The world is rapidly urbanizing, which may influence physical activity (PA) levels - although little evidence is available for low- and middle-income countries. We evaluated associations between urbanization and total PA, as well as work-, leisure-, home-, and transport-specific PA, for 138,206 adults (35-70 years) living in 698 communities across 22 countries within the Prospective Urban and Rural Epidemiology (PURE) study. The 1-week total PA long-form International PA Questionnaire (IPAQ) was administered at baseline (2003-2015) and we used satellite-derived population density and impervious surface area to quantify levels of urbanization for 5 and 10 years prior to PA measurements. Generalized linear mixed effects models were used to examine associations between urbanization measures and PA, controlling for individual, household and community factors. Higher community baseline levels of population density (-12.4%, 95% CI: -16.0%, -8.7% per IQR) and impervious surface area (-29.2%, 95% CI: -37.5%, -19.7% per IQR), as well as 5-year population density change (-17.2%, 95% CI: -25.7%, -7.7 per IQR) was associated with lower total PA. Important differences in the associations between urbanization metrics and PA were observed between PA domains, country income levels, urban and rural status, and gender. These findings provide new information on the complex associations between urbanization and PA.
Article
Urbanization and Physical Activity in the Global Prospective Urban and Rural Epidemiology Study
https://doi.org/10.21203/rs.3.rs-1538322/v1
This work is licensed under a CC BY 4.0 License
Journal Publication
published 06 Jan, 2023
You are reading this latest preprint version
Approximately 55% of the world’s population live in urban areas. This is expected to increase to 68% (approximately 7 billion people) by 20501. Most of this increase, up to 90%, will occur in Asia and Africa1. Trends in urbanization are altering the landscape of human settlement, with significant implications for living conditions, lifestyles, work conditions, and the environment in different parts of the world2. In particular, urbanization may influence physical activity (PA) levels3,4.
PA is one of the most important components of successful health promotion and disease prevention5–8. Physical inactivity causes more than 3 million deaths per year, of which 2.6 million occur in low- and middle-income countries (LMICs)9. Despite the health benefits associated with being physically active10, levels of PA among adults have been decreasing over the last decades11. Globally, only 1 in 4 adults achieve current WHO guidelines for PA (at least 150–300 minutes of moderate-intensity aerobic PA per week)12,13. This ranges from a high of 39% in the Americas to a low of 22% in African regions11. The global variability in urbanization and PA has been examined at an ecological level14–17, but few studies have used individual-level data to examine the association between urbanization and PA, especially in LMICs3,4,
Urbanization is associated with substantial lifestyle changes that may impact total PA, but specific changes will likely vary by different PA domains. Major PA domains include transport, leisure/recreational, household, and occupational18,19, and the proportion of PA from each domain varies greatly between countries and geographic regions20. For example, PA in LMICs is most often accumulated in occupational, household, and transport domains, whereas leisure-time PA contributes more to total PA in high-income countries (HICs)20. These differences may be due partly to urbanization; for example, changes in work settings and a transition from labor intensive jobs to more sedentary work reduces occupational PA21. Similarly, urbanization and changes in home environments and amenities will reduce household PA22. Urbanization also brings changes in the built environment that can influence recreational and transport PA23–25. Given the potential impact of urbanization as an upstream driving factor of PA26,27,there is surprisingly little research that examines the association between urbanization and individual components of PA in LMICs3,4,17, countries where the most rapid urbanization is currently occuring28,29. The few studies that have been done in HICs have mostly evaluated the influence of urbanization on leisure/recreational and transport PA, with conflicting results4,30.
The objective of this research is to determine how urbanization is associated with total and domain specific PA across 698 diverse communities in 22 low-, middle-, and high-income countries using individual data on 138,206 adults in the Prospective Urban Rural Epidemiology (PURE) study. This analysis will provide new information on how urbanization is associated with total and domain specific PA, and how these relationships vary between country-income status and by individual socio-demographic characteristics. Understanding how PA is related to urbanization is central to reducing population level physical inactivity, especially among rapidly developing and urbanizing LMICs.
