Environmental factors and cardiovascular diseases: the association of income inequality and green spaces in elderly residents of São Paulo, Brazil.

OBJECTIVE: We aimed to analyse the individual and contextual determinants associated with cardiovascular diseases (CVDs) morbidity among the elderly.
METHODS: The sample consisted of 1333 individuals aged 60 or older residing in the city of São Paulo, from the Health, Welfare and Aging (SABE) study survey performed in 2010. The association between CVD with both income inequality and green spaces was analysed using Bayesian multilevel models, controlling for individual and contextual factors.
RESULTS: We found a significant association between income inequality and green spaces, and risk of CVD. In comparison to elderly residents in areas with low-income inequality, there was an increased risk for CVD among those residing in the medium-low (OR=1.35, 95% CI 1.15 to 1.59), medium-high (OR=2.71, 95% CI 2.18 to 3.36) and high (OR=1.43, 95% CI 1.14 to 1.79) quartiles of income inequality. Those living in medium-low (OR=0.44, 95% CI 0.39 to 0.49), medium-high (OR=0.56, 95% CI 0.48 to 0.65) and high (OR=0.48, 95% CI 0.43 to 0.55) green spaces levels had lower risk of CVD.
CONCLUSIONS: These findings highlight the importance of area-level characteristics on CVD risk and the need to develop healthcare policies focused on the effect of individual and contextual characteristics. Published by the BMJ Publishing Group Limited. For permission to use (where not already granted under a licence) please go to http://www.bmj.com/company/products-services/rights-and-licensing/

To best of our knowledge, the results of our analysis are the first to highlight the importance of area-level income inequality and green areas for cardiovascular diseases morbidity in a large and highly unequal city of Latin America.

This study indicates the importance of contextual characteristics of the neighbourhood of residence for cardiovascular diseases in the elderly, an age group that will require ever more attention in the near future.

Individual data are amenable to recall bias, especially given the age structure of the sample.

The sample was not representative for the residential areas, so one should not draw conclusions specifics for each of the administrative areas.

Introduction

Population ageing is a global phenomenon that has led to important social and healthcare changes.1 According to the WHO,2 life expectancy has steadily increased worldwide, although it is still much higher in developed countries compared with middle-income and low-income countries.

This increased longevity means a longer exposure period to risk factors of non-communicable diseases. Cardiovascular disease (CVD) is now considered the most important cause of death worldwide.3 In Latin America and the Caribbean, CVD was responsible for the highest proportion of preventable mortality in 2011: 42% in men and 50% in women.4

In Brazil, mortality rates attributed to CVD have recently decreased, but CVD still remains the main cause of death, especially in older residents.5 In 2013, CVDs were responsible for 28.1% of all deaths in the country, and if one excludes deaths from violence and other external causes, it is responsible for 32.1% of total mortality.6

Morbidity from CVD is also a growing concern in public health. Total years lived with a chronic illness among older adults has increased over time, especially due to the advances in treatments of infectious diseases and the longer survival of individuals with chronic diseases. CVD is now responsible for the highest number of years spent with a disability among the elderly.7

Contextual characteristics of the residential area are important determinants of health outcomes, especially in developing countries.8 9 Recent studies have found that socioeconomic characteristics of the neighbourhoods (such as income inequality, violence and median income) are associated with risk of mortality in developing countries.10 11 Contextual characteristics have also been found to be associated with other health outcomes, such as the association of income inequality with exposure to chronic diseases risk factors12 and risk for depression.13 However, studies that analysed the association of characteristics of the neighbourhood and risk for CVD are still scarce in Brazil, a country with a wide range of socioeconomic and environmental factors that could significantly impact health. The present study aimed to analyse the effect of individual and contextual determinants of CVD morbidity among elderly residents of the largest city in Latin America.

