Characteristics of the environmental microscale and walking and bicycling for transportation among adults in Curitiba , Paraná State , Brazil

1 Programa de Pós-graduação em Educação Física, Universidade Federal do Paraná, Curitiba, Brasil. 2 Programa de Pós-graduação em Educação Física, Universidade Tecnológica Federal do Paraná, Curitiba, Brasil. 3 Brown School, Washington University in St. Louis, St. Louis, U.S.A. 4 Grupo de Pesquisa em Atividade Física e Qualidade de Vida, Pontifícia Universidade Católica do Paraná, Curitiba, Brasil. ARTIGO ARTICLE


Introduction
Physical inactivity is a leading cause of death worldwide, with serious health problems, notably in low and middle-income countries 1 .This can be partly attributed to the increase in motor vehicles, chaotic urbanization, and precarious public safety and security that affect the way people commute on a daily basis 2,3,4 .These and other characteristics have been widely studied at different scales of influence by the built environment in cities (micro, meso, and macro), especially the scales related to transportationrelated physical activity 4,5,6,7 .Such characteristics include land use, street layout, presence and quality of sidewalks and bike lanes, access to public transportation, aesthetics, and public safety and security 4,6,7,8 .
Evidence points to an association between variables in the perceived 5,9,10 and built environment 5,9,11,12 and transportation-related physical activity.Other studies have shown the positive effect of environmental changes on commuting behavior 12,13,14,15 .However, a large share of the evidence comes from studies in high-income countries, which may not represent the urban and sociocultural characteristics of lower-income countries like those of Latin America 3,6,7 .
In Brazil, different methodologies have been used to explore the association between variables from the built environment, such as "walkability" and transportation-related physical activity in adults 3,7,16,17,18,19 , while several characteristics vary according to the neighborhood's socioeconomic level 20,21 .However, a recent review failed to identify studies that used systematic observation of environment for this evaluation in adults 19 .Thus, the lack of precise and detailed information on the variables in the built and social microscale raises an important question for investigation in Brazilian cities 8,22 .
Certain objective measures can thus contribute to a better understanding of the relationship between cities' structure and physically active commuting 23,24 .Data from the environmental microscale allow a detailed description of conditions in the quality of sidewalks, aesthetics, access to public transportation, lighting, and others that can be modified at low cost and in less time when compared to macroscale attributes, generally identified through geoprocessing data 5,8 .
Streetscape auditing is thus a potential option for obtaining microscale data, due to its low costs and the potential for producing precise details on urban variables 5,8 .The objectivity of the measures requires clear protocols, comprehensive auditing items, and the possibility of capturing the environment's characteristics, which are dynamic, portraying the population's daily reality and potentially affecting individual choices as to mode of transportation 25,26 .The aim of this study was thus to analyze the association between the characteristics of the built and social environmental microscale and walking and bicycling as modes of transportation for adults in Curitiba, Paraná State, Brazil.

Study design, location, and ethical issues
This was a cross-sectional observational study using a household survey.The data are part of a research project in Curitiba, capital of Paraná State, in 2009.The projectʼs principal objective was to assess the health characteristics, leisure-time habits, and physical activities of adults residing near the cityʼs parks.The study was approved by the Institutional Review Board of the Federal University of Pelotas (case review 005/2008).

Selection of sites and households
In order to enhance the variability of socioeconomic status and characteristics of the built environment, eight urban parks were selected with different structures for physical activities, located in different areas of the city.The neighborhoods were classified in four groups: (1) high environmental quality and high socioeconomic status, (2) high environmental quality and low socioeconomic status, (3) low environmental quality and high socioeconomic status, and (4) low environmental quality and low socioeconomic status.Further details on the classification and selection criteria for the sites have been published elsewhere 27,28 .
Cad. Saúde Pública 2018; 34(1):e00203116 The households were selected from a 500-meter buffer around each of the eight parks, generated with the ArcGIS software (http://www.esri.com/software/arcgis/index.html).This distance has been adopted in similar studies 23 and considers a walking distance of five to 10 minutes from home to the park 27 .All street segments inside this radius (n = 1,899) were assessed to identify those with eligible households.One household was selected per segment, based on a random numbers table generated with the EpiInfo software (Centers for Disease Control and Prevention, Atlanta, USA).In all, 1,538 street segments with households were tallied.However, 361 (29%) did not have happen to have households and were thus excluded from the analysis.