We restricted our analyses to the 138,206 PURE participants who had complete PA and urbanization measures (Table 1). The mean age of all participants was 50.5 years, 58.7% of participants were females, 57.4% lived in urban areas, and 11.8% of participants were from HICs, 75.0% from MICs and 13.2% from LICs. The mean total PA level was 4,141 MET minutes per week (sd = 4984), varying from 4,480 MET minutes per week for the lowest quartile (Q1) of population density change to 3,825 MET minutes per week for the highest quartile (Q4) of population density change. (Table 2)
| Characteristic | N (%) |
|---|---|
| Participants, n (%) | 138206 |
| Age, mean (SD) | 50.5 (9.8) |
| Female, n (%) | 81051 (58.7) |
| Education, n (%) | |
| None, primary, or unknown | 55809 (40.4) |
| Secondary | 54019 (39.1) |
| Trade, college, or university | 28128 (20.4) |
| Household wealth index, n (%) | |
| Low | 42181 (30.5) |
| Middle | 46784 (33.9) |
| High | 47665 (34.5) |
| BMI | |
| <20 | 10572 (7.7) |
| 20–30 | 94936 (68.7) |
| > 30 | 24589 (17.8) |
| Missing | 8109 (5.9) |
| Chronic conditions at baseline | |
| No | 110674 (80.1) |
| Yes | 27532 (19.9) |
| Urban/Rural Status, n (%) | |
| Urban | 69347 (50.2) |
| Rural | 68859 (49.8) |
| Country Income Level, n (%) | |
| High | 16302 (11.8) |
| Middle | 103666 (75.0) |
| Low | 18238 (13.2) |
| 5-Year Community Population Density Change | |||||
|---|---|---|---|---|---|
| All Participants | Quartile 1 | Quartile 2 | Quartile 3 | Quartile 4 | |
| No. Individuals | 138206 | 36742 | 33170 | 34240 | 34054 |
| Total PA, MET × min per week, mean (SD) | 4140 (4984) | 4480 (5035) | 4259 (4883) | 3975 (4820) | 3825 (5156) |
| Transport PA | 629 (1237) | 597 (1210) | 640 (1194) | 689 (1259) | 596 (1281) |
| Recreational PA | 511 (1140) | 453(1046) | 573 (1123) | 552 (1118) | 472 (1263) |
| Household PA | 1469 (2367) | 1613 (2286) | 1317 (2195) | 1509 (2465) | 1424 (2502) |
| Occupational PA | 1533 (3983) | 1819 (4169) | 1729 (3997) | 1229 (3650) | 1338 (4049) |
Distribution of Urbanization Metrics
The distribution of urbanization metrics for PURE participants is shown in Fig. 1, stratified by country income level. Due to outliers and the highly right-skewed distribution for impervious area change we present log-transformed measures. The baseline median (IQR) population density and percent impervious area for PURE communities was 802 people/sq km (4428) and 35% (73) respectively. Over a 5-year period prior to baseline, the median (IQR) annual population density and impervious surface area change rate were 2% (1.5) and 0.4% (2) per year, respectively. For each urbanization measure, we observed a range of baseline levels and change rates by HIC, MIC and LICs, indicating that urbanization is occurring in all country income classes but is greatest in LMICs. There were varying levels of correlation between our different urbanization metrics and spatial and temporal scales of analyses (Appendix, Table S1. For different buffer areas around community centroids (1 and 5 km’s), and for different timescales of change (5 and 10 years) there were high correlations for all metrics (all r > 0.89 and > 0.91, respectively). For this reason, we only present 1 km and 5-year urbanization metrics below.
Associations between Population Density and Physical Activity
Table 3 summarizes models examining baseline population density and change in population density for total and domain specific PA. In adjusted models, an IQR increase in baseline population density was associated with a -12.4% (95% CI: -16.0, -8.7) lower population median of MET-min of total PA, while an IQR increase in the 5-year change in population density was associated with a -17.2% (95% CI: -25.7, -7.7) lower population median of total PA. The largest negative associations were observed for occupational and household PA, while positive associations were observed for transport and recreational PA. Statistically significant differences in the association of baseline population density on total PA, transport PA, recreation PA and occupation PA were observed for each of HIC, MIC and LIC. On the association of change in population density, statistically significant differences on household PA were observed for HIC, MIC and LIC. Larger magnitudes of associations were typically observed for HICs, compared to MIC or LICs, expect for population density change and recreational PA.
