Latin American Study of Nutrition and Health
The Latin American Study of Nutrition and Health (Estudio Latinoamericano de Nutrición y Salud; ELANS) is an eight-country (Argentina, Brazil, Chile, Colombia, Costa Rica, Ecuador, Peru, and Venezuela), observational epidemiological study using a common design and comparable methods across countries. The study uses a large representative sample (15-65 years old) from these eight countries and focuses on urban populations [20]. Data collection dates ranged from September 2014 to February 2015. The overarching ELANS protocol was approved by the Western Institutional Review Board (#20140605) and although the study is not a clinical trial, the protocol is registered at ClinicalTrials.gov (#NCT02226627) in order to increase the transparency of the methods and results and avoid criticism about publication bias. Ethical approval was obtained from each local institutional review boards and participants` informed consent/assent was obtained.
The entire ELANS study consisted of 9218 (4409 men) participants who were chosen using a random complex, multistage sampling frame with a random selection of Primary Sampling Units (PSUs) and Secondary Sampling Units (SSUs). The participants were recruited from PSUs areas (e.g., counties, municipalities, neighborhoods, residential areas) within each selected city in each country. An “n” size proportional to population weight was used for the selection of PSUs. In this case, a simple random sampling of “n” with replacement was performed to adhere to the principle of statistical independence of the selection of the areas included in the PSU sample. For these random selections, the probability proportional to size method was applied. Thus, within each of the areas included in the PSU distribution, a representative sample of SSUs was randomly designated using the probability proportional to size method.
For the selection of households, we implemented a four-step, systematic randomization procedure by establishing a selection interval (k): a) the total urban population was used to proportionally describe the main regions and to select cities representing each region; b) the sampling points (survey tracts) of each city were randomly designated, and c) clusters of households were selected from each sampling unit; d) the designated respondent within each household was selected using the birthday method. Details about participant sampling and recruitment strategies have been published elsewhere [20, 21].
In each country, stratified recruitment of individuals was done across sex, age group (15-19.9, 20-34.9, 35-49.9, and 50-65 years), and socioeconomic level (low, medium, high). Socioeconomic levels were weighted according to the national indices of each country. The number of individuals required per socioeconomic level has been addressed in more detail elsewhere [20]. In total, 92 cities participated in this study (7 to 23 cities in each country) [22]. The required sample size was calculated using a power analysis with a 95% confidence level, a maximum error of 3.5%, and a survey design effect of 1.75, resulting in a required sample size of 9090.
For household selection within cities, the systematic randomization method was used. The exclusion criteria adopted were: a) pregnant and lactating women; b) persons with physical or mental disabilities; c) unsigned consent form; d) individuals living in non-family residential environments; and e) individuals who could not read. More information on the ELANS study is provided in Fisberg et al. [20]
The perceived neighborhood built environment and PA protocol used in ELANS includes self-reported data collected by questionnaires. In this study, the questionnaires (perceived neighborhood built environment and PA) were interviewer-administered during the home visit, and 8185 (18-65 years old) participants had complete data. We excluded adolescents (15 to 17 years old) from the analyses because the ELANS study did not include adolescents of all ages. Also, adolescents may have restricted independent mobility [23] that may yield different environment-PA associations than those observed in adults. In addition, PA guidelines for adolescents differ from those for adults [24].
Perceived neighborhood built environment
To assess perceptions of the neighborhood built environment characteristics, the Neighborhood Environment Walkability Scale - Abbreviated (NEWS-A) [25] previously translated into Spanish and adapted for use in Latin America [14, 26]. Apart from translating the NEWS-A from English into the local language of the participating countries, scale adaptation also encompassed the addition of two items to the safety from crime subscale, an item measuring the proximity of shopping centers and three items gauging the proximity to three types of public open space (metropolitan parks, playgrounds and public squares) as shown in Table 1. The reliability and validity of NEWS-A have been evaluated in several countries with all included scales having test-retest reliability intraclass correlations >0.50 [27, 28].
The following characteristics were assessed: land use mix–diversity, land use mix-access, street connectivity, walking/cycling facilities, aesthetics, safety from traffic, and safety from crime. The land use mix-diversity scale is assessed by the perceived walking proximity from home to 23 different types of destinations, with responses ranging from 1-5-min walking distance (coded as 5, indicative of high walkability) to >30-min walking distance (coded as 1, indicative of low walkability). The remaining six scales are average ratings of items answered on a four-point Likert scale (1=strongly disagree to 4=strongly agree). Scales were scored in a direction consistent with higher walkability and safety, with individual items reversed when necessary. The scales ‘street connectivity’ and ‘safety from traffic’ from NEWS-A were not included in the results due to low internal consistency. Instead, the individual items were analysed separately (Table 1).