Selection of participants and data collection
One resident per household was randomly selected based on the inclusion criteria of adult age (≥ 18 years), no physical limitations, and having lived in the neighborhood for at least a year.Interviews were conducted in 95% of the eligible segments (n = 1,461), and in 5% of the remaining segments there were no eligible residents.The refusal rate was 8% (n = 121), and quality control was performed in 13% of the interviews via telephone contact to verify the date and time of the interview and to confirm some key study variables.

Dependent variables
Walking and bicycling for transportation during a normal week were assessed with the International Physical Activity Questionnaire (IPAQ), long version 29 .Variables were calculated as the ratio between weekly frequency and mean daily amount for each activity.Based on the literature 23 , walking was operationalized as two different outcomes according to weekly amount ("≥ 10min/week" and "≥ 150min/ week"), while bicycling was categorized as ≥ 10min/week.

Independent variables
Objective evaluation of the environment used an instrument for systematic observation of the streetscape.The Inventory for the Evaluation of the Community Environment Related to Physical Activity (ICAF in Portuguese) was translated from the Active Neighborhood Checklist 26 and adapted to the Brazilian context, showing an inter-evaluator agreement of 85-98% 30 .
The items comprise the community environmental microscale, and the instrument consists of various characteristics, grouped in six dimensions: land use, public transportation, streetscape, conditions and aesthetics, places for walking and cycling, and social environment.A total of 105 items were assessed, and variables were grouped according to the dimensions of the built and social environment suggested in the original instrument, which considers the quantity, variety, and quality of the attributes related to physical activities 30 .
In the "land use" dimension, we evaluated occupation of the lots (45 items), "public transportation" included taxi stands, bus stops, and Bus Rapid Transit on one or both sides of the street segment (3 items)."Streetscape" assessed structures for slowing traffic or facilitating safe street crossings for pedestrians, such as speed limits and pedestrian lanes and signs (14 items).The "conditions and aesthetics" dimension included public improvements on the street segment, trash cans, and benches, versus signs of vandalism and graffiti (14 items).The dimension "places for walking and cycling" identified characteristics of the street and sidewalk that could hinder or facilitate walking or bicycling, such as obstacles, marked lanes, and signs, plus width of the sidewalks (sidewalks: 10 items; streets: 5 items).Finally, the "social environment" dimension assessed positive aspects like police presence, people engaged in physical activities and/ or conversing on the streets, versus negative characteristics such as people arguing or fighting, stray dogs and other animals, panhandlers, illegal parking attendants, and drunkards (12 items).The presence of items with a potential negative association with walking and cycling was recoded from "1" to "0" (e.g., dead-end streets, etc.) In order to characterize the environment of the street segments, the attributes for each dimension were tallied and classified at three levels of quality by tertiles: low (1 st tertile), medium (2 nd tertile), and high (3 rd tertile).However, the items public transportation and places for walking and bicycling on streets showed low frequency of attributes, so we opted to analyze them dichotomously (0 items versus ≥ 1 items and 0-2 items versus ≥ 3 items, respectively).

• Individual variables
Based on the conceptual model proposed by Saelens et al. 9 , some individual variables were identified that might confound the association between the variables in the built and social environment and the target outcomes: sex, age bracket, marital status, socioeconomic status, nutritional status, self-rated health, self-rated quality of life, perceived crime in the neighborhood, number of motor vehicles in the household, and use of public transportation.For example, female gender is positively associated with walking ≥ 10min/week, but inversely associated with bicycling as transportation 23 .Meanwhile, vehicle ownership is inversely associated with the three outcomes analyzed in the current study 23 .These and other variables were identified, measured, and tested as possible covariates.The description of these measures and the way they are categorized are discussed next.
Sex (male, female) was recorded, age bracket was classified in three categories (18-39.9years, 40-59.9years, ≥ 60 years), and marital status categorized as single or married.Socioeconomic status was assessed using the questionnaire of the Brazilian Association of Market Research Companies (ABEP) 31 , and individuals were classified in three levels: low (classes C, D, and E), medium (class B), and high (class A).
Nutritional status was obtained from self-reported weight and height, and participants were classified as "normal weight" and "excess weight", according to body mass index.Self-rated health and quality of life were assessed with the World Health Organization Quality of Life (WHOQOL) scale, with answers according to a five-point Likert scale 32 .Self-rated health was assessed with the question, "Are you satisfied with your health?" (very dissatisfied, dissatisfied, neither satisfied nor dissatisfied, satisfied, very satisfied).For analytical purposes, the first three categories were grouped and operationalized as "negative self-rated health" and the other two as "positive self-rated health".Self-rated quality of life was assessed with the question, "How do you rate your quality of life?" (very bad, bad, neither bad nor good, good, and very good).The first three categories were operationalized as "negative self-rated quality of life" and the other two as "positive self-rated quality of life".Finally, perceived crime was assessed with the question, "Are there many crimes in your neighborhood?",with the dichotomous (no, yes) answer from the Neighborhood Environment Walkability Scale (NEWS) 33 .
The number of motor vehicles in the household was assessed with the ABEP questionnaire 31 and operationalized in three categories (0, 1, and ≥ 2).Use of public transportation was assessed with the question: "How many days a week do you use the city's public transportation?"(0 to 7 days), and the variable was operationalized in three categories (0 days, 1-2 days/week, and ≥ 3 days/week).The principal means of daily transportation was assessed to characterize the participants, who could choose one of eight possible answers: car, public bus, walking, motorcycle, bicycle, private (company) bus, taxi, and other.

