Sao Paulo Health Survey (ISA)
This study used data from the Sao Paulo Health Survey. Data collection was completed in 2015, with 4,043 adult participants who lived in five health administrative areas in Sao Paulo city. The sampling process has been described in more detail elsewhere[23]. Georeferencing resulted in 3,145 participants having their residential address geocoded: [18]. More details can be obtained from other publications[18, 19, 24]. This dataset will be used such as the baseline of the longitudinal study denominated of “ISA: Physical Activity and Environment” that will have such as objective to verify the relationship between the built environment and physical activity in adults from Sao Paulo city, Brazil.
Sedentary Behavior
Sedentary behavior data were collected using the International Physical Activity Questionnaire (IPAQ)[25] and measured on the basis of two questions: 1) Total sitting time on a usual weekday; 2) Total sitting time on a usual weekend day. The questions asked about minutes per day that people spent sitting in all domains of life (work, leisure, transportation, and household). For analysis, two types of outcomes were used: 1) Continuous measures of minutes of sitting time in a typical weekday and in a typical weekend day; and 2) categorical measures which employed a cut-off point of four hours per day to classify sedentary behavior in a typical weekday and in a weekend day. The basis for this cut-off point was a meta-analysis of the relationship between sedentary behavior, physical activity, and mortality in more than one million people in a second Lancet series about physical activity and health[4].
Mix of Destinations
Walkable destinations within each participant’s residential catchment were captured using georeferencing procedures [18, 19] applied to publicly available datasets, and included: transportation (bus stops, train and subway stations), green areas (parks, squares), public recreation centres, bike paths, and primary health care units. The dataset for these items pertain to places in 2016 and were obtained mainly from the open site GEOSAMPA <http://geosampa.prefeitura.sp.gov.br/PaginasPublicas/_SBC.aspx>. In addition, we captured supermarkets, food stores, bakeries and coffee shops using the Health Surveillance Registration database from Sao Paulo city associated with National Economic Activity Classification in November 2016. We calculated the number of these destinations within a 500m radial buffer of each participant’s home address: we used this distance because previous studies conducted in Sau Paulo city with the same sample found significant associations between the built environment and walking [18, 19].
The eight destination-types were operationalized in two ways. First, we calculated the median number of destinations for the sample [median=3 and range 0 to 8] and grouped participants into two categories using a median-split (1=above median, 0=at or below median). Second, the distribution of destinations was divided into tertiles (high, medium, low): a similar measure using data from the Sao Paulo Health Survey found that participants who lived in areas with a greater mix of destinations were significantly more likely to walk for transport [19].
Covariates
We used age (18-29 years, 30-39 years, 40-49 years, 50-59 years, 60 years or more), education (in four categories for logistic models: incomplete elementary school, incomplete high school, complete high school, incomplete undergraduate or above), marital status (singles, married/with partners, separated/widowers), obesity (in two categories: BMI <30 kg/m2: no or yes ≥ 30 kg/m2), physical activity (evaluated by IPAQ and we used the cut-off point of 150 minutes per week: yes or no according to work, household, leisure, and transportation activities), diseases (self-report of diseases diagnosed by physicians - arterial hypertension; diabetes; myocardial infarction; cardiac arrhythmia; other heart disease; cancer; arthritis, rheumatism or arthrosis; osteoporosis; asthma or asthmatic bronchitis; emphysema, chronic bronchitis or chronical obstructive pulmonary diseases; rhinitis; chronic sinusitis; other lung disease; tendonitis, repetitive strain injury or work-related musculoskeletal disorders; cerebral vascular accident or stroke; spine disease or spine problem: presence or not of diseases), smoking (yes or no), car or motorcycle ownership (yes or no); time living in the same residence (in three levels : <1 year, ≥1 year and <5 years, >5 years), and place where people living in Sao Paulo city (North, South, Midwest, Southeast, and East). These variables were selected based on the findings of systematic reviews [9, 12, 13], and also because they were found to be important variables in other studies that examined the relationship between the built environment and physical activity in the same sample [22-24].
Statistical Analysis
The analysis uses four outcomes variables: 1) sitting time during a typical weekday in minutes; 2) sitting time during a typical weekend day in minutes; 3) four hours or more of sedentary behavior in a typical weekday (yes or not); and 4) four hours or more of sedentary behavior in a typical weekend day (yes or not). We used bivariate descriptive analyses to examine how sitting time was distributed by each of the environmental, sociodemographic, and health-related variables. Sample weights were calculated considering the survey design effect, post-stratification by sex and age, and the non-response rate. We used the chi-square (Wald test p-value) to identify associations between categorical variables, and linear regression to identify differences between quantitative variables.
Linear and logistic multilevel analysis of the association between mix of destinations and sitting time were conducted in two stages, firstly without adjustment, and secondly, with adjustment for those factors found to be statistically significant in the bivariate analysis: the covariates used for the adjusted models are shown in Table 1. The multilevel analysis accounted for clustering within census-tract and household.
For multilevel analysis, we examined the relationship between the outcomes with the mix of destinations scores. Firstly, we examined the relationship without adjustment. Secondly, we added demographics, social, health, and environmental variables that had significant association in bivariate analysis. The places where people lived and time living in the same residence were added regardless of statistical significance because these variables are important for adjusting the analysis between environmental and behavior variables. We used both linear and logistic regression models. We used age, education, marital status , obesity , physical activity , diseases
We opted to run separate models for males and females [9, 13]. We used the xtmixed command for linear models and the results are presented as beta coefficients (β) with 95% confidence intervals, and the xtmelogit command for logistic models and the results are presented as odds ratios (OR) with 95% confidence intervals. All analyses were conducted in Stata software (Stata version SE 12.1, StataCorp).
Ethics Approval
The Ethics Committee of the School of Arts, Sciences, and Humanities at the University of Sao Paulo approved the study (process number 55846116.6.0000.5390).