine (1982) , 74 , 1073 – 1081 . doi: 10.1016/j.socscimed.2011.12.028 Google Scholar CrossRef Search ADS PubMed Payne , C. , Gómez-Olivé , F. X. , Kahn , K. , & Berkman , L . ( 2017 ). Physical function in an aging population in rural South Africa: Findings from HAALSI and cross-national comparisons with HRS sister studies . The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 72 , 665 – 679 . doi:10.1093/geronb/gbx030 Google Scholar CrossRef Search ADS Popkin , B. M . ( 2006 ). Global nutrition dynamics: The world is shifting rapidly toward a diet linked with noncommunicable diseases (NCDs) . The American Journal of Clinical Nutrition , 84 , 289 – 298 . doi:10.1093/ajcn/84.1.289 Google Scholar PubMed Popkin , B. M. , Horton , S. , & Kim , S . ( 2001 ). The nutritional transition and diet-related chronic diseases in Asia: Implications for prevention . Washington, DC : Food Consumption and Nutrition Division Discussion Paper International Food Policy Research Institute . Retrieved from http://www.ifpri.org/publication/nutritional-transition-and-diet-related-chronic-diseases-asia Radloff , L. S . ( 1977 ). The CES-D scale: A self-report depression scale for research in the general population . Applied Psychological Measurement , 1 , 385 – 401 . Google Scholar CrossRef Search ADS Riley , J. C . ( 2005 ). Bibliography of works providing estimates of life expectancy at birth and estimates of the beginning period of health transitions in countries with a population in 2000 of at least 400,000 . Retrieved from http://www.lifetable.de/RileyBib.htm de Romo , A. C. R. , & de Pérez , M. E. R . ( 1998 ). Historia de la salud pública en México: siglos XIX y XX [History of public health in Mexico: 19th and 20th century] . História, Ciências, Saúde-Manguinhos , 5 , 293 – 310 . doi: 10.1590/S0104-59701998000200002 Google Scholar CrossRef Search ADS PubMed Rosero-Bixby , L. , & Dow , W. H . ( 2009 ). Surprising SES gradients in mortality, health, and biomarkers in a Latin American population of adults . The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 64B , 105 – 117 . doi: 10.1093/geronb/gbn004 Google Scholar CrossRef Search ADS Schmidhuber , J. , & Shetty , P . ( 2005 ). The nutrition transition to 2030: Why developing countries are likely to bear the major burden . Paper presented at the 97th Seminar of the European Association of Agricultural Economists, University of Reading , England . Smith , K. V. , & Goldman , N . ( 2007 ). Socioeconomic differences in health among older adults in Mexico . Social Science & Medicine , 65 , 1372 – 1385 . doi: 10.1016/j.socscimed.2007.05.023 Google Scholar CrossRef Search ADS United Nations Development Program . ( 2016 ). Human development report 2016 . Washington, DC : United Nations . Unwin , N. , & Alberti , K. G . ( 2006 ). Chronic non-communicable diseases . Annals of Tropical Medicine and Parasitology , 100 , 455 – 464 . doi: 10.1179/136485906X97453 Google Scholar CrossRef Search ADS PubMed Vargas-Uricoechea , H. , & Casas-Figueroa , L. Á . ( 2015 ). An epidemiologic analysis of diabetes in Colombia . Annals of Global Health , 81 , 742 – 753 . doi: 10.1016/j.aogh.2015.11.001 Google Scholar CrossRef Search ADS PubMed Vellakkal , S. , Millett , C. , Basu , S. , Khan , Z. , Aitsi-Selmi , A. , Stuckler , D. , & Ebrahim , S . ( 2015 ). Are estimates of socioeconomic inequalities in chronic disease artefactually narrowed by self-reported measures of prevalence in low-income and middle-income countries? Findings from the WHO-SAGE survey . Journal of Epidemiology & Community Health , 69 , 218 – 225 . doi: 10.1136/jech-2014–204621 Google Scholar CrossRef Search ADS Wilkinson , R. G . ( 1996 ). Unhealthy societies: The afflictions of inequality . London : Routledge . Google Scholar CrossRef Search ADS World Bank Group . ( 2017a ). GINI index (World Bank estimate) . Retrieved from http://data.worldbank.org/indicator/SI.POV.GINI World Bank Group . ( 2017b ). GNI per capita, atlas method (current US$) . Retrieved from http://data.worldbank.org/indicator/NY.GNP.ATLS.CD World Bank Group . ( 2017c ). LAC equity lab: Income inequality - Urban/rural inequality . Retrieved from http://www.worldbank.org/en/topic/poverty/lac-equity-lab1/income-inequality/urban-rural-inequality World Health Organization . ( 2000 ). The world health report 2000: Health systems: Improving performance ( A. Haden , & B. Campanini Eds.). Geneva, Switzerland : World Health Organization . World Health Organization . ( 2015 ). World health statistics 2015 . Geneva, Switzerland : World Health Organization . Yan , S. , Li , J. , Li , S. , Zhang , B. , Du , S. , Gordon-Larsen , P. ,… Popkin , B. M . ( 2012 ). The expanding burden of cardiometabolic risk in China: The China health and nutrition survey . Obesity Reviews , 13 , 810 – 821 . doi: 10.1111/j.1467-789X.2012.01016.x Google Scholar CrossRef Search ADS PubMed Yesavage , J. A. , Brink , T. L. , Rose , T. L. , Lum , O. , Huang , V. , Adey , M. , & Leirer , V. O . ( 1982–1983 ). Development and validation of a geriatric depression screening scale: A preliminary report . Journal of Psychiatric Research , 17 , 37 – 49 . doi: 10.1016/0022-3956(82)90033–4 Google Scholar CrossRef Search ADS © The Author(s) 2018. Published by Oxford University Press on behalf of The Gerontological Society of America. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png The Journals of Gerontology Series B: Psychological Sciences and Social Sciences Oxford University Press

Patterns of SES Health Disparities Among Older Adults in Three Upper Middle- and Two High-Income Countries

Loading next page...
 
/lp/ou_press/patterns-of-ses-health-disparities-among-older-adults-in-three-upper-rKh80GvHKp
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of The Gerontological Society of America. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
ISSN
1079-5014
eISSN
1758-5368
D.O.I.
10.1093/geronb/gby050
Publisher site
See Article on Publisher Site

Abstract

Abstract Objectives To examine the socioeconomic status (SES) health gradient for obesity, diabetes, and hypertension within a diverse group of health outcomes and behaviors among older adults (60+) in upper middle-income countries benchmarked with high-income countries. Method We used data from three upper middle-income settings (Colombia-SABE-Bogotá, Mexico-SAGE, and South Africa-SAGE) and two high-income countries (England-ELSA and US-HRS) to estimate logistic regression models using age, gender, and education to predict health and health behaviors. Results The sharpest gradients appear in middle-income settings but follow expected patterns found in high-income countries for poor self-reported health, functionality, cognitive impairment, and depression. However, weaker gradients appear for obesity, hypertension, diabetes, and other chronic conditions in Colombia and Mexico and the gradient reverses in South Africa. Strong disparities exist in risky health behaviors and in early nutritional status in the middle-income settings. Discussion Rapid demographic and nutritional transitions, urbanization, poor early life conditions, social mobility, negative health behavior, and unique country circumstances provide a useful framework for understanding the SES health gradient in middle-income settings. In contrast with high-income countries, the increasing prevalence of obesity, an important risk factor for chronic conditions and other aspects of health, may ultimately change the SES gradient for diseases in the future. Diabetes, Health disparities, Hypertension, Middle-income countries, Obesity, Socioeconomic status Cross-national studies provide additional insight into the understanding of the socioeconomic status (SES) health gradient but there have been only a few cross-national studies of middle- and high-income countries that focus on the SES health gradient for older adults (Beltrán-Sánchez, Palloni, Riosmena, & Wong, 2016; Goldman, Turra, Rosero-Bixby, Weird, & Crimmins, 2011; Payne, Gómez-Olivé, Kahn, & Berkman, 2017). Although health is comprised of physical, mental, and social dimensions, few cross-national studies of middle- and high-income countries examine the SES gradient for older adults across a broad group of dimensions. The relationship between SES and health is complex (Davey Smith & Lynch, 2004), making international comparisons challenging. Most studies in the developed world show a strong positive relationship between poor health and lower SES status. However, one of the more intriguing questions facing social science researchers is the degree to which this relationship holds true in developing economies. Middle-income countries are particularly of interest given that they are experiencing large increases in diet-related conditions—obesity, diabetes, and heart disease—among a growing population of adults over the age of 60 (He, Goodkind, & Kowal, 2016; World Health Organization, 2000). Existing evidence from middle-income settings shows mixed results in regards to the SES health gradient, in particular for obesity (Dinsa, Goryakin, Fumagalli, & Suhrcke, 2012). One of the most comprehensive studies of a middle-income country to date showed reversals of the gradient for mortality, obesity, metabolic syndrome, and other diet-related health conditions in older adults; in contrast, other aspects of health followed an expected positive pattern (Rosero-Bixby & Dow, 2009). In other studies of older adults in middle-income countries, education, body mass index, and waist circumference were either not related (Goldman et al., 2011) or contrasting patterns appeared depending on rural/urban residence (Smith & Goldman, 2007). Demographic Regimes of the 20th Century and Nutritional Transitions in Middle-income Countries There are several reasons for cross-national differences observed in the SES gradient for health. Unequal societies tend to have a more negative impact on health (Wilkinson, 1996) and high levels of inequality mean that certain subgroups within a population remain at lower levels of development even though the country as a whole may prosper (López-Alonso, 2007). Limited social mobility and limited access