Data analysis
Multilevel Poisson regression was used to test the association between the variables of the built and social environment and walking and bicycling for transportation.Since the study design took into account the sample selection considering the eight parks as the primary sampling units, multilevel random intercept modeling was used to consider the cluster effect between the sampling units.The variance partition coefficient (VPC) was calculated for each combination of outcome and exposure, indicating the values of the constant and the proportion of variance attributed to the place level and the variation between individuals from the same location.
We first tested the association between the individual variables and the study's dependent variables.This allowed identifying variables that were potentially associated with each outcome (p < 0.20) and that could be kept for analysis in the adjusted model.To construct the final multilevel regression model, we analyzed the association between the characteristics of the built and social environment and the transportation variables, adjusted for individual variables that presented p < 0.20.The analyses were performed with the Stata 12.0 software (StataCorp LP, College Station, USA), using the xtmepoisson command, with significance kept at 5%.

Results
Of the 1,461 interviewees, 42 were excluded due to lack of complete data on some target variable, resulting in a final sample of 1,419 individuals (Table 1).The majority of the participants were female (63.6%), 40-59 years of age (45.8%), married (57.4%), with medium socioeconomic status (49.8%), with normal nutritional status (51.7%), and with positive self-rated health (71.6%) and quality of life (73.9%),and with low perception of crimes in neighborhood (53.3%).In addition, 62.4% of the participants reported using public transportation and 76.2% had at least one motor vehicle in the household.The principal means of daily transportation was by car (46.7%), followed by public bus (34.3%) and walking (12.9%) (Table 1).
As for the variables in the built and social environment, the highest proportion of participants lived on street segments with medium-quality land use (41.2%), public transportation (84.5%), low-quality streetscape (60.4%), low aesthetic quality (44.7%), few favorable items on places for walking and bicycling on the streets (0-2 items: 79.6%), and medium quality of the social environment (40.5%).In addition, 69.1% of participants spent less time walking for transportation (≥ 10min/week), while 20.3% walked for transportation at the recommended levels (150min/week), and only 9% reported bicycling for transportation at low levels (≥ 10min/week) (Table 1).
Table 2 shows the association between individual variables and walking and bicycling.Socioeconomic status (p < 0.05) and number of motor vehicles (p < 0.001) were inversely associated with walking ≥ 10min/week, while weekly use of public transportation was positively associated (p < 0.001) and married marital status was inversely and potentially associated with this outcome (p < 0.20).Socioeconomic status, number of motor vehicles, and rate of weekly use of public transportation were associated with walking ≥ 150min/week (p < 0.001), while age bracket and marital status were potentially associated with this outcome (p < 0.20).Sex, age bracket, self-rated health, perceived crime in the neighborhood, and rate of weekly use of public transportation showed significant association with bicycling (p < 0.05), while self-rated quality of life and number of motor vehicles in the household were potentially associated with this outcome (p < 0.20).
In the bivariate association between the environmental variables and the outcomes (Table 3), medium streetscape was inversely associated with walking ≥ 150min/week (prevalence ratio -PR = 0.62; 95%CI: 0.41-0.92)and bicycling (PR = 0.53; 95%CI: 0.29-0.97).No significant values were observed in the test for trend among the categories of target variables.

