Built environment and social environment : associations with overweight and obesity in a sample of Brazilian adults

The aim of this study was to assess associations between the built environment and social environment and excess weight in an urban population. Participants were selected from the Surveillance System for Risk Factors for Chronic Diseases (VIGITEL). The study used data from the city of Belo Horizonte, Minas Gerais State, Brazil. A total of 3,425 interviews from the years 2008 and 2009 were used. Georeferenced data on parks, squares, and locations for physical exercise, population density, and food stores were used to assess the built environment. Description of the social environment used income and homicide rate for the neighborhood. Environmental variables associated independently with excess weight were population density, presence of parks, squares, and locations for physical exercise, and self-reported presence of locations for physical exercise. The findings show that residential neighborhood characteristics are associated with excess weight in urban adults. Controlled Environment; Environment and Public Health; Obesity; Overweight Resumo O objetivo deste estudo foi avaliar as associações das variáveis do ambiente construído e social com o excesso de peso em uma população urbana. Os participantes foram selecionados com base no Sistema de Vigilância de Fatores de Risco para Doenças Crônicas (VIGITEL). O trabalho foi realizado com dados da cidade de Belo Horizonte, Minas Gerais, Brasil. Foram utilizadas 3.425 entrevistas referentes aos anos de 2008 e 2009. Informações georreferenciadas de parques, praças, lugares para a prática de atividade física, densidade populacional e estabelecimentos alimentares foram usadas para avaliar o ambiente construído. Para caracterizar o ambiente social foi utilizada renda e taxa de homicídio da vizinhança. As variáveis ambientais associadas independentemente com excesso de peso foram densidade populacional, presença de parques, praças e locais para a prática de atividade física e o autorrelato de locais para a prática de atividade física. As evidências deste estudo mostram que as características das vizinhanças onde as pessoas residem estão associadas ao excesso de peso de adultos do meio urbano. Meio Ambiente Construído; Meio Ambiente e Saúde Pública; Obesidade; Sobrepeso http://dx.doi.org/10.1590/0102-311X00078112 BUILT ENVIRONMENT AND SOCIAL ENVIRONMENT 1989 Cad. Saúde Pública, Rio de Janeiro, 29(10):1988-1996, out, 2013 Introduction Obesity is a chronic illness with a complex etiology and multiple associated factors such as life habits and genetic and biological susceptibility 1. Recently, the characteristics of the environment in which people live, including neighborhood socioeconomic status, availability and affordability of healthy foods, and opportunities for physical activity have been widely associated with the current obesity epidemic in various countries 2,3,4. When such environments encourage physical inactivity and unhealthy food choices, they are described as obesogenic, i.e., responsible for promoting obesity 5. In this context, the current obesity epidemic has been related to four main factors: built environment; social environment; individual behavior; and individual biological factors 6. The built environment can present opportunities for unhealthy eating, such as the absence of stores with fresh and healthy food produce and barriers to energy expenditure in daily routines 7,8,9, such as absence of sidewalks along the streets for walking to and from places, low connectivity between streets, insufficient lighting, and neighborhoods with limited recreational facilities, which can hinder physical exercise 10. In addition, neighborhoods with lower residential and population density, less connected streets, and low mixed land use are associated with increased risk of obesity 11,12,13,14,15,16. In addition to the built environment, the social environment can also affect eating patterns and level of physical activity in neighborhoods and thereby contribute to weight increase. The social environment includes elements related to individual living conditions such as income, schooling, crime record, social support networks, and level of trust, which are associated with greater or lesser social disorder and social deprivation in the neighborhood 17. Thus, understanding the environment’s role in the increased prevalence of obesity is fundamental for developing effective strategies to prevent this disease and thereby reduce its negative impacts on the population’s health. Various observational studies in recent decades have shown an increasing consensus concerning the existence of obesogenic environments 18,19,20, but there is still controversy on the formalization, definition, measurement, and characterization of the environmental components related to weight gain and difficulties in defining the environments and the unit used to define neighborhood, where some studies used census tracts and others use boroughs, buffers, and other specific neighborhood units in countries where they are located 21,22,23,24. Although obesogenic environments have been extensively studied in developed countries 23,24, the theme needs to be explored in greater depth in Brazilian studies with the use of different methodologies, populations (children, adolescents, adults, and the elderly), and contexts (households, schools, and the workplace). The current study thus aimed to estimate associations between variables from the built and social environments and excess weight in an urban setting in Brazil.