to quality health care affect the magnitude of differences between SES groups (CHDS, 2008). Levels of country development may also affect health. Reversals in the gradient appear for obesity according to levels of development; at higher levels of development, there is a shift toward an expected positive gradient of low SES and obesity (Dinsa et al., 2012; Monteiro, Conde, Lu, & Popkin, 2004; Pampel, Denney, & Krueger, 2012). However, level of development alone does not completely explain the SES gradient given that, among middle-income countries, positive, negative and no associations appear between SES and health (Dinsa et al., 2012). The rapid demographic, epidemiological, and nutritional transitions from the 20th century to the present are relevant explanations for the mixed patterns found in the SES gradient in middle-income countries. Three aspects of these rapid transitions are notable. First, the demographic transitions of the 20th century characterized by mortality decline did not affect all countries equally in terms of the timing, pace, and reasons for the transitions. Even within countries, transitions have been uneven, particularly in rural areas where existing disparities between rural and urban are large and transitions have lagged behind urban areas. Rapid urbanization produced even greater differences between rural and urban areas (Flórez, Guataquí, Mendez, & Cote, 2016). The increase in obesity and chronic conditions among older adults is predominantly concentrated in urban areas (Carrillo et al., 2016; Ford, Patel, & Narayan, 2017; Unwin & Alberti, 2006) and suggests that the impact of rapid demographic and nutritional transitions in terms of health is most evident in urban areas. Weak associations between SES and health could thus reflect the stark contrast between rural and urban areas in their demographic and nutritional transitions. Urban areas (particularly large metropolitan areas) may be more similar to developed economies in SES health patterns whereas rural areas may produce a reversal of the gradient for obesity because lower SES groups still maintain a more traditional diet while higher SES do not (Smith & Goldman, 2007). On the other hand, higher inequality in urban areas in some developing economies (World Bank Group, 2017c) could produce sharp SES gradients when examining urban areas only. Second, older adults born in the mid-20th century in middle-income countries experienced rapid changes in nutrition later in life (FAO, 1946, 2010) with increased exposure to foods high in saturated fats amidst increased urbanization (Lloyd-Sherlock, Beard, Minicuci, Ebrahim, & Chatterji, 2014; Popkin, 2006; Schmidhuber & Shetty, 2005). The transition away from more traditional food to a diet with higher saturated fat content in developing economies increases the risk of poor health (Popkin, 2006; Popkin, Horton, & Kim, 2001) and could account for differences in the SES health gradient. Reversals of the gradient in middle-income countries could be a result of worse health behavior among higher SES groups. In more developed economies, higher SES groups tend to have access to healthier foods and obesity is more prevalent in lower SES groups (Dinsa et al., 2012). However, higher SES groups in some middle-income countries may be engaging in risky behavior by consuming foods higher in saturated fat content, suggesting a deteriorating health profile for those with higher SES who engage in risky health behaviors (Rosero-Bixby & Dow, 2009), even in settings with good health care coverage and quality health care. A third aspect of the transitions has yet to be thoroughly examined. That is, we know very little regarding the long-term consequences of the demographic regimes of the 20th century on adult health in developing economies. Mortality declined in the 20th century in many countries but the type of demographic regime most of interest is that characterized by rapid mortality decline without parallel improvement in standard of living. The dramatic decrease in mortality in the mid to late 20th century in developing economies, in particular among infants and children, produced cohorts characterized by increasing survivorship of poor early life conditions (nutrition and infectious diseases in particular). Improvements in early life health occurred largely in the context of stagnant improvements in standard of living resulting in continued exposure to poor environmental conditions during childhood. Poor early life conditions not only affected childhood health but also may affect the health of these unique cohorts at older ages (Palloni & Souza, 2013). There is now ample evidence of the long-term consequences of poor childhood conditions for older adult health (Barker, 1998; Bateson & Gluckman, 2011; Crimmins & Finch, 2006; Elo & Preston, 1992; Gluckman & Hanson, 2005). The historical circumstances of the 20th century provide a pertinent explanation for the SES health gradient among older adults in developing economies (McEniry, 2014; Palloni & Souza, 2013). Poor early life nutrition can affect cognitive and non-cognitive skills that lead to poor educational attainment (Martorell, Stein, & Schroeder, 2001). Cohorts increasingly characterized by survivorship of poor early life conditions could result in a larger group of individuals in low SES groups with lower educational attainment and less opportunity for social mobility producing sharper SES health disparities (Palloni & Souza, 2013). Alternatively, for these unique cohorts, mismatches between early life and later nutritional environments, in combination with increased social mobility among lower SES groups and negative health behavior among higher SES groups, could explain weak SES gradients. Predictive physiological and metabolic adaptations to an undernourished environment in early life lead to mismatches later in life if there is more favorable nutrition-enriched environments at older ages, resulting in disease (Bateson & Gluckman, 2011; Gluckman & Hanson, 2005). Improved economic status and social mobility for lower SES groups born into poor nutritional environments could lead to changes in food consumption resulting in a more enriched diet higher in saturated fat content. These circumstances could produce mismatches, increasing the risk of obesity, diabetes, and heart disease. Weak gradients appear because there are fewer differences between middle and higher SES groups in terms of diet, especially if higher SES groups also engage in negative health behavior and consume a diet rich in saturated fat. Taken as a whole, the complexity of historical circumstances in developing economies in transition and the importance of early life conditions may produce conflicting patterns of the SES gradient for health—reversals, sharp or weak gradients—in contrast to what occurred in developed economies with different historical circumstances. The demographic regimes of high-income countries in the 20th century are characterized by earlier and more gradual mortality decline (early 1900s or before) attributed to both improvements in standard of living and public health (McEniry, 2014). These regimes produced the pervasive increase in the older adult population and in chronic conditions and the established pattern of low SES being associated with poor health. In contrast, these increases are a more recent phenomenon in middle-income countries, more evident in urban areas, and more characterized by survivorship of poor early life conditions. Our Study Our study expands upon other studies of the SES health gradient in middle-income countries in two important ways. First, we compare the SES gradient and health across developed and developing world settings in the context of the demographic regimes of the 20th century. One type of regime (urban settings in upper-middle income countries) is characterized by rapid mortality decline without parallel improvement in standard of living and which has resulted in relatively recent increases in the aging population and chronic conditions. The other type of regime (high-income countries) experienced an earlier, more gradual mortality decline, where increases in the older population and chronic conditions have been pervasive for quite some time. Although other authors have made cross-national comparisons of older adult health in middle and high-income countries (Goldman et al., 2011; Payne et al., 2017), there have been few studies that use the framework of the demographic regimes of the 20th century to examine the health of older adults born during those regimes. Second, we compare gradients for obesity, diabetes, and hypertension relative to several other health outcomes along with adult lifestyle and early life nutritional status. With a notable exception (Rosero-Bixby & Dow, 2009), there have been few studies which have examined the SES gradient with such a diversity of outcomes in middle-income countries. We select three upper middle-income countries (Colombia, Mexico, and South Africa) and two high-income countries (United States and England) with the purpose of comparing, where possible, SES gradients from the study of the upper middle-income country of Costa Rica (Rosero-Bixby & Dow, 2009). The selected middle-income countries experienced similar rapid demographic transitions in the mid-20th century, but are also similar in terms of present-day country-level national income. The United States and England are settings with similar demographic transitions and levels of income but where health results and health systems are quite different. Rapid mortality decline occurred in Colombia beginning in the mid-1930s (Flórez et al., 2016) and in Mexico and South Africa in the 1940s, although, in the case of South Africa, political and economic events may have had slowed improvements for segments of the population until much later (Beinart & Dubow, 1995; Riley, 2005; de Romo & de Pérez, 1998). Poor nutritional environments were pervasive for older adults born in these settings during the 1930s–1940s although there was rapid improvement in nutrition in the decades after, in particular for Colombia and Mexico (FAO, 