Physical activity assessment
Participants reported their PA levels by completing the long-form of the last 7 days, interview version of the International Physical Activity Questionnaire (IPAQ) in Spanish [29]. We adapted the IPAQ by using only the questions that covered the active transport-related and leisure-time domains [22]. The long-form IPAQ (last 7 days) has been validated internationally using CSA accelerometer (model 7164) to assess total PA in a variety of contexts (occupational, transport, household, leisure) and at different intensities (moderate, vigorous, walking, cycling) in individuals aged 18–65 years from twelve countries with Spearman's correlation coefficients ranging from 0.46 to 0.96 [29].
The participants were instructed to report the frequency and duration (bouts of >10 minutes) of PA in the domains of active transport and leisure. Specifically, the following questions were asked: a) “During the last 7 days, did you walk or use a bicycle (pedal cycle) for at least 10 minutes continuously to get to and from places?” (Yes, No); b) “During the last 7 days, on how many days did you walk or ride a bicycle for at least 10 minutes at a time to go from place to place?”; c) “How much time did you usually spend on one of those days to bicycle or walk from place to place?” These questions were asked separately for walking and cycling. Concerning leisure PA, the following questions were asked: a) “During the last 7 days, did you walk, or do any moderate or vigorous PA for at least 10 minutes continuously?” (Yes, No); b) “During the last 7 days, on how many days did you walk, or do moderate or vigorous PA for at least 10 minutes at a time in your leisure time?”; c) “How much time do you spend walking or doing moderate-to-vigorou PA in your leisure time?” Questions were asked separately for walking, moderate-intensity, and vigorous-intensity activities. Details on the assessment of PA by IPAQ have been published elsewhere [22].
IPAQ PA data are reported as min/week of walking, moderate and vigorous PA during leisure-time, and min/week of walking and cycling for transport-related purposes. Time (min/week) spent in each of the PA domains (i.e., transport-related and leisure-time) was calculated and used in analysis. In this study, we used transport-related PA (walking + bicycle) and leisure-time PA (walking + moderate + vigorous) separately.
An international comparison showed that IPAQ had comparable reliability and validity to other self-report PA assessment methods [29]. There is evidence for an acceptable degree of reliability and validity of the transport-related and leisure-time PA items of the IPAQ. For these items, intraclass correlation coefficients ranging from 0.42 to 0.75 have been reported [30, 31]. Also, moderate correlations (0.50–0.63) were found between diary measures of transport-related and leisure-time PA and the corresponding IPAQ items. In our study, the intraclass correlation coefficients of self-reported total weekly minutes of walking, transport-related, and leisure-time PA with accelerometry-estimated weekly minutes of moderate-intensity PA were between 0.24 and 0.35 (unpublished data).
Statistical analysis
Statistical analyses were carried out with the software SPSS v.26. Means, standard deviations (SD), median (interquartile range: IQR), and percentages were computed, as appropriate, to describe the variables. Weighting was done according to sociodemographic characteristics, sex, socioeconomic level, and country [20].
Cronbach’s alpha was used to assess the internal consistency of the neighborhood environment characteristics’ scales. Because the PA variables were positively skewed and had a large number of zeros, two different multilevel regression models, with the country as the second level, were used to estimate the associations of neighborhood characteristics with PA (transport-related and leisure-time): a logistic regression model (odds ratio: OR; confidence interval 95%: 95%CI) with a binary dependent variable (0 = “<10 minutes of PA/week”, 1 = “≥10 minutes of PA/week”) followed by a linear model with the min/week of PA as the dependent variable. The linear regression model (β; 95%CI) was estimated using data from the respondents who reported ≥10 minutes of PA per week. Due to the non-normality of data, the minutes of PA were log-transformed for the linear models and the unstandardized coefficient values were back-transformed into min/week aiming to increase the clinical utility of the findings. Both models were adjusted for sex, age, socioeconomic level, and country. We present the overall (i.e., pooled) and country-specific results (Additional file 1: Table S1-S4). A significance level of 5% was adopted.