| Population Density at baseline (β per IQR = 4428 people/sq. km) | 5-year change in population density (β per IQR = 2% increase per year) | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| All | HIC | MIC | LIC | p-value† | All | HIC | MIC | LIC | p-value† | |
| Total PA | -12.4 (-16.0, -8.7) | -16.2 (-26.5, -4.6) | -4.3 (-11.4, 3.3) | -16.2 (-20.2, -12.4) | 0.04 | -17.2 (-25.7, -7.7) | -45.7 (-50.6, -40.2) | -13.3 (-24.8, -0.1) | -22.6 (-38.5, -2.4) | 0.17 |
| Transport | 0.8 (-4.7,6.6) | 103.9 (65.6, 151.2) | 29.3 (17.5, 42.2) | -16.2 (-20.9, -11.3) | < 0.001 | 7 (-7.0, 23.0) | -31.6 (-44.5, -15.6) | 13.8 (-4.4, 35.4) | 16.2 (-35.8, 9.3) | 0.30 |
| Recreation | 9.3 (3.7, 15.1) | 6.4 (-12.9, 29.9) | 36.9 (24.5, 50.7) | -4.3 (-8.4, 0.2) | < 0.001 | 6.3 (-7.1, 21.6) | 6.4 (-16.8, 36.1) | 11.6 (-6.2, 32.8) | -22.2 (-36.1, -5.2) | 0.53 |
| Household | -16.2 (-19.9, -12.3) | -38.6 (-47.9, -27.5) | -16.2 (-22.9, -9.0) | -12.4 (-16.6, -8.0) | 0.16 | -23.1 (-31.9, -13.3) | -38.9 (-52.5, -21.3) | -21.0 (-32.3, -7.8) | -15.3 (-33.5, -8.0) | 0.01 |
| Occupation | -16.2 (-21.1, -11.1) | -23.3 (-38.3, -4.8) | -26.7 (-34.5, -17.9) | -8.5 (-13.6, -3.0) | 0.01 | -21.9 (-32.9, -9.1) | -35.1 (-46.0, -21.8) | -19.5 (-34.4, -1.1) | -11.4 (-32.1, -15.6) | 0.37 |
| Adjusted Model: Age, sex, baseline year, education, wealth index, country income level, chronic disease at baseline, BMI, and nested random intercept for each community by center. Change models include baseline population density. † p-value for interaction between country income level and each urbanization metric. # β coefficients represent percent change in MET minutes per week of total or domain specific physical activity. | ||||||||||
Associations between Impervious Area and Physical activity
Table 4 summarizes models examining baseline impervious surface area and change in impervious surface area and total and domain specific PA. In fully adjusted models, an IQR increase in baseline impervious surface area was associated with a -29.2% (95% CI: -37.5, -19.7) lower population median of total PA, while an IQR increase in impervious surface area change (log transformed) was not associated with total PA. Increase in impervious surface area was associated with − 11.6% (95% CI: -21.7, -0.1) lower recreational PA, but was not associated with any other PA outcome. The largest negative associations for baseline impervious surface area and PA were observed in LICs and for household PA, while positive associations were observed for transport and recreational PA in HICs and MICs. Statistically significant differences in the association of baseline impervious surface area on transport PA was observed for HIC, MIC.
| Impervious Area at baseline (β per IQR = 73%) | 5-year change in impervious area (log) (β per IQR = 2% increase) | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| All | HIC | MIC | LIC | p-value† | All | HIC | MIC | LIC | p-value† | |
| Total PA | -29.2 (-37.5, -19.7) | -20.6 (-33.2, -5.4) | -21.8 (-33.1, -8.7) | -64.4 (-74.9, -49.4) | 0.01 | 2.0 (-7.3, 12.2) | -2.9 (-13.0, 8.2) | 1.5 (-11.1, 16.1) | 3.9 (-24.6, 43.5) | 0.98 |
| Transport | 67.2 (42.1, 96.8) | 154.5 (88.9, 242.9) | 102.7 (67.7, 144.9) | -68.4 (-78.6, -53.4) | < 0.001 | -1.9 (-13.5, 11.1) | -8.5 (-23.6, 9.5) | 1.7 (-14.4, 20.8) | -10.4 (-35.9, 25.1) | 0.85 |
| Recreation | 54.9 (32.9, 80.8) | 15.2 (-11.6, 50.1) | 81.3 (49.2, 120.3) | -22.0 (-42.6, 5.9) | 0.30 | -11.6 (-21.7, -0.1) | 2.6 (-13.5, 21.9) | -17.5 (-30.5, -2.0) | -6.1 (-27.8, 22.2) | 0.14 |
| Household | -45.5 (-52.3, -37.7) | -43.9 (-55.4, -29.5) | -43.5 (-52.2, -33.3) | -59.6 (-71.8, -42.4) | 0.24 | 8.30 (-3.11, 21.06) | 6.7 (-7.8, 23.5) | -1.2 (-1.8, -0.5) | 1.6 (-27.0, 41.6) | 0.55 |
| Occupation | -51.5 (-59.3, -41.6) | -34.1 (-50.5, -12.2) | -51.8 (-61.5, -39.6) | -36.4 (-57.0, -6.1) | 0.66 | 15.46 (1.17, 31.78) | 25.6 (3.7, 52.2) | -3.7 (-19.9, 15.8) | 17.1 (-16.9, 65.2) | 0.01 |
| Adjusted Model: Age, sex, baseline year, education, wealth index, country income level, chronic disease at baseline, BMI, and nested random intercept for each community by center. Change models include baseline impervious surface area. † p-value for interaction between country income level and each urbanization metric. # β coefficients represent percent change in MET minutes per week of total or domain specific physical activity. | ||||||||||
Sub-Group Analysis of Urbanization and Physical Activity
