Sao Paulo Health Survey
This was the cross-sectional study that used the Sao Paulo Health Survey (ISA) dataset (2015). The final face-to-face interviews were conducted with 4,043 study individuals (12 years or older). More details can be obtained in Alves et al. .
ISA: Physical Activity and Environment
This study is part of the longitudinal project to verify the relationship between built environment, physical activity, and nutritional status in adults that living in Sao Paulo city, Brazil.
Weight and height variables were obtained by self-reported according to two questions: “What is your weight?”, whose response was given in kilograms and grams; and: “What is your height?”, with response options in meters and centimeters. There was calculated the body mass index (BMI) from the formula of weight in kilograms divided by the square of height in meters (Kg/m2), in order to classify their nutritional status according to WHO standards . Adults were considered as overweight (≥ 25.0 Kg/m2) and obese (≥ 30.0 Kg/m2) . The self-report data was validated in a previous study with the same population and the study showed acceptable results .
Places of the City Where People Lived
We used the classification of the health administration areas of the municipality government of Sao Paulo city . These areas are divided into five regions until 2015: East, South, Southeast, North, and Midwest .
The mix of destinations was characterized by the presence of the following items: a) bus stop; b) train and subway stations; c) parks; d) squares; e) public recreation centers; f) bike paths; g) primary health care units; h) supermarkets; i) bakeries; j) restaurants (food stores), and k) coffee-shops. These variables were calculated using a geographic information system (QGIS 2.14). We delineated radial buffers of 500 meters according to residential address. More details about the mix of destination score can be obtained in Florindo et al. .
Family Per Capita Income
The family per capita income using the total residence income divided by total the number of persons living in the same household was measured through the question “What was the average overall net household income last month?” and analyzed in quartiles: 1st quartile (lowest), second, third and fourth (highest).
The physical activity was evaluated by International Physical Activity Questionnaire long version. The score was calculated in minutes per week by sum of minutes in each domain: occupational, leisure, transportation and household [38-40]. We used the cutoff point of 150 minutes per week (0-149 minutes/week, ≥150 minutes/week).
Social and Demographic Variables
Age groups (18–29, 30–39, 40–49, 50–59, and 60 years or older), sex (male, female), level of education (incomplete elementary school, complete elementary to incomplete high school, complete high school, undergraduate incomplete to complete) and, length of living in the same residence (up to one year, between one and five years, >five years).
Chi-square test for overweight and obesity was calculated according to social and demographics variables, and health characteristics. Complex sample design according to the census tract (primary unit of the sample) in five health areas in Sao Paulo (strata), and the sample weight was used.
Two outcomes were used, 1) BMI in Kg/m2; and 2) Obesity (BMI ≥ 30.0 Kg/m2; yes or no). The place where people lived were the main independent variable and was based in five health administrative areas of the city: Midwest, North, Southeast, South, and East. Normality test was performed (Kolmogorov-Smirnov), since the data had no normal distribution, the mean was compared by the Kruskal-Wallis test and multiple post-hoc comparisons were performed using the Bonferroni test.
The modeling was undertaken took into account clustering by census tract and household in four stages: 1) Firstly, we conducted the analysis of BMI and obesity with places where people lived without adjust; 2) Secondly, we examined the analysis with adjustment for sex, age groups, education, length of living in the same residence and total physical activity; 3) Thirdly, we used all variables of the model 2 with mix of destinations 500 m buffers’ size and 4) Finally, we used all variables of the model 3 with family per capita income. We used the xtmixed command for linear models and the results were presented as beta coefficients (β) with 95% confidence intervals, and the xtmelogit command for logistic models and the results were presented as odds ratios (OR) with 95% confidence intervals.
All analyses were conducted in Stata software (Stata version SE 12.1, StataCorp).
The Ethics Committee of the School of Arts, Sciences, and Humanities at the University of Sao Paulo approved this study (process number 55846116.6.0000.5390).