1946, 2010). Both Colombia and Mexico experienced some degree of increased social mobility toward the end of the 20th and beginning of the 21st centuries (directly benefiting either older adults or their children), increasing purchasing power for families, particularly in urban areas (Franco, Leon, & Atria, 2007). The selected middle-income settings show similarities in per capita gross national income (World Bank Group, 2017b) and human development (United Nations Development Program, 2016). Nevertheless, there are country differences in life expectancy, health systems, inequality, and history in the selected settings. Life expectancy at age 60 is lowest in South Africa (World Health Organization, 2015) and the quality of its health system is rated much lower than either Colombia or Mexico (World Health Organization, 2000). Apartheid dominated the political and economic system in South Africa for much of the mid to late 20th century; the HIV epidemic greatly impacted population health (Hosegood & Timaeus, 2006). Inequality is much higher in South Africa than in Colombia and Mexico; South Africa is one of the most unequal countries in the world; inequality in the United States and England is much lower than all three middle-income countries (World Bank Group, 2017a). Large discrepancies exist between rural and urban areas in the middle-income countries. However, although there is greater concentration of wealth and resources in urban settings, better economic opportunities, and better access to health care than rural areas, inequality is higher in urban areas than in rural areas in Colombia, Mexico, and South Africa (Bhorat, van der Westhuizen, & Jacobs, 2009; World Bank Group, 2017c). Our focus is on the direction and magnitude of the SES gradient for obesity, diabetes, and hypertension relative to other health outcomes and lifestyle. We examine reversals in the gradient and the degree to which the gradient is sharper or weaker than expected especially for obesity, diabetes, and hypertension. We also examine the degree to which there are similar patterns across countries with other health outcomes and behaviors. We expect that the gradient will be consistently strong and positive in high-income countries and mixed in middle-income settings. We expect to observe sharper inequalities in the middle-income countries, in particular in South Africa. Using the framework of rapid demographic, epidemiological, and nutritional transitions, we identify the most plausible explanations for cross-national differences and similarities in the SES health gradient for the middle-income settings. Methods Data Data on older Colombian adults come from SABE-Bogotá. SABE Bogotá is a representative cross-sectional sample of 2,000 older adults, 60 years and older within the metropolitan area of Bogotá from 2012 and is based on the original SABE study carried out in seven major cities of Latin America and the Caribbean. We select respondents 60 years and older living in urban areas in Mexico and South Africa using the WHO Study on Global Ageing and Adult Health Study (SAGE, 2007–2008) and in the United States and England using the Health and Retirement Study (HRS, 2000, 2006–2008) and the English Longitudinal Study on Ageing (ELSA, 2007–2008). For the United States, we obtained measured hypertension, diabetes, and performance measures first collected in 2006–2008. Race is an important distinguishing factor in South Africa; we focus on the two majority racial groups in the SAGE survey using their racial categories of South African blacks and colored. The final sample sizes were Bogotá (2,000), Mexico (1,286), South Africa (911), the United States (12,527), and England (6,183). The data form part of a NIA-funded project to harmonize data from major surveys of older adults or households (McEniry, 2015). The studies are representative of the older adult population either nationally or in major country provinces. All studies had good interviewer training, good questionnaire design, concern for data quality, and high response rates. All selected surveys collected measured blood pressure in addition to other dimensions of physical and mental health. Measures We use education for SES because education is a comparable measure of SES in older adults in cross-national studies (Goldman et al., 2011). However, the definition of educational categories can influence study results (Montez, Hummer, & Hayward, 2012). Because educational attainment differs between middle-income and high-income countries, we use a slightly different coding of education to reflect low, middle, and high educational groups. For the middle-income countries, we use three categories similar to educational categories used in other cross-national studies of middle-income countries (Goldman et al., 2011; Rosero-Bixby & Dow, 2009): incomplete primary, incomplete secondary, and complete secondary plus. For the United States and England, we use three categories defined by cross-national studies (Banks, Marmot, Oldfield, & Smith, 2006): low (0–11 years England, 0–12 years United States), middle (12–13 years England, 13–15 years United States), and high (more than 13 years England, more than 15 years United States). For health outcomes, we define obesity as a body mass index (BMI) of greater than or equal to 30. We define hypertension as systolic blood pressure rates greater than or equal to 140 mmHg, diastolic blood pressure rates greater than or equal to 90 mmHG, or taking medication to control hypertension. Diabetes is self-reported except for in the United States and England where we also include diabetes based on the biomarker HbA1c using a cutoff point of 6.5% or greater (of total hemoglobin), or taking diabetes medication (Yan et al., 2012). Although important in some settings, underestimation of diabetes may not be so problematic in some middle-income countries such as Colombia (Vargas-Uricoechea & Casas-Figueroa, 2015). Other chronic conditions are self-reported (heart disease, arthritis, respiratory disease, cancer, and stroke). We define poor self-reported health (SRH) based on a question asking respondents about their overall health. We use comparable and well-validated measures across countries for other dimensions of health, using established cutoff points where possible. Cognitive impairment is based on learning and immediate recall of words. For SABE-Bogotá, we defined impaired cognition as a short-term memory of 0–1 out of 3 words from a modified version of the Minimental State Exam. HRS, ELSA, and SAGE all had immediate recall of 10 words, with a cutoff point of less than 4 to define impaired cognition. Depression is defined in terms of major depressive symptoms. SABE-Bogotá uses the Geriatric Depression Scale (Yesavage et al., 1982–1983), which includes 15 items, and major depression is identified by scoring 10 points or higher on this scale. Depression in the SAGE surveys is based on the DSM IV criteria reviewed by psychiatrists. We define major depression as reporting at least one of two major symptom categories (mood and loss of interest), daily symptoms lasting at least 2 weeks, plus having at least four additional symptoms (insomnia, weight loss, psychomotor agitation, feelings of worthlessness, diminished ability to think, recurrent thoughts about death, or low energy). HRS and ELSA use the CES-D (Radloff, 1977) scale based on eight items. We define a score of 4 or greater as depressed. Difficulty with functionality is harmonized across countries and based on difficulty with at least one ADL (bathing, dressing, transferring, or toilet use). We define trouble walking (an indicator of frailty) in terms of the highest quintile of time to perform a normal walking test and weak grip strength in terms of a dynamometer reading of less than or equal to 20 kg. For adult lifestyle or risky health behavior (smoking, exercise, and diet), we define smoking as those who currently smoke, no rigorous exercise as those who do not perform consistent and weekly rigorous exercise, and poor diet as those who consume less than two servings of fruits and vegetables daily. We use the first quartile of height as a marker of poor early life nutritional status. Analyses For each country and health outcome and behavior, we estimate logistic regression models using age, gender, and educational groups with the reference group as the middle educational group. We control for race in South African models. To put our analyses of South African blacks and colored into better perspective, we estimate models for the entire population of South African older adults living in urban areas including the white and Asian populations and we estimate models for U.S. blacks. Comparing similar racial/ethnic groups may help better understand health disparities; national context is important and environmental factors differ across countries (Agyemang et al., 2010). Because there may be differences in cohorts in terms of diet, we estimate models with interactions between age and education for obesity, hypertension, and diabetes as a very broad dietary evaluation of cohorts. We summarize results with graphs showing patterns of the gradient across all countries along with joint SES significance levels. We estimate predicted probabilities and calculate the ratio of the predicted probabilities between low and high education groups as a measure of the magnitude of the gradient (Beltrán-Sánchez et al., 2016). To reduce the amount of cluttering in tables we show only the odds ratio with its significance level. The 95% confidence intervals for models are available upon request. Missing values were more notable in the middle-income countries (see Supplementary Notes for an explanation regarding missing values). Results Sample Characteristics and Health Profiles The selected samples are all similar in terms of a majority female composition (56%–61%) with the exception of Mexico where only 51% are female (Table 1). Age patterns are similar across all countries but there are noted differences in education. The large metropolitan area of Bogotá has a higher percent (26%) of respondents with at least a secondary