We observed statistically significant differences in associations in our sub-group analyses by sex, age, educational status and urban/rural status. Figure 2 illustrated stratified models for total PA and our four urbanization measures. For example, an annual IQR increase in population density change over a 5 year period was associated with lower population median total PA among individuals living in rural areas (-21.1%; 95% CI: -31.0, -9.7) compared to urban areas (-14.3%; 95% CI -28.7,3.0); among males (-18.0%; 95% CI -27.6,-7.8) compared to females (-13.5%; 95% CI: -22.6, -3.3); among individuals > 55 years (-19.6%; 95% CI -29.5,-8.4) compared to individuals ≤ 55 years (-16.2%; 95% CI -24.9,-6.5); and among individuals with < High School education (-21.5%; 95% CI -31.0,-10.7) compared to individuals with high school or greater (-12.1; 95% CI -21.8,-1.2). Similar trends are observed in these sub-group analyses when we stratified by HIC, LIC and MIC (Figure S5-S19).
Sensitivity analysis
Our model results were robust to sensitivity analyses. Adding country to our models to control for potential confounding in urbanization and PA levels between countries, did not alter the results (Table S3 & Table S4). Models that included both population density and impervious surface area measures were consistent in terms of direction of association with minimal changes in the magnitude of association in models compared to using only either population density or impervious surface area measures (Table S5 to Table S6). When we used un-transformed PA measures as well as the untransformed impervious surface area change measure we observed similar associations and conclusions (Table S7 to Table S8).
This is the first study to examine the association between urbanization and different domains of PA for high-, middle-, and low-income countries using individual participant data on PA levels. Across all 698 communities from 22 countries, population density and impervious surface area were independently associated with lower total PA levels, with the largest negative associations for household and occupational PA and small increased associations for transport and recreation PA. Change in population density levels for the 5 years prior to study baseline was also significantly associated with lower total PA, household PA, and occupational PA, but also increased associations for recreational and transport PA. Less robust associations were observed for change in community impervious surface area. Important differences in the associations between urbanization metrics and specific domains of PA were observed between country income levels, by urban and rural status, and gender.
Consistent with our results, other studies in HICs, MICs and LICs reported that urbanization is associated with lower overall PA16,31−33. Specifically, a study in China, a MIC, reported community urbanization measures (population, density, access to markets for household goods, economic wellbeing, transportation, communications, educational institutions, health facilities, sanitation and housing infrastructures) were associated with approximately 57% and 40% declines in total PA for men and women respectively over a 9 year period16. In Iran, a MIC, researchers found that urbanization, measured using an index created from demographic, socioeconomic, and health-related variables, was associated with 7% and 2% higher odds of lower PA among males and females respectively33. Findings from a study in Kenya, a LIC, on the effect of urbanization, measured using an urban-rural classification, observed lower PA levels among urban residents compared to rural counterparts17. While the measures of urbanization in these studies are very different, they suggest that overall, increasing urbanization is associated with decreased total PA, supporting the results of our study.
Importantly, different urbanization measures vary in their impact on different domains of PA, and the magnitude of change also varies across high-, middle-, and low- income countries. These findings imply that policy makers globally need to address specific pathways and mechanisms by which urbanization is related to PA. The larger reductions in occupational PA and household PA observed in our study, especially among LICs and MICs compared to HICs, could be explained by the changes in occupational activities, socioeconomic status, and social and physical structures that happen with the rapid urbanization occurring in LICs. These changes may impact the lifestyles of individuals directly or indirectly, leading to changes in PA levels22. For example, significant urbanization generally increases white collar jobs34, especially in service industries, which decreases PA levels. Similarly, urbanization may lead to technological advances that result in an increase in effortless household equipment and appliances that reduce domestic PA22. We observed the largest associations between urbanization and PA in LIC’s, driven by occupation and household PA declines. Alternatively, recreation and transport PA was higher for some urbanization metrics, especially in HIC and MICs, suggesting potential mechanisms to increase total PA through urban planning that encourages walking and cycling.