Abstract
The aim of this study was to assess associations between the built environment and social environment and excess weight in an urban population.Participants were selected from the Surveillance System for Risk Factors for Chronic Diseases (VIGITEL).The study used data from the city of Belo Horizonte, Minas Gerais State, Brazil.A total of 3,425 interviews from the years 2008 and 2009 were used.Georeferenced data on parks, squares, and locations for physical exercise, population density, and food stores were used to assess the built environment.Description of the social environment used income and homicide rate for the neighborhood.Environmental variables associated independently with excess weight were population density, presence of parks, squares, and locations for physical exercise, and self-reported presence of locations for physical exercise.The findings show that residential neighborhood characteristics are associated with excess weight in urban adults.

Introduction
Obesity is a chronic illness with a complex etiology and multiple associated factors such as life habits and genetic and biological susceptibility 1 .Recently, the characteristics of the environment in which people live, including neighborhood socioeconomic status, availability and affordability of healthy foods, and opportunities for physical activity have been widely associated with the current obesity epidemic in various countries 2,3,4 .When such environments encourage physical inactivity and unhealthy food choices, they are described as obesogenic, i.e., responsible for promoting obesity 5 .
In this context, the current obesity epidemic has been related to four main factors: built environment; social environment; individual behavior; and individual biological factors 6 .The built environment can present opportunities for unhealthy eating, such as the absence of stores with fresh and healthy food produce and barriers to energy expenditure in daily routines 7,8,9 , such as absence of sidewalks along the streets for walking to and from places, low connectivity between streets, insufficient lighting, and neighborhoods with limited recreational facilities, which can hinder physical exercise 10 .In addition, neighborhoods with lower residential and population density, less connected streets, and low mixed land use are associated with increased risk of obesity 11,12,13,14,15,16 .
In addition to the built environment, the social environment can also affect eating patterns and level of physical activity in neighborhoods and thereby contribute to weight increase.The social environment includes elements related to individual living conditions such as income, schooling, crime record, social support networks, and level of trust, which are associated with greater or lesser social disorder and social deprivation in the neighborhood 17 .Thus, understanding the environment's role in the increased prevalence of obesity is fundamental for developing effective strategies to prevent this disease and thereby reduce its negative impacts on the population's health.
Various observational studies in recent decades have shown an increasing consensus concerning the existence of obesogenic environments 18,19,20 , but there is still controversy on the formalization, definition, measurement, and characterization of the environmental components related to weight gain and difficulties in defining the environments and the unit used to define neighborhood, where some studies used census tracts and others use boroughs, buffers, and other specific neighborhood units in countries where they are located 21,22,23,24 .
Although obesogenic environments have been extensively studied in developed countries 23,24 , the theme needs to be explored in greater depth in Brazilian studies with the use of different methodologies, populations (children, adolescents, adults, and the elderly), and contexts (households, schools, and the workplace).The current study thus aimed to estimate associations between variables from the built and social environments and excess weight in an urban setting in Brazil.