Methods

Data

This study is part of the Health, Welfare and Aging Study (SABE), a representative multiple-stage probabilistic sample of individuals aged 60 years and older residing in the city of São Paulo.14 Data were obtained from a household questionnaire administered in 2010, which comprised of 11 thematic blocks covering information on sociodemographic factors, cognitive evaluation, health conditions, functional status, mobility test, drugs use, access to services, family network, social support, work history and housing characteristics. This study is coordinated by the Department of Epidemiology of the School of Public Health of the University of São Paulo.14

Study design

A cross-sectional study was conducted to analyse the association between the individual and environmental characteristics with CVD morbidity in elderly residents of São Paulo, Brazil.

Variables

The dependent variable is the self-reported presence of a CVD confirmed by a medical diagnosis. At the individual level, the demographic and socioeconomic variables of interest are gender, age (categorised into 60–64, 65–69, 70–74, 75–79, 80–84 and 85 or more years), race (white, mixed, black and others), educational attainment (categorised according of the total number of years of formal education), income (categorised in terms of minimum wages of the year of data collection) and marital status (with or without a partner).

The models also included behavioural factors such as alcohol consumption, risk of alcoholism, smoking and body mass index (BMI) to control for individual confounding. Data for alcohol consumption was obtained by a self-referred 3-month consumption question (yes or no). Risk of alcoholism was assessed according to the Short Michigan Alcoholism Screening Test-Geriatric (SMAST-G).15 Smoking history was divided into three categories: never smoked, used to smoke and currently smoking. The BMI of each respondent was calculated by using the individual's weight and height (kg/m2). Respondents were then subsequently categorised into underweight (<22 kg/m2), normal weight (≥22 and <27 kg/m2) and overweight (≥27 kg/m2) as recommended by the Technical Standard of Food and Nutrition Surveillance for reference values of the elderly.16 We also included the two most prevalent chronic diseases (diabetes and hypertension) to test their independent associations with CVD morbidity.

Socioeconomic and environmental characteristics of area of residence were calculated using data from the 2010 Census for each of the 32 administrative areas of the municipality of São Paulo, known as subprefeituras. The contextual characteristics of interest were income inequality (measured by the Gini Index), green spaces per inhabitant (total green area per square metre) and average per capita income of the administrative area. The contextual variables were categorised in quartiles and classified using a qualitative terminology (low, medium–low, medium–high and high, respectively).

Statistical analysis

Multilevel logistic regression was conducted to analyse the association between contextual characteristics and risk of CVD in order to account for the clustering within the neighbourhood, assuming the non-independence of observations. The first level of the model referred to individual-level characteristics and the second level to the area-level (contextual) characteristics. We applied Bayesian multilevel models, a recommended approach to decrease bias in multilevel analysis with dichotomous-dependent variables. The approach also allows the comparison of model fit by comparing the Bayesian Information Criterion (BIC) of each model, where decreasing values indicate better fit.17

The analyses were performed with Stata V.13.1 (Stata Corporation, College Station, Texas, USA, 2013). For the descriptive statistics of demographic, socioeconomic, behavioural variables and for the bivariate analysis according to CVDs presence, we included the survey mode procedure, which considers the complex sample design regarding sampling weight and individual clustering (secondary sampling units) within census tracts (primary sampling units). Multilevel analyses were performed using the gllamm command, including the corresponding weighting to take into account the complex sampling design.

Results

Sample characteristics

The sample consisted of 1333 individuals aged 60 or older residing in the city of São Paulo in 2010. Descriptive analysis indicates that most of the individuals were females (59.9%), 58.6% reported being white and around 35% being black or brown (29.3% brown and 6.7% black, respectively). Most of the elderly had low educational attainment, with more than 70% having only 7 years of formal education or less. In relation to income, most of the elderly earned three minimum wages or less, and 46.3% had income between 1 and 2 minimum wage per capita (table 1).