There is no universal definition or measurement method for “urbanization”, and a range of different metrics have been used to study associations between urbanization and PA or health35,36,37. Urbanization is an extremely complex phenomenon and different urbanization measures likely capture different aspect of the urban environment and change. In our study we observed moderate correlations between two measures (community population and impervious area) with PA, and these measures likely capture different components of urbanization (with impervious area capturing more development-related components of urbanization). Urbanization can also be context-specific; thus, an urbanization indicator may not necessarily capture the same construct in different communities (i.e. HIC, MIC or LIC country settings)38. Nevertheless, our two urbanization measures are objective measures that can be applied to any community, and are common urban metrics that capture large-scale upstream urban characteristics important to PA. Our results will help policy makers and urban planners understand how urbanization is associated with PA and ultimately how urbanization can be optimized to increase population PA levels. Further, our findings better inform urban public health policy makers about the major health problems that may arise with urbanization in their regions without the required social supports and infrastructure changes to support overall population PA. This is especially important in rapidly urbanizing communities in developing countries.
While our study has several strengths, there are some potential limitations to highlight. First, this is a cross-sectional study, thus making it difficult to assess the causality of related factors. However, we used a five-year change measure prior to enrollment to capture urbanization changes in each community. In addition, the PURE population is diverse and captures different populations and community settings, adding to the generalizability of our findings. Second, this study relies on self-reported measures of PA, which may be influenced by social desirability and recall biases. However, the use of self-reported PA estimates in this study has shown good validity and reliability against accelerometer data and other self-reported measures39. In addition, the inclusion of multiple domains of PA (transport, leisure, occupation and household) is unique and important for determining how urbanization is associated with different domains of PA, especially in developing countries where there is a paucity of studies. Third, we measured urbanization only using population density and impervious surface area. Other important dimensions of urbanization, such as of infrastructure, economic and demographic characteristics, were not directly measured. Nevertheless, these two objectively derived measures capture different aspects of urbanization that are important to PA, as well as for policy and planning. Finally, while we had detailed information on individual, household and community variables, there are unmeasured factors that are likely important to our analyses. For example, sociocultural norms may discourage outdoor PA for girls and women aside from within the home (household PA) or built environment design may influence the amount of transport of recreation PA available. While these factors are important and will be examined in future research, here we examined the potential impact of these unmeasured factors on our analysis using center fixed effects, which did not change our over-all results.
Urbanization was associated with lower total PA in this diverse global study. Lower household and occupational PA were most strongly associated with urbanization, especially in LICs and MICs. In HICs, urbanization was associated with higher recreation PA. Important differences in the associations between urbanization metrics and PA were observed between country income levels, as well as by urban and rural status and gender. These findings provide new information on the complex associations between urbanization and PA.
Study Design and Participants
Detailed methodology and design of the PURE study have been described elsewhere40,41. The PURE study consists of approximately 183,503 adults aged 35–70 years residing in 800 urban and rural communities in 27 countries42. Here we restrict to the 138,206 adults with complete PA outcomes assessed at baseline using the long-form International PA Questionnaire (IPAQ). Our study population represents individuals from 698 communities located in 22 countries, including 5 low income countries (LICs) (Bangladesh, India, Pakistan, Tanzania, and Zimbabwe); 13 middle income countries (MICs)(Argentina, Brazil, Chile, China, Colombia, Iran, Malaysia, Palestine, Philippines, Poland, Russia, South Africa, and Turkey); and 4 high income countries (HICs) (Canada, Saudi Arabia, Sweden, and the United Arab Emirates)42. The countries are grouped into economic levels based on World Bank classifications at study baseline43. The choice and number of countries selected in PURE reflects a balance involving many countries at different economic levels with substantial heterogeneity in social and economic circumstances. Within larger PURE countries data collection was conducted at the center level, which represents distinct geographic regions. Participants were sampled from communities, representing neighborhoods in urban areas and small villages in rural areas. Within each defined community, households were approached, and individuals aged 35–70 years who intended to live at their current address for at least another 4 years were invited to participate in the study. Baseline data collection occurred at the community, household and individual levels. Variables such as sociodemographic factors, medical history, lifestyle behaviors, and risk factors including baseline chronic diseases were recorded in the PURE study using standardized measures and procedures44. All participants in the PURE study provided written informed consent. The study was coordinated by the Population Health Research Institute (Hamilton Health Sciences, Hamilton, ON, Canada) and approved by the Hamilton Health Sciences Research Ethics Board and local ethics committees at each center. The study protocol conformed to the ethical guidelines as stated in the Declaration of Helsinki.