educational level as compared with urban Mexican and South African respondents (10% and 11%). The United States and England have similar percent in the lowest educational group (63% and 60%). Table 1. Basic Sample Characteristics Colombia Bogotá Mexico Urban S Africa Urban England United States Female (%) 57 51 61 56 59 Age group (%)  60–64 35 24 34 23 24  65–69 25 22 31 22 22  70–74 17 17 14 18 20  75–79 12 19 9 16 17  80+ 12 18 12 20 17 Education (%)  Low 39 57 56 60 63  Middle 35 34 34 15 18  High 26 10 11 26 18 Colombia Bogotá Mexico Urban S Africa Urban England United States Female (%) 57 51 61 56 59 Age group (%)  60–64 35 24 34 23 24  65–69 25 22 31 22 22  70–74 17 17 14 18 20  75–79 12 19 9 16 17  80+ 12 18 12 20 17 Education (%)  Low 39 57 56 60 63  Middle 35 34 34 15 18  High 26 10 11 26 18 Source. SABE Bogotá; SAGE-Mexico, urban dwellers; SAGE-South Africa, urban dwellers, African blacks and colored; ELSA-England; HRS-US. Note: Weighted; numbers do not always sum to 100 due to rounding. View Large Table 1. Basic Sample Characteristics Colombia Bogotá Mexico Urban S Africa Urban England United States Female (%) 57 51 61 56 59 Age group (%)  60–64 35 24 34 23 24  65–69 25 22 31 22 22  70–74 17 17 14 18 20  75–79 12 19 9 16 17  80+ 12 18 12 20 17 Education (%)  Low 39 57 56 60 63  Middle 35 34 34 15 18  High 26 10 11 26 18 Colombia Bogotá Mexico Urban S Africa Urban England United States Female (%) 57 51 61 56 59 Age group (%)  60–64 35 24 34 23 24  65–69 25 22 31 22 22  70–74 17 17 14 18 20  75–79 12 19 9 16 17  80+ 12 18 12 20 17 Education (%)  Low 39 57 56 60 63  Middle 35 34 34 15 18  High 26 10 11 26 18 Source. SABE Bogotá; SAGE-Mexico, urban dwellers; SAGE-South Africa, urban dwellers, African blacks and colored; ELSA-England; HRS-US. Note: Weighted; numbers do not always sum to 100 due to rounding. View Large Large differences exist in health outcomes and risk factors across countries (Table 2). We observe particularly higher percentages with obesity, hypertension, and trouble walking in South Africa (48%, 86%, and 29%), heart disease in the United States and England (25% and 29%), and arthritis in the United States (61%). For overall health, we find high percentages in poor SRH in Bogotá (48%), cognitive impairment in Mexico (40%), and weak grip strength in Bogotá and urban Mexico (48% and 46%). For risky behavior, we find a high percent of older adults do no regular strenuous exercise in the middle-income countries (88%–90%) or in England (86%), and eat less than two servings of fruit or vegetables in Bogotá and urban Mexico (33% and 35%). On average, female respondents in middle-income country settings are between 3–9 (7–13) cm shorter than their English (United States) counterparts. Likewise, male respondents are 9 (14) cm shorter than their English (United States) counterparts. Table 2. Overall Prevalence of Health/Risk Factors Colombia Bogotá Mexico Urban S Africa Urban England United States Overall health (%) Poor self-reported health 48 14 17 32 26 Difficulty with functionality 6 25 25 25 16 Trouble walking 21 19 29 21 19 Weak grip strength 48 46 27 29 28 Cognitive impairment 10 40 18 15 16 Depression 7 8 1 16 15 Chronic conditions (%) Hypertension (measured) 76 73 86 54 73 Diabetes 18 21 12 9 22 Heart disease 5 10 10 29 25 Stroke 4 6 5 7 7 Respiratory disease 15 7 4 8 10 Arthritis 28 13 30 42 61 Cancer 6 9 14 Adult risk factors (%) Obesity 26 30 48 28 21 Current smoker 8 21 23 13 12 No vigorous exercise 90 89 88 86 57 Less than 2 fr/veg daily 33 35 21 Early life nutritional status Height (cm)  Female 150 (6) 149 (7) 155 (11) 158 (7) 162 (7)  Male 163 (7) 163 (7) 163 (10) 172 (7) 177 (7) Colombia Bogotá Mexico Urban S Africa Urban England United States Overall health (%) Poor self-reported health 48 14 17 32 26 Difficulty with functionality 6 25 25 25 16 Trouble walking 21 19 29 21 19 Weak grip strength 48 46 27 29 28 Cognitive impairment 10 40 18 15 16 Depression 7 8 1 16 15 Chronic conditions (%) Hypertension (measured) 76 73 86 54 73 Diabetes 18 21 12 9 22 Heart disease 5 10 10 29 25 Stroke 4 6 5 7 7 Respiratory disease 15 7 4 8 10 Arthritis 28 13 30 42 61 Cancer 6 9 14 Adult risk factors (%) Obesity 26 30 48 28 21 Current smoker 8 21 23 13 12 No vigorous exercise 90 89 88 86 57 Less than 2 fr/veg daily 33 35 21 Early life nutritional status Height (cm)  Female 150 (6) 149 (7) 155 (11) 158 (7) 162 (7)  Male 163 (7) 163 (7) 163 (10) 172 (7) 177 (7) Source. SABE Bogotá; SAGE-Mexico, urban dwellers; SAGE-South Africa, urban dwellers, African blacks and colored; ELSA-England; HRS-US. Note: Early life nutritional status shows averages with standard deviations. Weighted. Prevalence for diabetes in England and United States based on biomarkers. View Large Table 2. Overall Prevalence of Health/Risk Factors Colombia Bogotá Mexico Urban S Africa Urban England United States Overall health (%) Poor self-reported health 48 14 17 32 26 Difficulty with functionality 6 25 25 25 16 Trouble walking 21 19 29 21 19 Weak grip strength 48 46 27 29 28 Cognitive impairment 10 40 18 15 16 Depression 7 8 1 16 15 Chronic conditions (%) Hypertension (measured) 76 73 86 54 73 Diabetes 18 21 12 9 22 Heart disease 5 10 10 29 25 Stroke 4 6 5 7 7 Respiratory disease 15 7 4 8 10 Arthritis 28 13 30 42 61 Cancer 6 9 14 Adult risk factors (%) Obesity 26 30 48 28 21 Current smoker 8 21 23 13 12 No vigorous exercise 90 89 88 86 57 Less than 2 fr/veg daily 33 35 21 Early life nutritional status Height (cm)  Female 150 (6) 149 (7) 155 (11) 158 (7) 162 (7)  Male 163 (7) 163 (7) 163 (10) 172 (7) 177 (7) Colombia Bogotá Mexico Urban S Africa Urban England United States Overall health (%) Poor self-reported health 48 14 17 32 26 Difficulty with functionality 6 25 25 25 16 Trouble walking 21 19 29 21 19 Weak grip strength 48 46 27 29 28 Cognitive impairment 10 40 18 15 16 Depression 7 8 1 16 15 Chronic conditions (%) Hypertension (measured) 76 73 86 54 73 Diabetes 18 21 12 9 22 Heart disease 5 10 10 29 25 Stroke 4 6 5 7 7 Respiratory disease 15 7 4 8 10 Arthritis 28 13 30 42 61 Cancer 6 9 14 Adult risk factors (%) Obesity 26 30 48 28 21 Current smoker 8 21 23 13 12 No vigorous exercise 90 89 88 86 57 Less than 2 fr/veg daily 33 35 21 Early life nutritional status Height (cm)  Female 150 (6) 149 (7) 155 (11) 158 (7) 162 (7)  Male 163 (7) 163 (7) 163 (10) 172 (7) 177 (7) Source. SABE Bogotá; SAGE-Mexico, urban dwellers; SAGE-South Africa, urban dwellers, African blacks and colored; ELSA-England; HRS-US. Note: Early life nutritional status shows averages with standard deviations. Weighted. Prevalence for diabetes in England and United States based on biomarkers. View Large Health Outcomes The effects of education on several health outcomes show expected consistent and significant patterns of lowest educational group being most at risk of being in poor health in the United States and England (Table 3). In Colombia, strong gradients appear in all variables except trouble walking and grip strength, and in Mexico, strong gradients appear with the exception of grip strength. The gradients in South Africa are significant for poor SRH, functionality, and trouble walking which shows a reversal in the gradient. The total urban South African population of older adults shows similar patterns with the blacks and colored; U.S. blacks tend to display sharper disparities than South African blacks/colored (Supplementary Table S1). Table 3. Effects of Education on Selected Health Outcomes Colombia Bogotá Mexico Urban S Africa Urban England United States Poor SRH L 1.46*** 1.87** 1.55* 1.96*** 2.14*** H 0.41*** 0.59 0.58 0.74** 0.68*** Difficulty with Functionality L 1.68** 1.88*** 0.71 1.47*** 1.50*** H 0.77 1.08 0.23*** 0.91 0.99 Trouble Walking L 1.22 1.85** 0.36*** 2.05*** 2.06*** H 0.79 0.62 1.43 0.83 1.03 Weak Grip Strength L 1.29 0.94 1.01 1.53*** 1.31** H 1.06 1.52 0.46 0.85 1.03 Cognitive Impairment L 1.86* 1.87*** 1.33 2.48*** 2.17*** H 0.77 0.78 0.46 0.90 0.67** Depression L 1.51* 2.04** 1.10 1.79*** 1.91*** H 0.38** 0.79 1.13 0.98 0.76** Colombia Bogotá Mexico Urban S Africa Urban England United States Poor SRH L 1.46*** 1.87** 1.55* 1.96*** 2.14*** H 0.41*** 0.59 0.58 0.74** 0.68*** Difficulty with Functionality L 1.68** 1.88*** 0.71 1.47*** 1.50*** H 0.77 1.08 0.23*** 0.91 0.99 Trouble Walking L 1.22 1.85** 0.36*** 2.05*** 2.06*** H 0.79 0.62 1.43 0.83 1.03 Weak Grip Strength L 1.29 0.94 1.01 1.53*** 1.31** H 1.06 1.52 0.46 0.85 1.03 Cognitive Impairment L 1.86* 1.87*** 1.33 2.48*** 2.17*** H 0.77 0.78 0.46 0.90 0.67** Depression L 1.51* 2.04** 1.10 1.79*** 1.91*** H 0.38** 0.79 1.13 0.98 0.76** Source. SABE Bogotá; SAGE-Mexico, urban dwellers; SAGE-South Africa, urban dwellers, African blacks and colored; ELSA-England; HRS-US. Note: Shown are odds ratios by education group (low/high). Reference group is middle educational group. Models significantly differently from null model at p < .001 for England and the United States; models also significant in other countries with some exceptions (see note in Supplementary Material). SRH = Self-reported health. *p < .05; **p < .01; ***p < .001. View Large Table 3. Effects of Education on Selected Health Outcomes Colombia Bogotá Mexico Urban S Africa Urban England United States Poor SRH L 1.46*** 1.87** 1.55* 1.96*** 2.14*** H 0.41*** 0.59 0.58 0.74** 0.68*** Difficulty with Functionality L 1.68** 1.88*** 0.71 1.47*** 1.50*** H 0.77 1.08 0.23*** 0.91 0.99 Trouble Walking L 1.22 1.85** 0.36*** 2.05*** 2.06*** H 0.79 0.62 1.43 0.83 1.03 Weak Grip Strength L 1.29 0.94 1.01 1.53*** 1.31** H 1.06 1.52 0.46 0.85 1.03 Cognitive Impairment L 1.86* 1.87*** 1.33 2.48*** 2.17*** H 0.77 0.78 0.46 0.90 0.67** Depression L 1.51* 2.04** 1.10 1.79*** 1.91*** H 0.38** 0.79 1.13 0.98 0.76** Colombia Bogotá Mexico Urban S Africa Urban England United States Poor SRH L 1.46*** 1.87** 1.55* 1.96*** 2.14*** H 0.41*** 0.59 0.58 0.74** 0.68*** Difficulty with Functionality L 1.68** 1.88*** 0.71 1.47*** 1.50*** H 0.77 1.08 0.23*** 0.91 0.99 Trouble Walking L 1.22 1.85** 0.36*** 2.05*** 2.06*** H 0.79 0.62 1.43 0.83 1.03 Weak Grip Strength L 1.29 0.94 1.01 1.53*** 1.31** H 1.06 1.52 0.46 0.85 1.03 Cognitive