Table 1

Distribution of demographic, socioeconomic and behavioural characteristics, presence of diabetes, hypertension and CVD of the elderly, 2010, São Paulo, Brazil

Demographic and socioeconomic characteristics	n*	Per cent†
Gender
Male	477	40.06
Female	856	59.94
Total	1333	100
Age categories, years
60–64	355	31.62
65–69	231	22.59
70–74	218	17.73
75–79	166	12.78
80 years or older	363	15.28
Total	1333	100
Race
White	775	58.67
Brown	383	29.30
Black	94	6.73
Other	68	5.30
Total	1320	100
Education, years of formal study completed
0–3	527	35.34
4–7	494	37.25
8+	311	27.41
Total	1332	100
Income, minimum wage
<1	38	3.18
1–2	579	46.65
2–3	238	18.98
3–4	124	11.24
4+	189	19.95
Total	1168	100
Marital status
With partner	659	54.97
Without partner	658	45.03
Total	1317	100
Alcohol ingestion	951	68.20
No	951	68.20
Yes	381	31.80
Total	1332	100
Risk of alcoholism
No	341	88.41
Yes	37	11.59
Total	378	100
Smoking
Never smoked	702	51.03
Former smoker	488	37.02
Current smoker	142	11.96
Total	1332	100
BMI
Underweight	169	11.2
Normal weight	425	33.7
Overweight	650	55.1
Total	1244	100
Diabetes
No	998	74.83
Yes	333	25.17
Total	1331	100
Hypertension
No	432	33.24
Yes	900	66.76
Total	1332	100
Cardiovascular diseases
No	1009	77.13
Yes	322	22.87
Total	1331	100

Source: SABE study, 2010.

*Absolute numbers on the unweighted sample.

†Weighted sample proportion.

BMI, body mass index; CVD, cardiovascular disease; SABE, Health, Welfare and Aging.

Of the sample, 70% reported not having ingested alcohol in the past 3 months. Among those reporting having ingested alcohol, most of the elderly were not at risk of alcoholism (88.1%), according to SMAST-G. Less than half the elderly never smoked (about 37% reported being a former smoker and a smaller proportion, 11.9%, reported being current smoker). Most of the elderly were categorised as being overweight (55.1%). A medical diagnosis of diabetes was reported by 25.2% of the elderly population. Prevalence estimates for hypertension and CVD were 66.7% and 22.8%, respectively (table 1).

Bivariate associations with CVD

The prevalence of CVD did not differ across sociodemographic groups, such as gender, race, education, income and marital status. Higher prevalence of CVD was observed in higher age groups (table 2).

Table 2

Demographic and socioeconomic characteristics according to CVDs presence, 2010, São Paulo, Brazil

	CVDs presence
	No (%)	Yes (%)	Total (%)	p Value*
Gender				0.489
Male	78.07	21.93	100
Female	76.50	23.50	100
Age categories				0.000
60–64	84.94	15.06	100
65–79	77.33	22.67	100
70–74	74.60	25.40	100
75–79	74.26	25.74	100
80 years or older	66.01	33.99	100
Race				0.449
White	76.58	23.42	100
Brown	79.36	20.64	100
Black	70.27	29.73	100
Other	76.77	23.23	100
Education, years of formal study completed				0.057
0–3	73.81	26.19	100
4–7	76.51	23.49	100
8+	82.13	17.87	100
Income, minimum wage				0.215
<1	84.59	15.41	100
1–2	74.97	25.03	100
2–3	75.68	24.32	100
3–4	83.78	16.22	100
4+	79.63	20.37	100
Marital status				0.301
With partner	75.75	24.25	100
Without partner	78.18	21.82	100
Alcohol ingestion				0.044
No	75.35	24.65	100
Yes	80.93	19.07	100
Risk of alcoholism				0.480
No	81.38	18.62	100
Yes	76.51	23.49	100
Smoking				0.005
Never smoked	77.33	22.67	100
Former smoker	73.68	26.32	100
Current smoker	86.76	13.22	100
BMI				0.313
Underweight	80.31	19.69	100
Normal weight	78.08	21.92	100
Overweight	75.29	24.71	100
Diabetes				0.000
No	81.15	18.85	100
Yes	65.03	34.97	100
Hypertension				0.000
No	88.35	11.65	100
Yes	71.59	28.41	100

*χ2 test.

BMI, body mass index; CVD, cardiovascular disease.