Physical Activity Levels
PA outcomes were assessed using the long-form IPAQ. Our outcome included the four domains of PA (occupational, transportation, housework, recreational) and total PA reported in metabolic equivalents (MET) × minutes per week (MET-minutes per week). Participants were asked to report frequency and duration of specific activities including vigorous and moderate intensity physical activities, and walking in terms of the frequency and duration in the last 7 days. For each individual, the recorded activities were converted to MET-minutes per week using equations adapted from the IPAQ data processing guidelines45. Specific rules for screening outliers are outlined in the IPAQ data screening method. A summary of the final equations used in deriving PA is presented in Table S1.
Urbanization Measures
Measuring the degree of urbanization is challenging due to varying definitions used, especially across different countries and settings46. Urbanization has been measured using a variety of measures such as population size and density, land, economy, society and ecologic variables35,36, or the presence of impervious surfaces or built up areas47. Impervious area refers to impervious materials such as concrete, metal, glass, tarmac, and plastic materials that cover the soil and serve as an important factor for the ecological impact of the built environment48. For the purposes of this study, population density and impervious surface area were used as our primary measures of urbanization, as these two measures capture different urbanization dimensions that can be objectively measured with satellite data using similar metrics across diverse communities and countries across the world.
Population Density: We used the fourth version of the Gridded Population of the World (GPWv.4)49 to measure population density for all PURE community locations. The GPWv4 models the distribution of human population (counts and densities) using sub-national population estimates from around the world, which are derived using periodic census data for each country and is gridded with an output resolution of 30 arc-seconds (approximately 1 km at the equator) and is available for the years 2000, 2005, 2010, and 2015. We calculated population density within 1 and 5 km buffer regions for each PURE community centroid. We examined the 1 km buffer region as a measure of local community change and the 5 km buffer as a measure of larger-scale urbanization in an area. We examined two primary urbanization measures using population density. The first is baseline population density for each PURE community at the time the IPAQ questionnaires were completed. We assigned the closest GPWv4 year to each year of enrollment in the PURE study (2003–2015). The second is the 5-year annual population density rate change (Eq. 1 below) prior to study baseline in each community. We use this equation to calculate the average annual change in population density per year to be consistent with other studies of urbanization50–52.
Impervious Area
We examined impervious area as another measure of urbanization. Impervious area is related to population density but captures a distinct component of urbanization related to land use/cover change53,54. We derived impervious area metrics for each PURE community using the global artificial impervious area (GAIA) dataset55. Briefly, GAIA is a global dataset that captures the growth of artificial impervious areas annually from 1985 to 2018. GAIA is made up of mapped annual artificial impervious area at a 30 m resolution from the full archive of Landsat images. The modelling approach to capturing impervious area demonstrated a mean overall accuracy over multiple years greater than 90%55. Similar to population density, we calculated the percent of impervious area within a 1 km and 5 km buffer regions for each PURE community centroid. We assigned the corresponding impervious surface area to each year of enrollment in the PURE study (2003–2015). We examined two primary urbanization measures using impervious area, similar to the approach used for population density. The first is baseline percent impervious area for each PURE community. The second is the 5-year impervious area rate change (Eq. 2 below). Because the 5-year impervious area rate change was highly skewed, we use the log transformed measure in all analyses.
Statistical Analysis
We first report descriptive statistics (e.g., mean, frequency, standard deviation) of all PA outcomes, urbanization measures and covariates. We excluded individuals with missing key continuous covariates and for missing categorical variables included a missing data category. Next, we examined correlations between all urbanization measures using Pearson’s correlations. We modelled cross-sectional associations between each urbanization measures and each PA outcome (total, leisure, transport, occupational, household) using generalized mixed effects models with a nested random intercept for communities by center to account for the clustered sampling design of the PURE study. We present adjusted models that include age, sex, baseline year, household wealth index, education, BMI, chronic disease status at baseline, country income level and a nested random intercept for communities by center. Covariates were determined a priori that may be related to urbanization and PA but not directly on the causal pathway. For models examining change in population density and impervious surface area we also include the baseline population density or impervious area in the models to control for differences in urbanization levels. We ran overall models and models stratified by country income levels (HIC, LIC, and MIC). Finally, we examined models stratified by key individual variables, including age (≤ 55, > 55), sex (male, female), educational status (< high school, > high school) and urban/rural status.