Impairment L 1.86* 1.87*** 1.33 2.48*** 2.17*** H 0.77 0.78 0.46 0.90 0.67** Depression L 1.51* 2.04** 1.10 1.79*** 1.91*** H 0.38** 0.79 1.13 0.98 0.76** Source. SABE Bogotá; SAGE-Mexico, urban dwellers; SAGE-South Africa, urban dwellers, African blacks and colored; ELSA-England; HRS-US. Note: Shown are odds ratios by education group (low/high). Reference group is middle educational group. Models significantly differently from null model at p < .001 for England and the United States; models also significant in other countries with some exceptions (see note in Supplementary Material). SRH = Self-reported health. *p < .05; **p < .01; ***p < .001. View Large Graphical depictions of the gradients by ordering the regression odds ratios summarize these initial general impressions for Colombia, Mexico, South Africa, United States (Figure 1), and England, which is similar to the United States (Supplementary Figure 1). Although the order of the odds ratios differs across countries, the general patterns for these aspects of health are similar in Colombia, Mexico, the United States, and England especially for cognitive impairment, depression, functionality, and poor SRH. South Africa shows fewer significant associations between education and health/adult lifestyle and more variation in the SES patterns. Figure 1. View largeDownload slide Cross national patterns of SES health disparities. SES = Socioeconomic status; see Supplementary Figure 2. Figure 1. View largeDownload slide Cross national patterns of SES health disparities. SES = Socioeconomic status; see Supplementary Figure 2. Predicted probabilities show the sharpest gradients in the middle-income countries in functionality, cognitive impairment, and depression (Colombia), and SRH, trouble walking (Mexico) and weak grip strength (South Africa) (Figure 2, Supplementary Table S3). The probability of poor health for the lowest SES group is between two to nearly four times that of the probability of poor health for those in the highest SES group for these aspects of health in the middle-income countries. Figure 2. View largeDownload slide Ratio of predicted probabilities for low/high SES. SES = Socioeconomic status; see Supplementary Figure 3. Figure 2. View largeDownload slide Ratio of predicted probabilities for low/high SES. SES = Socioeconomic status; see Supplementary Figure 3. Chronic Conditions and Obesity The middle-income settings show a weaker pattern for the effects of obesity, diabetes, and hypertension with the exception of obesity and diabetes in South Africa where a reversal of the gradient appears showing a strong likelihood of being obese among high SES and protective effects of being obese and diabetic for low SES (Table 4). Models for obesity with all urban South African older adults show similar results for low SES but the significant effect in high SES disappears; U.S. blacks tend to display sharper disparities than South African blacks, although the gradients for obesity and diabetes are flat for U.S. blacks (Supplementary Table S1). High SES shows protective effects on diabetes (Colombia) and hypertension and obesity (Mexico). In contrast, strong gradients in England and the United States appear in the expected direction. Predicted probabilities (Figure 2, Supplementary Table S3) show weaker gradients for hypertension, diabetes, and obesity across all countries, although the gradient is sharpest for diabetes in Colombia and Mexico and sharpest for obesity in the United States and England. The predicted probability of diabetes for the lowest SES group is almost 1.5 times the predicted probability of diabetes for the highest SES group in Colombia and Mexico; similarly for obesity in the United States and England (Figure 2, Supplementary Table S3). Models with interactions for age and education produced no consistent significant interactions across countries and are not shown. Table 4. Effects of Education on Hypertension, Diabetes, Obesity, Adult Risk Factors, Childhood Conditions Colombia Bogotá Mexico Urban S Africa Urban England United States Hypertension L 1.09 0.96 1.14 1.15 1.36*** H 0.86 0.57* 1.59 0.82* 0.94 Diabetes L 1.03 1.05 0.64* 1.40* 1.38*** H 0.69* 0.64 1.21 0.96 0.93 Obesity L 0.92 0.77 0.48*** 1.54*** 1.38*** H 0.85 0.55* 2.72*** 0.93 0.82* Smoker L 1.10 1.10 1.86*** 1.89*** 1.33*** H 0.79 0.82 0.45 0.73* 0.57*** No Vigorous Exercise L 1.78* 0.64* 0.78 1.61*** 1.36*** H 0.65* 0.43** 0.92 0.77* 0.95 LT 2 Fr/Veg Daily L 1.64*** 1.42** 2.29*** H 0.49*** 0.81 0.35* Low Height L 1.80*** 1.79*** 1.16 1.70*** 1.49*** H 0.77 0.78 0.92 0.83 0.85 Colombia Bogotá Mexico Urban S Africa Urban England United States Hypertension L 1.09 0.96 1.14 1.15 1.36*** H 0.86 0.57* 1.59 0.82* 0.94 Diabetes L 1.03 1.05 0.64* 1.40* 1.38*** H 0.69* 0.64 1.21 0.96 0.93 Obesity L 0.92 0.77 0.48*** 1.54*** 1.38*** H 0.85 0.55* 2.72*** 0.93 0.82* Smoker L 1.10 1.10 1.86*** 1.89*** 1.33*** H 0.79 0.82 0.45 0.73* 0.57*** No Vigorous Exercise L 1.78* 0.64* 0.78 1.61*** 1.36*** H 0.65* 0.43** 0.92 0.77* 0.95 LT 2 Fr/Veg Daily L 1.64*** 1.42** 2.29*** H 0.49*** 0.81 0.35* Low Height L 1.80*** 1.79*** 1.16 1.70*** 1.49*** H 0.77 0.78 0.92 0.83 0.85 Source. SABE Bogotá; SAGE-Mexico, urban dwellers; SAGE-South Africa, urban dwellers, African blacks and colored; ELSA-England; HRS-US. Note: Shown are odds ratios by education group (low/high). Reference group is middle educational group. Models significantly differently from null model at p < .001 for England and the United States; models also significant in other countries with some exceptions (see note in Supplementary Material). Models for diabetes for England and the United States use diabetes based on biomarkers; models using self-reported diabetes produced similar patterns and are not shown in table. *p < .05; **p < .01; ***p < .001. View Large Table 4. Effects of Education on Hypertension, Diabetes, Obesity, Adult Risk Factors, Childhood Conditions Colombia Bogotá Mexico Urban S Africa Urban England United States Hypertension L 1.09 0.96 1.14 1.15 1.36*** H 0.86 0.57* 1.59 0.82* 0.94 Diabetes L 1.03 1.05 0.64* 1.40* 1.38*** H 0.69* 0.64 1.21 0.96 0.93 Obesity L 0.92 0.77 0.48*** 1.54*** 1.38*** H 0.85 0.55* 2.72*** 0.93 0.82* Smoker L 1.10 1.10 1.86*** 1.89*** 1.33*** H 0.79 0.82 0.45 0.73* 0.57*** No Vigorous Exercise L 1.78* 0.64* 0.78 1.61*** 1.36*** H 0.65* 0.43** 0.92 0.77* 0.95 LT 2 Fr/Veg Daily L 1.64*** 1.42** 2.29*** H 0.49*** 0.81 0.35* Low Height L 1.80*** 1.79*** 1.16 1.70*** 1.49*** H 0.77 0.78 0.92 0.83 0.85 Colombia Bogotá Mexico Urban S Africa Urban England United States Hypertension L 1.09 0.96 1.14 1.15 1.36*** H 0.86 0.57* 1.59 0.82* 0.94 Diabetes L 1.03 1.05 0.64* 1.40* 1.38*** H 0.69* 0.64 1.21 0.96 0.93 Obesity L 0.92 0.77 0.48*** 1.54*** 1.38*** H 0.85 0.55* 2.72*** 0.93 0.82* Smoker L 1.10 1.10 1.86*** 1.89*** 1.33*** H 0.79 0.82 0.45 0.73* 0.57*** No Vigorous Exercise L 1.78* 0.64* 0.78 1.61*** 1.36*** H 0.65* 0.43** 0.92 0.77* 0.95 LT 2 Fr/Veg Daily L 1.64*** 1.42** 2.29*** H 0.49*** 0.81 0.35* Low Height L 1.80*** 1.79*** 1.16 1.70*** 1.49*** H 0.77 0.78 0.92 0.83 0.85 Source. SABE Bogotá; SAGE-Mexico, urban dwellers; SAGE-South Africa, urban dwellers, African blacks and colored; ELSA-England; HRS-US. Note: Shown are odds ratios by education group (low/high). Reference group is middle educational group. Models significantly differently from null model at p < .001 for England and the United States; models also significant in other countries with some exceptions (see note in Supplementary Material). Models for diabetes for England and the United States use diabetes based on biomarkers; models using self-reported diabetes produced similar patterns and are not shown in table. *p < .05; **p < .01; ***p < .001. View Large No significant gradients appear in other chronic conditions (heart disease, stroke, arthritis, or respiratory disease) in Colombia, Mexico, or South Africa with the exception of respiratory disease in Colombia. There are strong gradients for these chronic conditions in the United States and England and a reversal of the gradient appears for cancer in Colombia, England, and the United States (Supplementary Table S2). South African urban residents and blacks/colored show similar patterns; the gradient for U.S. blacks is mostly flat with the exception of arthritis (Supplementary Table S2). Predicted probabilities (Supplementary Table S3) show the sharpest gradients appear mostly in the middle-income countries with the exception of arthritis and respiratory disease. Colombia shows the sharpest gradient for stroke and cancer and South Africa the sharpest gradient for heart disease. The predicted probability of stroke for the lowest SES group is nearly three times that of the highest SES group in Colombia. The predicted probability of heart disease for the lowest SES group in South Africa is nearly two times that of the highest SES group. Being in the lowest SES group reduces the probability of reporting cancer by about 50% as compared to being in the highest SES group in Colombia. The predicted probability of respiratory disease (arthritis) for the lowest SES group in England is about two (1.3) times more than that of the highest SES group. Risky Adult Behavior and Early Life Nutritional Status We find strong gradients for smoking in South Africa, England, and the United States (Table 4). Strong gradients also appear for exercise in Colombia, England, and the United States but a reversal in low SES in Mexico and no significant gradient in South Africa. Strong gradients appear for consumption of fruits and vegetables across all middle-income settings. With the exception of South Africa, strong gradients appear for low height across all countries. South African urban residents follow similar patterns as South African urban blacks and colored; U.S. blacks show a sharp gradient for exercise but flatter gradients for smoking and height (Supplementary Table S1). Predicted probabilities show a weak gradient for exercise across all countries with the exception of the United States. The sharpest gradients for smoking and consumption of fruits and vegetables appear in South