We found a statistically significant association of alcohol consumption and smoking with risk for CVD. Presence of CVDs was less frequent for individuals who reported alcohol ingestion during the past 3 months (19.1%). Current smokers had the lowest presence of CVD (13.2%), followed by individuals who never smoked (22.6%) and former smokers (26.3%; table 2).

Statistically significant differences in the prevalence of CVDs were also observed for individuals with diabetes and hypertension. Among individuals who reported being diagnosed with diabetes, ∼35% also had been diagnosed with a CVD, while among those reporting having hypertension the proportion was 28% (table 2).

Multilevel regression models

Results from the adjusted models indicated a statistically significant association between some of the individual characteristics and CVD. Older individuals had statistically higher odds of CVD. Completing 8 years or more of formal education was also independently associated with greater risk for CVD. Those living without a partner were significantly less likely to report CVD in comparison with those living with a partner (OR=0.57, 95% CI 0.40 to 0.81). We did not find any significant association between race, gender and individual income with CVD risk (table 3). The analysis of behavioural characteristics indicated that alcohol consumption was significantly associated with lower odds of having a CVD (OR=0.74, 95% CI 0.56 to 0.99). We found no significant association between smoking and BMI with CVD risk (table 3). On the other hand, both diabetes (OR=2.04, 95% CI 1.39 to 2.98) and hypertension (OR=2.33, 95% CI 1.37 to 3.95) were associated with CVD risk (table 3).

Table 3

Multilevel logistic models of CVD prevalence according to demographic, socioeconomic, behavioural and environmental characteristics in the elderly, 2010, São Paulo, Brazil

	Empty model (n=1257)		Model 1 (n=1005)		Model 2 (n=1005)		Model 3 (n=1005)
	OR	95% CI	OR	95% CI	OR	95% CI	OR	95% CI
1° Level
Intercept	0.22**	0.21 to 0.23	0.07**	0.02 to 0.18	0.06**	0.02 to 0.17	0.08**	0.03 to 0.24
Gender
Female			1.04	0.79 to 1.38	1.04	0.79 to 1.38	1.05	0.79 to 1.38
Age categories
65–69			1.89*	1.17 to 3.04	1.84*	1.12 to 3.03	1.92*	1.16 to 3.19
70–74			2.07*	1.21 to 3.55	2.05*	1.19 to 3.56	2.15*	1.24 to 3.74
75–79			1.82*	1.15 to 2.88	1.82*	1.13 to 2.92	1.88*	1.17 to 3.01
80–84			3.44**	2.19 to 5.39	3.45**	2.18 to 5.47	3.56**	2.26 to 5.59
Race
Brown			1.01	0.77 to 1.32	0.99	0.76 to 1.30	0.98	0.75 to 1.29
Black			1.33	0.76 to 2.34	1.33	0.76 to 2.34	1.31	0.75 to 2.28
Other			0.95	0.48 to 1.86	0.95	0.48 to 1.87	0.97	0.49 to 1.91
Education (years)
4–7			0.92	0.64 to 1.31	0.92	0.64 to 1.32	0.92	0.64 to 1.33
8+			0.62*	0.39 to 0.97	0.62*	0.39 to 0.99	0.62*	0.39 to 0.98
Income, minimum wages
1–2			1.46	0.76 to 2.82	1.51	0.78 to 2.89	1.48	0.77 to 2.84
2–3			1.53	0.80 to 2.92	1.59	0.83 to 3.04	1.57	0.83 to 2.97
3–4			1.10	0.44 to 2.71	1.15	0.46 to 2.82	1.13	0.46 to 2.79
4+			1.81	0.79 to 4.12	1.89	0.83 to 4.34	1.83	0.80 to 4.16
Marital status
Without partner			0.56*	0.39 to 0.81	0.57*	0.40 to 0.81	0.57*	0.40 to 0.81
Alcohol ingestion
Yes			0.74*	0.56 to 0.98	0.74*	0.56 to 0.99	0.74*	0.56 to 0.99
Smoking
Former smoker			1.51	0.94 to 2.42	1.53	0.95 to 2.46	1.51	0.94 to 2.44
Current smoker			0.77	0.43 to 1.36	0.77	0.44 to 1.36	0.76	0.43 to 1.36
BMI
Normal weight			1.18	0.72 to 1.94	1.21	0.73 to 2.00	1.19	0.72 to 1.97
Overweight			1.38	0.87 to 2.19	1.41	0.89 to 2.24	1.39	0.87 to 2.21
Diabetes
Yes			2.01**	1.37 to 2.95	2.04**	1.39 to 2.98	2.02**	1.37 to 2.97
Hypertension
Yes			2.32*	1.37 to 3.94	2.33**	1.37 to 3.95	2.35*	1.38 to 4.01
2° Level: neighbourhood
Income inequality, quartile
2					1.17	0.99 to 1.39
3					2.38**	2.03 to 2.78
4					0.90	0.77 to 1.05
Green space m2/inhabitant, quartile
2							0.44**	0.39 to 0.49
3							0.56**	0.48 to 0.65
4							0.48**	0.43 to 0.55
BIC (ICC)	1 313 315	(0.017)	969 023	(0.018)	970 132	(0.053)	968 930	(0.03)