Methodology
This was a cross-sectional, descriptive, and analytical epidemiological study using the VIGITEL database, with access granted by the Brazilian Ministry of Health 25,26  This study used the interviews from the samples in the years 2008 and 2009 in the city of Belo Horizonte.Sampling of participants was done in two stages: (1) random selection of households with landline telephones and (2) selection of interviewees 18 years or older.For each Brazilian capital city, the VIGITEL system aims to obtain probabilistic samples of the population of adult residents in households served by a landline telephones, and sampling weights were subsequently assigned to the individuals interviewed by the system, with the aim of at least partially correcting biases resulting from non-universal coverage of the landline telephone system.This study considered 4,000 eligible interviews, of which 3,661 were used because they contained the data on weight and height.Of these, 131 interviews (3.6%) were excluded, from participants with body mass index (BMI) less than 18.5kg/m 2 .There was an additional loss of 105 individuals due to locating errors in the data georeferencing process.The final sample thus consisted of 3,425 individuals.
Data were used from the standardized questionnaire conducted with a telephone interview, with information on demographics, weight, and height, all self-reported.
The dependent variable for the current study was BMI.BMI was categorized in two groups: normal (18.5kg/m 2 ≤ BMI < 25kg/m 2 ) and excess weight (BMI ≥ 25kg/m 2 ) according to the values proposed by the World Health Organization (WHO) 27 .Weight and height were self-reported.These data have been widely used in epidemiological studies 28 and were previously validated for Brazilian adults 29 .
The environmental variables selected for this study were: population density, supermarkets and hypermarkets, fruit and vegetable stores, parks, public squares, and places for practicing physical exercise, homicide rate, and mean family income (Table 1).
To verify the spatial distribution of variables and analyze them, the VIGITEL was geocoded using the postal codes of the participants' homes, obtaining values for latitude and longitude based on the centroid of the street corresponding to the address's postal code.Individuals were later clustered according to the coverage age where they resided, also based on latitude and longitude.
Characterization of the built environment and social environment used the development of a geocoded base, incorporating each participant's personal data into the sample.An existing, updated list was then used, based on the registries of the current addresses of supermarkets and hypermarkets in the city.The latter was used as an indicator of the availability of different foods.Georeferenced data on parks, public squares, places for physical exercise, and population density were also used to comprise the data on the built environment.These data were provided by the Belo Horizonte Municipal Company for Information and Information Technology (Prodabel).
The social environment was characterized with georeferenced data on the locations where homicides occurred, based on latitude and longitude, thus used to create a homicide rate for the coverage areas of the primary health units.These data were provided by the Integrated Center for Social Defense Information of the Minas Gerais State Military Police.Mean family income was obtained from the databases of the Brazilian Institute for Geography and Statistics (IBGE) and was also used as a social environment variable.
Analysis of environmental data began with the Geographic Information System, with the aim of stratifying the geocoded data in layers of information and isolating spatial relations.Cross-analyses of the georeferenced data were translated into variables for the analyses.These procedures were conducted with MapInfo, version 10.5 (MapInfo Corp., New York, USA).
Distribution of cases was analyzed with information on exact location (latitude and longitude), based on the coverage area of the primary health unit (set of territorially demarcated census tracts), which are 147 subdivisions of the nine Health Districts in the city of Belo Horizonte.These areas were proposed considering the boundaries of the census tracts set by IBGE and digitized by Prodabel; for the current study, the coverage area was used as the neighborhood unit.Coverage areas averaged 2.55km 2 .The total population was 2,081,147, with 693,773 households.
Data analysis used the survey module of Stata, version 12 (Stata Corp., College Station, USA), which considers the various aspects of the complex sampling design.The sample was characterized with frequency distribution tables for the target variables according to the subjects' BMI.Environmental variables were described by measures of central tendency and dispersion.
In the univariate and multivariate analyses, the measure of association was prevalence ratio (PR), calculated with the aid of Poisson regression with an estimator of robust variance 30 .Statistical differences were assessed according to the pseudo-likelihood ratio and Wald test.The criterion used for selection of variables in the bivariate analysis was p ≤ 0.20.After the bivariate analysis, the association was verified between excess weight and the environmental variables adjusted for potential confounding factors: gender, age, conjugal status, and physical inactivity.
Table 3 shows the descriptive measures of the target environmental variables.Georeferencing included a total of 174 supermarkets, food marts, and hypermarkets, 171 fruit and vegetable stores, and 107 squares, parks, and locations for physical exercise.
Table 4 shows the prevalence rates and estimators of crude and adjusted associations between the built environment and social environment and excess weight.The crude analysis showed low prevalence rates for excess weight