To account for residual non-normality distributions of the models and to describe the full distribution of PA data, we used log transformed PA outcome variables in all analyses using y = ln(y + 1). To interpret our model results, we exponentiated the coefficient, subtracted one from this number, and multiplied by 100. This gives the percent increase (or decrease) in the response variable for every one-unit increase in the independent urbanization variables. For example, for every one-unit increase/decrease in the independent variable, the population median of our dependent variable increases/decreases by x%. Thus, in the present study, percent change was approximated by setting both random effects equal to zero and using ln(y) as an approximation to ln(y + 1). We examined quartiles for total PA and the domains of PA, which did not demonstrate statistically significant departures from a linear relationship and therefore present linear results for all models.
We conducted several sensitivity analyses to evaluate the robustness of our main model results. First, we examined joint models that included both population density and impervious surface area measures to determine if these urbanization metrics have independent associations with PA. Second, we examined adding country to our models to control for potential confounding in urbanization and PA levels between countries. Third, we examined model results using un-transformed PA measures. All statistical analyses were performed using SAS® software 9.4.56 We present results for fully adjusted models in our tables.
ACKNOWLEDGEMENTS
The PURE study is funded by the Population Health Research Institute, Hamilton Health Sciences Research Institute, the Canadian Institutes of Health Research (including through the Strategy for Patient-Oriented Research via the Ontario SPOR Support Unit), the Heart and Stroke Foundation (ON, Canada), the Ontario Ministry of Health and Long-Term Care. The PURE-AIR study is funded by the Canadian Institutes for Health Research (grant 136893) and by the Office of the Director, National Institutes of Health (award DP5OD019850). It is also funded by unrestricted grants from several pharmaceutical companies, with major contributions from AstraZeneca (Canada), Sanofi-Aventis (France and Canada), Boehringer Ingelheim (Germany and Canada), Servier Laboratories, and GlaxoSmithKline, and additional contributions from Novartis, King Pharma, and from several national and local organizations in participating countries.
AUTHOR CONTRIBUTIONS
P.H and K.B conceptualized the research questions, analyzed the data, interpreted the results, and assisted with the writing, review and editing of the manuscript. SY designed the overall PURE study, obtained funding, and oversaw its conduct since its inception 18 years ago. R.I, O.B, J.D, Y.A, R.A, R.Y, R.K, P.J, R.I, P.S, A.R, P.P, P.L, R.K, K.Z, B.H, L.Y, C.W, K.Y, J.C, K.A, A.A, A.D, S.L and S.Y assisted with interpreting the findings, provided feedback on study designs and provided a critical review. All authors have read and agreed to the published version of the manuscript.
DATA AVAILABILITY
The datasets generated and/or analysed during the current study are not publicly available due to the confidential nature of the PURE health data but are available from the corresponding author on reasonable request.
Additional Information
The authors declare no competing interests.
The authors declare they have nothing to disclose.
- Nations, U. 2018 Revision of World Urbanization Prospects. United Nations https://www.un.org/en/desa/2018-revision-world-urbanization-prospects.
- Urbanization. https://www.un.org/en/development/desa/population/theme/urbanization/index.asp.
- Huang, N.-C., Kung, S.-F. & Hu, S. C. The Relationship between Urbanization, the Built Environment, and Physical Activity among Older Adults in Taiwan. Int. J. Environ. Res. Public. Health 15, (2018).
- Monda, K. L., Gordon-Larsen, P., Stevens, J. & Popkin, B. M. China’s transition: The effect of rapid urbanization on adult occupational physical activity. Soc. Sci. Med. 64, 858–870 (2007).
- Physical activity. https://www.who.int/news-room/fact-sheets/detail/physical-activity.
- Taylor, D. Physical activity is medicine for older adults. Postgrad. Med. J. 90, 26–32 (2014).
- Organization, W. H. Global Action Plan on Physical Activity 2018–2030: More Active People for a Healthier World. (World Health Organization, 2019).
- Almeida, O. P. et al. 150 minutes of vigorous physical activity per week predicts survival and successful ageing: a population-based 11-year longitudinal study of 12 201 older Australian men. Br. J. Sports Med. 48, 220–225 (2014).
- WHO | New physical activity recommendations for reducing disease and prevent deaths. WHO http://www.who.int/chp/media/news/releases/2011_2_physicalactivity/en/.
- CDC. Benefits of Physical Activity. Centers for Disease Control and Prevention https://www.cdc.gov/physicalactivity/basics/pa-health/index.htm (2021).
- WHO | Prevalence of insufficient physical activity. WHO http://www.who.int/gho/ncd/risk_factors/physical_activity_text/en/.