Discussion
This is the first study in Brazil to explore the association between characteristics of the built and social environment, obtained by systematic and direct observation of the environment, and walking and bicycling for transportation in adults.The methodology allowed a geographic representation of the places in neighborhoods with different environmental and social attributes for physical activity, besides obtaining and objectively measuring quantitative and qualitative attributes of the built and social environment that are not possible to identify in data obtained with geoprocessing or perception of the environment 24 .
These are thus the current study's strengths and innovative characteristics.Streetscape was the only score inversely associated with walking greater than 150 min/week and bicycling.A recent study in Recife, Pernambuco State, Brazil 30 , evaluated the same attributes of the built and social environment around schools and found that public transportation, social environment, and overall environmental score were related to active commuting to school by preschool children (3-5 years).Two studies tested the association between the environment's characteristics (by systematic observation) and walking and bicycling in adults 8,25 .However, these studies were done in cities in the United States, where the characteristics of the environment differ greatly in comparison to Brazilian cities 34 .For example, a multicenter study in 14 cities in 10 countries showed that the density of street intersections and mixed land use were greater in Curitiba, than in cities in New Zealand and the United States 34 .These variables, among others, are important predictors of walking and bicycling for transportation 5,9,34 .
Cad. Saúde Pública 2018; 34(1):e00203116   Medium streetscape reduced by 41% and 46%, respectively, the likelihood of walking ≥ 150 min/ week and bicycling.These results differ from a study in the US cities of Seattle, San Diego, Baltimore, and Washington DC, where the overall streetscape score was positively associated with weekly frequency of active commuting (walking + bicycling) 34 .The inverse association shown in the current study may have several explanations.First, the majority of the streets in the selected neighborhoods show low commercial density, so residents do not need to walk or pedal to local shops on a daily basis 9 .In fact, in Curitiba a positive association was found between the proportion of business districts close to the household and walking for transportation 23 .Second, proximity to parks may be associated with use of these places for leisure-time physical exercise only 27,35 , and not for transportation.Third, since parks may pose natural barriers to "short" errands on foot or bicycle to neighborhood shops and services, the high household vehicle ownership rate (76%) may favor the use of vehicles for these errands 28 .In fact, some 50% of the interviewees reported using their car or motorcycle as their main means of daily transportation (Table 1).Finally, although the medium "streetscape" score indicates street segments with better characteristics which may facilitate walking (pedestrian crossings and lights, Cad. Saúde Pública 2018; 34(1):e00203116 the environment such as bus stops could be positively associated with less walking per week (≥ 10min/ week), since the distance to bus stops is short (average of 175 meters) 23 .The sample does not represent the city's entire adult population, since it was limited to individuals living in the vicinity of eight parks with the potential for recreational physical activities 27 .Some studies have suggested that the presence of parks in the neighborhood is associated with greater appreciation of the surrounding areas, which would in turn lead to better characteristics in the built environment and little variability in the indicators between places 39 .The scores on the items from the environment were calculated from the measurement of the variables observed on the street segments of the participants' households, so it was not possible to calculate an "environmental quality " score within an area that the individual might be exposed to or cross when walking or pedaling, for example a circular buffer or street network buffer (including sausage buffer and detailed trimmed buffer, both with 25cm or 75cm radius on either side of street and; detailed biffer) measuring 300, 500, or 1,000 meters 40 .Finally, the cross-sectional design limits the causal interpretation between variables.

Conclusion
Streetscape was inversely associated with walking and bicycling as modes of transportation.
Future experimental studies should be conducted in lower-income countries like Brazil to test the effects of environmental changes on walking and bicycling for transportation in adults.In addition, triangulation of methods in the same study (evaluation of the perception of the built environment, geoprocessing data, systematic observation of the built environment, use of Global Positioning System equipment and focus group interviews) can also add to the understanding of the relationship between the built environment and transportation-related physical activity in adults.

Contributors
A. A. S. Lopes participated in the data collection, initial study conception, data analysis, literature review, and writing and critical revision of the article in all its stages.M. Kienteka, participated in the data collection, initial study conception, and first draft of the article.R. C. Fermino participated in the data collection, initial study conception, literature review, and writing and critical revision of the article in all its stages.R. S. Reis was responsible for the research project's conception and coordination and participated in the critical revision of the article.

Table 1
Description of individual and environmental variables, transportation-related physical activity, and means of transportation in adults.Curitiba, Paraná State, Brazil, 2009 (N = 1,419).

Table 1 (continued)
Description of individual and environmental variables, transportation-related physical activity, and means of transportation in adults.Curitiba, Paraná State, Brazil, 2009 (N = 1,419).

Table 2
Multilevel bivariate association between individual variables and walking and bicycling for transportation in adults.Curitiba, Paraná State, Brazil,