Africa and for low height in Colombia (Supplementary Table S3). The predicted probability of being a current smoker (poor diet) for the lowest SES group is more than three (almost 5) times that of the highest SES group in South Africa. The predicted probability of short height for the lowest SES group is over two times that of the highest SES group in Colombia. Discussion We examined the SES health gradient using education for an extensive group of health outcomes and risky adult behavior in three middle-income urban settings, benchmarking with England and the United States within the framework of the demographic and nutritional transitions of the 20th century. High-income countries consistently showed strong gradients for health in the expected direction (low SES predicts poor health) whereas mixed patterns appeared in middle-income countries. The sharpest gradients appeared mostly in middle-income countries but followed expected patterns found in the United States and England especially in health outcomes such as poor SRH, functionality, trouble walking, cognitive impairment, and depression; however, weaker gradients appeared for obesity, hypertension, diabetes, and other chronic conditions in Colombia and Mexico with reversals observed in South Africa. Reversal of the gradient appeared for cancer in Bogotá, England, and the United States. Strong disparities existed in risky health behavior related to diet and early life nutritional status in the middle-income settings. The results contribute to expanding knowledge regarding cross-national comparisons of the SES health gradient in middle- and high-income settings. There have been few cross-national studies that examine the SES gradient across such a diverse group of health outcomes and risky health behaviors. The selection of urban populations in Colombia, Mexico, and South Africa is relevant given the increased prevalence of obesity and chronic conditions among older adults in the context of rapid demographic and nutritional transitions, increased urbanization, and high inequality. We do not claim that our selection of middle-income countries is representative of other upper middle-income countries. However, the selected countries are comparable in terms their demographic, epidemiological, and nutritional transitions and level of national income. The comparison with United States and England provides the perspective of well-established patterns from developed countries, which experienced much earlier growth of the older adult population and associated chronic conditions. We highlight four main points in regards to the SES gradient and health in middle-income countries. First, high level of inequality due to rapid urbanization is a relevant explanation for the sharpness of the gradient found in some health outcomes. Consistent with other studies in middle-income countries (Rosero-Bixby & Dow, 2009) and high-income countries (Banks et al., 2006), our study showed similar patterns in the expected direction for health such as poor SRH, cognition, depression, functionality, and trouble walking. However, the sharpest gradients appear in urban settings in middle-income countries. Sharper gradients are not surprising given that there are high levels of inequality found in urban settings in middle-income countries (World Bank Group, 2017a). Rapid urbanization accompanied the rapid demographic and nutritional transitions affecting middle-income countries (Flórez et al., 2016). Second, the combination of rapid transitions, poor early life nutritional environment, social mobility, and an enriched nutritional environment later in life provides a pertinent explanation for the weaker gradients for obesity, diabetes, and hypertension in Colombia and Mexico. The sharp disparities in poor early life nutritional status and current diet suggest a history of poor nutrition among low SES groups. The argument for the switch in the gradient in health conditions such as obesity according to country-level gross national income (Monteiro et al., 2004) does not adequately explain the patterns encountered in obesity in the middle-income countries in our study. The weaker gradient for obesity (particularly in Colombia and Mexico) and the differences noted in the gradient (particularly in South Africa) suggest multiple and opposing factors. The historical circumstances of cohorts born in the 20th century and characterized by increasing survivorship of poor early life conditions (Palloni & Souza, 2013) combined with uneven development affecting different SES groups and changes in food consumption is a plausible explanation for the weaker gradients. Both Colombia and Mexico experienced increased social mobility in the late 20th and early 21st century (Franco et al., 2007), potentially resulting in changes in food consumption characterized by higher consumption of saturated fats and processed foods and producing higher risk of obesity, diabetes, and hypertension. Older adults from cohorts of the 20th century with increased social mobility may be experiencing a mismatch in nutritional environment later in life (Bateson & Gluckman, 2011; Gluckman & Hanson, 2005) leading to higher risk of disease. In contrast, those in the lowest SES may not yet be experiencing large nutritional changes and not everyone in the highest SES may be better consumers of healthy food and value healthy behavior. Thus, even though the determinants of obesity, diabetes, and hypertension are different between groups (higher SES due to adult lifestyle and middle SES due to early life plus adult lifestyle), weaker gradients appear between middle and high SES groups. The weaker gradients could partially reflect a weaker association among educational attainment in the selected middle-income settings. However, improved access to education among lower SES groups due to public policy is probably not a major confounding factor in explaining the gradient in settings such as Colombia and Mexico because strong gradients appear with other health outcomes such as cognitive impairment, depression, functionality, and poor SRH. On the other hand, weaker gradients could reflect the long-term negative consequences of economic inequality unique to segments of the South African population. Educational attainment for some segments of the South African population may not have resulted in increased wealth. The coverage and quality of the health care system in countries such as Colombia, noted to have a good health system (World Health Organization, 2000), could provide an explanation for the weaker gradients. Better health observed in several health outcomes among Colombians due to a good health care system and better health across all SES groups would be an expected product of more equalized access to quality health care. However, similar results in the gradients appear for Mexico with a much lower rated health system. Additionally, studies from Costa Rica, which has a quality health system, showed reversals in the gradient for obesity and diet-related health conditions (Rosero-Bixby & Dow, 2009). Thus, better health care may not be the prime explanation for the observed weaker gradients. Underestimation of health conditions due to self-reported health could partially explain the reasons for weaker gradients in chronic conditions in the middle-income countries (Vellakkal et al., 2015). However, in some settings, underestimation of chronic conditions, such as diabetes, produces reliable estimates of disease (Banks et al., 2006; Goldman, Lin, Weinstein, & Lin, 2003) and estimates of prevalence based on self-reported diabetes and diabetes based on biomarkers are very similar in Colombia (Vargas-Uricoechea & Casas-Figueroa, 2015). Access to quality health care can explain underestimation of chronic conditions. However, urban areas, which are more prone to better resources and health care, make it more likely that individuals are aware of their health, which suggests that underreporting may be less problematic in our study. The reversal of the gradient for cancer in Colombia, England, and the United States points to the role of health education creating awareness of cancer risks in terms of early detection and diagnosis. Educational awareness provides a likely explanation for the reversal in developed world economies (Kinsey, Jemal, Liff, Ward, & Thun, 2008). In that sense, the large metropolitan city of Bogotá exhibits a pattern similar to high-income countries although the sharper gradient indicates higher inequality in health education. A third main point resolves around unique country histories. National context and differences in environmental factors matter in terms of interpreting cross national health patterns (Agyemang et al., 2010; Hebert et al., 1998). Explanations for the SES gradient are particularly complex in South Africa given its political and economic history (see Supplementary Material). South Africa is one of the most unequal societies in the world and income inequality is highest among South African blacks and colored (Bhorat et al., 2009; World Bank Group, 2017a). Our study showed a very high prevalence of obesity and reversals of the gradient for obesity, diabetes and trouble walking in South Africa similar to other studies (Alaba & Chola, 2014); obesity is a risk factor for diabetes and difficulty with functionality (Kuh & Ben-Shlomo, 2004). Costa Rica also showed a reversal in the SES gradient for obesity (Rosero-Bixby & Dow, 2009), but the likelihood of being obese among high SES groups is much greater among South African blacks/colored. The sharp SES disparity in obesity and higher prevalence of obesity in South Africa could thus indicate the added impact of difficult economic conditions experienced by the cohort of African black and colored populations born before the 1950s. As a result, even in a middle-income country, circumstances for a subgroup of the population may be similar to that of low-income countries, where the poorest experience food shortages and the richest have access to excess food and live a more sedentary lifestyle leading to the reversal of the gradient (Dinsa et al., 2012). The sharp