Discussion
The study's results illustrate the independent relationship between environmental variables and excess weight.Higher population density and the presence of parks, squares, and other locations for physical exercise were associated with lower prevalence of excess weight.This study showed that living in neighborhoods with higher population density (3 rd tertile) was associated with lower prevalence of excess weight, an association that was maintained after adjusting for gender, age, and conjugal status.Neighborhoods with high population density often display mixed land use and greater connectivity between streets.They are thus more amenable to circulating on foot, which can be related to lower BMI 10,31,32 .
Studies in North American cities 33,34 found that persons living in neighborhoods with higher population density showed significantly lower BMI than those living in areas with low population density.An inverse association was seen between BMI and mixed land use, number of bus and subway stops, and number of street intersections.This suggests that higher population density in the neighborhood is associated with more circulation on foot, such that individuals living in neighborhoods with higher population density used automobiles less and consequently engage in more physical activity by walking back and forth on a routine basis, thus presenting lower obesity prevalence 34 .
The current study also showed associations with the presence of parks, squares, and locations for physical exercise in the neighborhood.Recent studies have suggested that parks, squares, and other green spaces for recreation in neighborhoods are important places for people to be more physically active 35 , especially in urban areas where access to green spaces is more limited 36 .There is also evidence that improved availability and access to green spaces for recreation in urban areas can be an effective public health measure, encouraging greater participation in leisure-time activities and thus reducing the risk of overweight and obesity 37,38 .
A study in the city of Curitiba, Paraná State, Brazil, with objective measures of the built environment found that adults living in neighborhoods with higher income, more gyms, and closer proximity to recreational and sports centers were more capable of following recommendations for leisure-time walking.However, the density of recreational infrastructures such as bicycle lanes, parks, wooded areas, and squares was not statistically associated with physical activity in the study 39 .For variables related to the social environment, no associations were observed, but other studies have shown that living in economically underprivileged neighborhoods is related to higher BMI and greater weight gain, and other studies have found that the odds ratio of overweight increased with the neighborhood's deprivation 40,41,42 .
Some findings from previous studies further suggested that the neighborhood's socioeconomic environment may be related to the prevalence of excess weight through mediators related to the built environment 43 .Thus, economically privileged neighborhoods show a higher absolute number and density of commercial establishments such as supermarkets and fruit and vegetable stores, which in turn are associated with greater availability of, access to, and consumption of healthy foods and lower overweight and obesity rates 44 .
Interestingly, when the current study explored the contributions by the built environment to the availability of specialized stores for healthy food in the neighborhoods, no statistical associations with excess weight were observed.Studies in devel-oped countries have shown that characteristics of the food environment, such as the type and size of food stores, are important factors associated with neighborhood food quality in large cities, but contradictory results have been found when the outcome is excess weight 44,45,46,47 .This study is subject to some limitations.First, despite the use of a causal model, the cross-sectional design does not allow such conclusions.The second limitation, inherent to the methodological proposal due to its practicality in health surveillance systems, results from the self-reported measurement of the dependent variable.However this limitation is minimized by validation studies conducted with participants from the VIGITEL sample, which showed high correlation indices when the measurements were compared to self-reported values and good sensitivity and specificity 29 .A third limitation relates to the geo-referenced data used to describe the characteristics of the built and social environments, which are secondary data and thus subject to imprecision.An additional limitation is in the sample, consisting of persons living in households with landline telephones.Still, sampling weights were used in the data analysis to adjust the sample to the Brazilian population's social and demographic composition.
The evidence in this study showed that the characteristics of the built environment presented significant associations with excess weight, even after adjusting for gender, age, and conjugal status.These findings are preliminary and require further studies in other Brazilian cities to confirm the relevance of neighborhood environments in the causal web of the current obesity epidemic and to develop effective strategies for the prevention of obesity in the Brazilian context.
Medio Ambiente Controlado; Medio Ambiente y Salud Pública; Obesidad; Sobrepeso Contributors G. Velásquez-Meléndez and L. L. Mendes participated in the study's conceptualization and analysis and writing of the article.C. M. P. Padez collaborated in the critical revision of the intellectual content.

Table 1
Description of environmental variables.
CDL: Belo Horizonte Chamber of Commerce; CEASA: Minas Gerais Food Produce Central Warehouse; CINDS: Integrated Center for Social Defense Data; IBGE: Brazilian Institute of Geography and Statistics; Prodabel: Belo Horizonte Municipal Company for Information and Information Technology; SMS: Belo Horizonte Municipal Health Secretariat.* At the time of the study, the results of the 2010 Population Census had still not been published.

Table 2
Socio-demographic characteristics of the total sample and according to body mass index (BMI) for the city of Belo Horizonte, Minas Gerais State,Brazil, 2008 and 2009.

Table 3
Description of variables in the social environment and built environment by coverage area.

Table 4
Prevalence rates and crude and adjusted associations between variables from the built environment and social environment and excess weight.The above models are independent, i.e., a single model does not contain more than one environmental variable.