*p<0.05.

**p≤0.001.

BMI, body mass index; BIC, Bayesian Information Criterion; CVD, cardiovascular disease; ICC, intraclass correlation coefficient.

At the contextual level, we observed significant associations between income inequality, green spaces, average income and CVD risk. In comparison to those living in areas with low-income inequality, those living in area of medium–high income inequality (OR=2.38, 95% CI 2.03 to 2.78) were at greater odds of having a CVD (table 3). In comparison to those living with the low green space, those in the medium–low, medium–high and high green space level (OR=0.44, 95% CI 0.39 to 0.49; OR=0.56, 95% CI 0.48 to 0.65 and OR=0.48, 95% CI 0.43 to 0.55, respectively) were significantly less likely to report having a CVD (table 3).

When adjusting for average household income within administrative areas of the São Paulo, in comparison with the residents in areas with lowest income inequality, those living in the medium–low (OR=1.35, 95% CI 1.15 to 1.59, medium–high (OR=2.71, 95% CI 2.18 to 3.36) and high-income inequality level (OR=1.43, 95% CI 1.14 to 1.79), had higher odds of CVD (table 4). In comparison to the spaces with the lowest amount of green space, those in the medium–low (OR=0.37, 95% CI 0.34 to 0.41), medium–high (OR=0.48, 95% CI 0.41 to 0.58) and high (OR=0.48, 95% CI 0.42 to 0.54) green space level had significantly lower risk for CVD (table 4).

Table 4

Multilevel logistic models of CVD prevalence adjusted by individual characteristics, according to area-level average income, income inequality and green spaces, 2010, São Paulo, Brazil

	Full modelIncome inequality (n=1005)		Full modelGreen spaces (n=1005)
	OR	95% CI	OR	95% CI
1° Level
Intercept	0.06**	0.02 to 0.17	0.12**	0.04 to 0.33
2° Level: neighbourhood
Average income, quartile
2	1.14	0.97 to 1.33	0.97	0.85 to 1.11
3	0.76*	0.63 to 0.92	0.98	0.86 to 1.11
4	0.49**	0.39 to 0.63	1.41**	1.24 to 1.60
Income inequality, quartile
2	1.35**	1.15 to 1.59
3	2.71**	2.18 to 3.36
4	1.43*	1.14 to 1.79
Green space m2/inhabitant, quartile
2			0.37**	0.34 to 0.41
3			0.48**	0.41 to 0.58
4			0.48**	0.42 to 0.54
BIC (ICC)	969 991	(0.062)	969 017	(0.071)

*p<0.05.

**p≤0.001.

BIC, Bayesian Information Criterion; CVD, cardiovascular disease.