gradient for consumption of fruits and vegetables in South Africa—double that of Colombia and Mexico—suggests the merit of this argument. The legacy of South Africa’s unique historical circumstances also provides explanations for the nonsignificance of the gradient in some aspects of health. Education may not be the best indicator of SES for older adult African blacks and colored born in the 20th century (Beinart & Dubow, 1995). Even though some cross-national studies suggest that education alone is a good indicator of SES for older adults (Goldman et al., 2011), multiple measures of SES may be more appropriate (Banks et al., 2006; Rosero-Bixby & Dow, 2009). Although our classification of education is very similar to that used by SAGE researchers (Vellakkal et al., 2015), and we follow other studies in classifying education into three categories (Banks et al., 2006; Rosero-Bixby & Dow, 2009), it may be that this grouping does not sufficiently distinguish inequality in South Africa. Admittedly, the sample size is smaller for urban South African blacks and colored but we observe similar patterns when we examine both urban and rural residents (results available upon request). Thus, sample size alone cannot explain the patterns observed. Explanations for the gradient in some aspects of health may not be unique to South Africa as other SAGE middle-income countries, for example, also show no education effects for depression (Anand & Phil, 2014). Although higher educational levels tend to engage in moderate or rigorous exercise (Malambo, Kengne, Lambert, De Villiers, & Puoane, 2016), it may that our measure of exercise (rigorous exercise) does not adequately distinguish between SES groups for older adults. A final main point is in regard to obesity. While sharp disparities in risk factors, such as smoking in South Africa, carry risks for health, it is not yet clear the degree to which the increasing prevalence of obesity will affect the SES gradient in middle-income countries. Obesity is an important risk factor for chronic conditions (Kuh & Ben-Shlomo, 2004) and other aspects of health such as cognition, depression, functionality. The large increase in obesity is a relatively recent phenomenon among older adults in middle-income countries. It may be too early to determine the impact of increasing obesity on the SES gradient for other aspects of health. The increasing prevalence of obesity in these populations ultimately may produce mixed patterns in the SES gradient in these other aspects of health. There are limitations to our study. First, cross-national comparisons require comparable measures of health and SES, which is not always possible in surveys of older adults. Having comparable measures for household wealth and health prevention, while enriching our study, was not possible given differences among surveys. There were no biomarkers for consumption of foods high in saturated fat in the middle-income settings. Second, cross-national comparisons are not precise due to country differences, which are often not measurable in surveys of older adults. Third, the meaning of educational attainment differs across countries signaling caution when making comparisons. Fourth, we used the legal definition of urban dwellers in the SAGE survey for Mexico and South Africa, making a direct comparison with older adults living in the capital city of Bogotá, Colombia not feasible. In spite of these limitations, rapid demographic and nutritional transitions, poor early life conditions, social mobility, an enriched nutritional environment later in life, negative health behavior, urbanization, and unique country circumstances provide a useful framework for understanding the SES gradient for health in middle-income countries. Given the importance of obesity as a risk factor, monitoring changes in the SES health gradient in related aspects of health in middle-income countries continues to be relevant. Supplementary Material Supplementary data is available at The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences online. Funding This work was supported by the Fulbright Commission in Colombia, the Colombian Department of Science, Technology, and Innovation (Colciencias), and the Aging Institute at the Pontificia Universidad Javeriana, Bogotá, within the framework of the Fulbright U.S. Scholar program, Cohort 2016–17. University of Wisconsin-Madison researchers are supported by National Institute of Health core grants to the Center for Demography and Ecology, University of Wisconsin (P2C HD047873); and the Center for Demography of Health and Aging, University of Wisconsin (P30 AG017266). Dr. Samper-Ternent was supported by the Sealy Center on Aging at the University of Texas Medical Branch in Galveston, Claude D. Pepper OAIC (P30-AG024832) and The Mexican Health and Aging Study (R01AG018016) from grants of the National Institutes of Health/National Institute on Aging. Conflict of Interest None reported. Acknowledgments The authors are particularly grateful to Robert M. Hauser, Alberto Palloni, and Emily Agree without whom this project would not have been possible. The authors thank the anonymous reviewers for their comments. Many thanks to Sarah Moen for editing the final version of the paper. Author contributions: M. McEniry provided background theories, conducted statistical analysis, and wrote the paper. R. Samper-Ternent, C. E. Flórez, R. Pardo contributed to significant discussions regarding Colombian data, measures, results, and revising the paper. C. Cano reviewed the paper and provided valuable comments. References Agyemang , C. , Kunst , A. , Bhopal , R. , Zaninotto , P. , Unwin , N. , Nazroo , J. ,… Stronks , K . ( 2010 ). A cross-national comparative study of blood pressure and hypertension between English and Dutch South-Asian– and African-origin populations: The role of national context . American Journal of Hypertension , 23 , 639 – 648 . doi: 10.1038/ajh.2010.39 Google Scholar CrossRef Search ADS PubMed Alaba , O. , & Chola , L . ( 2014 ). Socioeconomic inequalities in adult obesity prevalence in South Africa: A decomposition analysis . International Journal of Environmental Research and Public Health 11 , 3387 – 3406 . doi: 10.3390/ijerph110303387 Google Scholar CrossRef Search ADS PubMed Anand, A., & Phil, M. (2014). Understanding depression among older adults in six low-middle income countries using WHO-SAGE survey. Behavioral Health, 1, 1–11. Banks , J. , Marmot , M. , Oldfield , Z. , & Smith , J. P . ( 2006 ). Disease and disadvantage in the United States and in England . The Journal of the American Medical Association , 295 , 2037 – 2045 . doi: 10.1001/jama.295.17.2037 Google Scholar CrossRef Search ADS PubMed Barker , D. J. P . ( 1998 ). Mothers, babies and health in later life ( 2nd ed .). Edinburgh : Churchill Livingstone . Bateson , P. , & Gluckman , P . ( 2011 ). Plasticity, robustness, development and evolution . Cambridge : Cambridge University Press . Google Scholar CrossRef Search ADS Beinart , W. , & Dubow , S . (Eds.). ( 1995 ). Segregation and apartheid in twentieth-century South Africa . London : Routledge . Beltrán-Sánchez , H. , Palloni , A. , Riosmena , F. , & Wong , R . ( 2016 ). SES Gradients among Mexicans in the US and in Mexico: A new twist to the hispanic paradox ? Demography , 53 , 1555 – 1581 . doi: 10.1007/s13524-016-0508-4 Google Scholar CrossRef Search ADS PubMed Bhorat , H. , van der Westhuizen , C. , & Jacobs , T . ( 2009 ). Income and non-income inequality in post-apartheid South Africa: What are the drivers and possible policy interventions? Pretoria, South Africa : Trade & Industrial Policy Strategies . Development Policy Research Unit (DPRU) Working Paper No. 09/138. doi:10.2139/ssrn.1474271 Carrillo , R. , Barnabe-Ortiz , A. , Pillay , T. D. , Gilman , R. H. , Sanchez , J. F. , Poterico , J. ,… Miranda , J. J . ( 2016 ). Obesity risk in rural, urban and rural-to-urban migrants: Prospective results of the PERU MIGRANT study . International Journal of Obesity , 40 , 181 – 185 . doi: 10.1038/ijo.2015.140 Google Scholar CrossRef Search ADS PubMed CHDS . ( 2008 ). Closing the gap in a generation: health equity through action on the social determinants of health . Final report of the commission on social determinants of health . Geneva : World Health Organization . PubMed PubMed Crimmins , E. M. , & Finch , C. E . ( 2006 ). Infection, inflammation, height, and longevity . Proceedings of the National Academy of Sciences of the United States of America , 103 , 498 – 503 . doi: 10.1073/pnas.0501470103 Google Scholar CrossRef Search ADS PubMed Davey Smith , G. , & Lynch , J . ( 2004 ). Life course approaches to socioeconomic differentials in health . In G. D. Smith , & J. Lynch (Eds.), A life course approach to chronic disease epidemiology (pp. 77 – 115 ). Oxford : Oxford University Press . Dinsa , G. D. , Goryakin , Y. , Fumagalli , E. , & Suhrcke , M . ( 2012 ). Obesity and socioeconomic status in developing countries: A systematic review . Obesity Reviews , 13 , 1067 – 1079 . doi: 10.1111/j.1467-789X.2012.01017.x Google Scholar CrossRef Search ADS PubMed Elo , I. T. , & Preston , S. H . ( 1992 ). Effects of early-life conditions on adult mortality: a review . Population Index , 58 , 186 – 212 . doi:10.2307/3644718 Google Scholar CrossRef Search ADS PubMed FAO . ( 1946 ). World Food Survey . Washington, DC : United Nations . FAO . ( 2010 , 2017). Food balance sheets . Retrieved from http://faostat.fao.org/site/368/default.aspx#ancor Flórez, C. E., Guataquí, J. C., Mendez, R., & Cote, H. (2016). Las transformaciones sociodemográfics en Colombia Durante el Siglo XX y principios del Siglo XXI, Working paper. Bogotá, Colombia: Department of Economics, Universidad del Rosario. Ford , N. D. , Patel , S. A. , & Narayan , K. M. V . ( 2017 ). Obesity in low- and middle-income countries: Burden, drivers, and emerging challenges . Annual Review of Public Health , 38 , 145 – 164 . doi: 10.1146/annurev-publhealth-031816-044604 Google Scholar CrossRef Search ADS PubMed Franco , R. , Leon , A. , & Atria , R . ( 2007 ). Stratification and social mobility in Latin America: Structural transformations from a quarter of a century . Santiago, Chile : ECLAC and LOM Editions . Gluckman , P. , & Hanson , M . ( 2005 ). The fetal matrix: Evolution, development and disease . Cambridge : Cambridge University Press . Goldman , N. , Lin , I. F. , Weinstein , M. , & Lin , Y. H . ( 2003 ). Evaluating the quality of self-reports of hypertension and diabetes . The Journal of Clinical Epidemiology , 56 , 148 – 154 . doi: 10.1016/S0895-4356(02)00580-2 Google Scholar CrossRef Search ADS PubMed Goldman , N. , Turra , C. M. , Rosero-Bixby , L. , Weird , D. , & Crimmins , E . ( 2011 ). Do biological measures mediate the relationship between education and health: A comparative study . Social Science & Medicine , 72 , 307 – 315 . doi: 10.1016/j.socscimed.2010.11.004 Google Scholar CrossRef Search ADS He , W. , Goodkind , D. , & Kowal , P . ( 2016 ). An aging world: 2015 . U.S. Census Bureau, International Population Reports, P95/16-1. Washington, DC : U.S. Government Publishing Office . Hebert , J. R. , Hurley , T. G. , Olendzki , B. C. , Teas , J. , Ma , Y. , & Hampl , J. S . ( 1998 ). Nutritional and socioeconomic factors in relation to prostate cancer mortality: A cross-national study . Journal of the National Cancer Institute , 90 , 1637 – 1647 . doi: 10.1093/jnci/90.21.1637 Google Scholar CrossRef Search ADS PubMed Hosegood , V. , & Timaeus , I. M . ( 2006 ). HIV/AIDS and older people in South Africa . In B. Cohen , & J. Menken (Eds.), Aging in Sub-Saharan Africa: Recommendation for furthering research (pp. 250 – 275 ). Washington, DC : The National Academies Press . Kinsey , T. , Jemal , A. , Liff , J. , Ward , E. , & Thun , M . ( 2008 ). Secular trends in mortality from common cancers in the United States by educational attainment, 1993–2001 . Journal of the National Cancer Institute , 100 , 1003 – 1012 . doi: 10.1093/jnci/djn207 Google Scholar CrossRef Search ADS PubMed Kuh , D. , & Ben-Shlomo , Y . (Eds.). ( 2004 ). A life course approach to chronic disease epidemiology . New York : Oxford University Press . doi:10.1093/acprof:oso/9780198578154.001.0001 Google Scholar CrossRef Search ADS Lloyd-Sherlock , P. , Beard , J. , Minicuci , N. , Ebrahim , S. , & Chatterji , S . ( 2014 ). Hypertension among older adults in low- and middle-income countries: prevalence, awareness and control . The International Journal of Epidemiology , 43 , 116 – 128 . doi: 10.1093/ije/dyt215 Google Scholar CrossRef Search ADS PubMed López-Alonso , M . ( 2007 ). Growth with inequality: Living standards in Mexico, 1850–1950 . The Journal of Latin American Studies , 39 , 81 – 105 . doi: 10.1017/S0022216X06002045 Google Scholar CrossRef Search ADS Malambo , P. , Kengne , A. P. , Lambert , E. V. , De Villiers , A. , & Puoane , T . ( 2016 ). Prevalence and socio-demographic correlates of physical activity levels among South African adults in Cape Town and Mount Frere communities in 2008–2009 . Archives of Public Health , 74 , 1–9. doi: 10.1186/s13690-016-0167-3 Martorell, R., Stein, A. D., & Schroeder, D. G. (2001). Early nutrition and later adiposity. The Journal of Nutrition, 131, 874S–880S. doi:10.1093/jn/131.3.874S McEniry , M. C . ( 2014 ). Early life conditions and rapid demographic changes in the developing world: Consequences for older adult health . Dordrecht, Netherlands: Springer Science+Business Media. doi:10.1007/078-94-007-6979-3 Google Scholar CrossRef Search ADS McEniry, M. C. (2015). Research on Early Life and Aging Trends and Effects (RELATE): A Cross-National Study. Ann Arbor, MI: Inter-university Consortium for Political and Social Research (ICPSR34241-v2). doi:10.3886/ICPSR34241.v2 Monteiro , C. A. , Conde , W. L. , Lu , B. , & Popkin , B. M . ( 2004 ). Obesity and inequities in health in the developing world . International Journal of Obesity and Related Metabolic Disorders , 28 , 1181 – 1186 . doi: 10.1038/sj.ijo.0802716 Google Scholar CrossRef Search ADS PubMed Montez , J. K. , Hummer , R. A. , & Hayward , M. D . ( 2012 ). Educational attainment and adult mortality in the United States: A systematic analysis of functional form . Demography , 49 , 315 – 336 . doi: 10.1007/s13524-011-0082-8 Google Scholar CrossRef Search ADS PubMed Palloni , A. , & Souza , L . ( 2013 ). The fragility of the future and the tug of the past: Longevity in Latin America and the Caribbean . Demographic Research , 29 , 543 – 578 . doi: 10.4054/DemRes.2013.29.21 Google Scholar CrossRef Search ADS PubMed Pampel , F. C. , Denney , J. T. , & Krueger , P. M . ( 2012 ). Obesity, SES, and economic development: a test of the reversal hypothesis . Social Science & Medicine (1982) , 74 , 1073 – 1081 . doi: 10.1016/j.socscimed.2011.12.028 Google Scholar CrossRef Search ADS PubMed Payne , C. , Gómez-Olivé , F. X. , Kahn , K. , & Berkman , L . ( 2017 ). Physical function in an aging population in rural South Africa: Findings from HAALSI and cross-national comparisons with HRS sister studies . The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 72 , 665 – 679 . doi:10.1093/geronb/gbx030 Google Scholar CrossRef Search ADS Popkin , B. M . ( 2006 ). Global nutrition dynamics: The world is shifting rapidly toward a diet linked with noncommunicable diseases (NCDs) . The American Journal of Clinical Nutrition , 84 , 289 – 298 . doi:10.1093/ajcn/84.1.289 Google Scholar PubMed Popkin , B. M. , Horton , S. , & Kim , S . ( 2001 ). The nutritional transition and diet-related chronic diseases in Asia: Implications for prevention . Washington, DC : Food Consumption and Nutrition Division Discussion Paper International Food Policy Research Institute . Retrieved from http://www.ifpri.org/publication/nutritional-transition-and-diet-related-chronic-diseases-asia Radloff , L. S . ( 1977 ). The CES-D scale: A self-report depression scale for research in the general population . Applied Psychological Measurement , 1 , 385 – 401 . Google Scholar CrossRef Search ADS Riley , J. C . ( 2005 ). Bibliography of works providing estimates of life expectancy at birth and estimates of the beginning period of health transitions in countries with a population in 2000 of at least 400,000 . Retrieved from http://www.lifetable.de/RileyBib.htm de Romo , A. C. R. , & de Pérez , M. E. R . ( 1998 ). Historia de la salud pública en México: siglos XIX y XX [History of public health in Mexico: 19th and 20th century] . História, Ciências, Saúde-Manguinhos , 5 , 293 – 310 . doi: 10.1590/S0104-59701998000200002 Google Scholar CrossRef Search ADS PubMed Rosero-Bixby , L. , & Dow , W. H . ( 2009 ). Surprising SES gradients in mortality, health, and biomarkers in a Latin American population of adults . The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 64B , 105 – 117 . doi: 10.1093/geronb/gbn004 Google Scholar CrossRef Search ADS Schmidhuber , J. , & Shetty , P . ( 2005 ). The nutrition transition to 2030: Why developing countries are likely to bear the major burden . Paper presented at the 97th Seminar of the European Association of Agricultural Economists, University of Reading , England . Smith , K. V. , & Goldman , N . ( 2007 ). Socioeconomic differences in health among older adults in Mexico . Social Science & Medicine , 65 , 1372 – 1385 . doi: 10.1016/j.socscimed.2007.05.023 Google Scholar CrossRef Search ADS United Nations Development Program . ( 2016 ). Human development report 2016 . Washington, DC : United Nations . Unwin , N. , & Alberti , K. G . ( 2006 ). Chronic non-communicable diseases . Annals of Tropical Medicine and Parasitology , 100 , 455 – 464 . doi: 10.1179/136485906X97453 Google Scholar CrossRef Search ADS PubMed Vargas-Uricoechea , H. , & Casas-Figueroa , L. Á . ( 2015 ). An epidemiologic analysis of diabetes in Colombia . Annals of Global Health , 81 , 742 – 753 . doi: 10.1016/j.aogh.2015.11.001 Google Scholar CrossRef Search ADS PubMed Vellakkal , S. , Millett , C. , Basu , S. , Khan , Z. , Aitsi-Selmi , A. , Stuckler , D. , & Ebrahim , S . ( 2015 ). Are estimates of socioeconomic inequalities in chronic disease artefactually narrowed by self-reported measures of prevalence in low-income and middle-income countries? Findings from the WHO-SAGE survey . Journal of Epidemiology & Community Health , 69 , 218 – 225 . doi: 10.1136/jech-2014–204621 Google Scholar CrossRef Search ADS Wilkinson , R. G . ( 1996 ). Unhealthy societies: The afflictions of inequality . London : Routledge . Google Scholar CrossRef Search ADS World Bank Group . ( 2017a ). GINI index (World Bank estimate) . Retrieved from http://data.worldbank.org/indicator/SI.POV.GINI World Bank Group . ( 2017b ). GNI per capita, atlas method (current US$) . Retrieved from http://data.worldbank.org/indicator/NY.GNP.ATLS.CD World Bank Group . ( 2017c ). LAC equity lab: Income inequality - Urban/rural inequality . Retrieved from http://www.worldbank.org/en/topic/poverty/lac-equity-lab1/income-inequality/urban-rural-inequality World Health Organization . ( 2000 ). The world health report 2000: Health systems: Improving performance ( A. Haden , & B. Campanini Eds.). Geneva, Switzerland : World Health Organization . World Health Organization . ( 2015 ). World health statistics 2015 . Geneva, Switzerland : World Health Organization . Yan , S. , Li , J. , Li , S. , Zhang , B. , Du , S. , Gordon-Larsen , P. ,… Popkin , B. M . ( 2012 ). The expanding burden of cardiometabolic risk in China: The China health and nutrition survey . Obesity Reviews , 13 , 810 – 821 . doi: 10.1111/j.1467-789X.2012.01016.x Google Scholar CrossRef Search ADS PubMed Yesavage , J. A. , Brink , T. L. , Rose , T. L. , Lum , O. , Huang , V. , Adey , M. , & Leirer , V. O . ( 1982–1983 ). Development and validation of a geriatric depression screening scale: A preliminary report . Journal of Psychiatric Research , 17 , 37 – 49 . doi: 10.1016/0022-3956(82)90033–4 Google Scholar CrossRef Search ADS © The Author(s) 2018. Published by Oxford University Press on behalf of The Gerontological Society of America. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

The Journals of Gerontology Series B: Psychological Sciences and Social SciencesOxford University Press

Published: Apr 19, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off