Discussion

The study highlights the association between sociodemographic, economic, behavioural and environmental factors and CVD risk among elderly residents of São Paulo. At the individual level, being older, having lower educational attainment, living with a partner, not consuming alcohol and having diabetes or hypertension were consistently associated with higher odds of CVDs. At the contextual level, area-level characteristics such as income inequality, green spaces and average income were also statistically associated with CVDs.

Older age has been consistently associated with increased exposure to risk factors and, consequently, to health disorders. In this context, the prevalence of chronic diseases has been increasingly associated with ageing.7 The significant association found here between CVDs and older age, reinforces the importance of specific strategies and healthcare in older individuals, especially given the fact that the presence of CVD occurs often concurrently with other chronic diseases.

Our analysis also found that individuals with higher education had overall lower odds of CVDs. Studies in Brazil18 and other countries19 have historically found a higher proportion of chronic diseases for the most socially vulnerable individuals. The lower odds of CVDs among the elderly with higher education found for the present analysis emphasises the close relationship between social inequality and poor health.

Regarding marital status and CVDs, the current literature indicates an inverse association between living with a partner and health.20 21 Gomes et al,22 in a study with elderly residents of São Paulo, described the association between living without a partner and higher risk of death. However, we found a significant association between living without a partner and lower risk for CVDs. This finding can be explained by the fact that older individuals living with a partner may have higher survival rates after being diagnosed with a disease, and that persons who lived without a partner may have died earlier than those with a companion, increasing the proportion of healthy elderly in the sample. Moreover, marital status may change over time and the results of the cross-sectional approach can be influenced by cohort effects.

Alcohol use, more specifically alcohol abuse, is frequently mentioned as a risk factor for several chronic diseases.23 However, evidence of a beneficial effect of moderate alcohol consumption for CVDs has been described in the literature.24 We found that individuals who reported alcohol consumption had a lower risk for CVD. It can be the case that alcohol consumption for this cohort is mostly moderate, since a small proportion of the elderly showed risk of alcoholism, according to SMAST-G results.

The improvement of health habits, such as healthy diet, not smoking, moderate alcohol consumption and regular practice of physical activity has been shown to have an important impact in reducing the risk of developing chronic diseases.1 Therefore, the intervention on modifiable risk factors and lifestyle should be encouraged in the elderly to control chronic diseases, especially when the disease has high comorbidity, such as CVDs with diabetes and hypertension, as was the case for the present study.

We found a consistently significant association between CVDs and the characteristics of the area of residence analysed by the study, that is, income inequality and green spaces. Higher income inequality was consistently associated with higher presence of CVD for the elderly. On a previous study of cardiovascular mortality in São Paulo, Farias25 found that residents living in regions with lower socioeconomic status had higher odds of CVD mortality. Similarly, the evidence found in this study indicates increased odds of having a CVD among those who live in very unequal areas, even after controlling for the average income of the residence area.

Access to green spaces has been previously associated with better health conditions, such as stress reduction.26 27 We found an inverse association between presence of green spaces and risk for CVD, indicating that residing in a region with more green spaces was protective against CVD, independently of socioeconomic factors and average income of the administrative area. This finding contributes to the discussion about the importance of green spaces for health promotion, and needs to be taken into consideration in decisions about urban planning.

The burden of CVDs is a growing public health challenge, especially in middle-income and low-income countries, where it represents the majority of mortality causes.28 Our analysis indicate the importance of contextual characteristics of the neighbourhood of residence for CVDs in the elderly, an age group that will require ever more attention in the near future, especially considering the rapid demographic transition occurring in Brazil.

The study has some limitations to be considered. First, individual data are amenable to recall bias, especially given the age structure of the sample. Second, the sample was not representative for the residential areas, so one should not draw specific conclusions for each of the administrative areas. Third, validated physical activity scores were not available from the study, which could influence the association between contextual characteristics and CVDs. Fourth, we used cross-sectional data, so a temporal association could not be established.

Conclusion

The study highlights the importance of contextual and individual factors associated with the risk for CVDs. The results found here for the largest Latin America city can reflect the future situation of the elderly in the region and indicate the need to develop healthcare policies focused on the effect of individual and contextual characteristics.