- Physical activity. https://www.who.int/news-room/fact-sheets/detail/physical-activity.
- Hallal, P. C. et al. Global physical activity levels: surveillance progress, pitfalls, and prospects. The Lancet 380, 247–257 (2012).
- Assah, F. K., Ekelund, U., Brage, S., Mbanya, J. C. & Wareham, N. J. Urbanization, Physical Activity, and Metabolic Health in Sub-Saharan Africa. Diabetes Care 34, 491–496 (2011).
- Machado-Rodrigues, A. M. et al. Urban–rural contrasts in fitness, physical activity, and sedentary behaviour in adolescents. Health Promot. Int. 29, 118–129 (2014).
- Ng, S. W., Norton, E. C. & Popkin, B. M. Why have physical activity levels declined among Chinese adults? Findings from the 1991–2006 China Health and Nutrition Surveys. Soc. Sci. Med. 1982 68, 1305–1314 (2009).
- Ojiambo, R. M. et al. Effect of Urbanization on Objectively Measured Physical Activity Levels, Sedentary Time, and Indices of Adiposity in Kenyan Adolescents. J. Phys. Act. Health 9, 115–123 (2012).
- Scarabottolo, C. C. et al. Relationship of different domains of physical activity practice with health-related quality of life among community-dwelling older people: a cross-sectional study. BMJ Open 9, e027751 (2019).
- WHO | Physical Activity and Adults. WHO https://www.who.int/dietphysicalactivity/factsheet_adults/en/.
- Macniven, R., Bauman, A. & Abouzeid, M. A review of population-based prevalence studies of physical activity in adults in the Asia-Pacific region. BMC Public Health 12, 41 (2012).
- Church, T. S. et al. Trends over 5 Decades in U.S. Occupation-Related Physical Activity and Their Associations with Obesity. PLOS ONE 6, e19657 (2011).
- Gong, P. et al. Urbanisation and health in China. Lancet 379, 843–852 (2012).
- Sturm, R. The economics of physical activity: societal trends and rationales for interventions. Am. J. Prev. Med. 27, 126–135 (2004).
- Pratt, M., Macera, C. A., Sallis, J. F., O’Donnell, M. & Frank, L. D. Economic interventions to promote physical activity: application of the SLOTH model. Am. J. Prev. Med. 27, 136–145 (2004).
- Cawley, J. An economic framework for understanding physical activity and eating behaviors. Am. J. Prev. Med. 27, 117–125 (2004).
- Heath, G. W. et al. Evidence-based intervention in physical activity: lessons from around the world. Lancet Lond. Engl. 380, 272–281 (2012).
- Armstrong, T. & World Health Organization. A guide for population-based approaches to increasing levels of physical activity: implementation of the WHO global strategy on diet, physical activity and health. (World Health Organization, 2007).
- Lambert, E. V. et al. Role of physical activity for health in communities undergoing epidemiological transition. World Rev. Nutr. Diet. 90, 110–126 (2001).
- Popkin, B. M., Paeratakul, S., Zhai, F. & Ge, K. A review of dietary and environmental correlates of obesity with emphasis on developing countries. Obes. Res. 3<bvertical-align:super;> </bvertical-align:super;><background-color:#C8BE84;bvertical-align:super;>Suppl 2</background-color:#C8BE84;bvertical-align:super;>, 145s–153s (1995).
- Ng, S. W., Howard, A.-G., Wang, H., Su, C. & Zhang, B. The Physical Activity Transition among Adults in China: 1991–2011. Obes. Rev. Off. J. Int. Assoc. Study Obes. 15, (2014).
- Dong, Y. et al. Prevalence of Type 2 diabetes in urban and rural Chinese populations in Qingdao, China. Diabet. Med. J. Br. Diabet. Assoc. 22, 1427–1433 (2005).
- Liu, S. et al. Prevalence of diabetes and impaired fasting glucose in Chinese adults, China National Nutrition and Health Survey, 2002. Prev. Chronic. Dis. 8, A13 (2011).
- WHO EMRO | Urbanization and noncommunicable disease (NCD) risk factors: WHO STEPwise Iranian NCD risk factors surveillance in 2011 | Volume 23, issue 7 | EMHJ volume 23, 2017. http://www.emro.who.int/emhj-volume-23-2017/volume-23-issue-7/urbanization-and-noncommunicable-disease-ncd-risk-factors-who-stepwise-iranian-ncd-risk-factors-surveillance-in-2011.html.
- Popkin, B. M. FAN. World Dev. 27, 1905–1916 (1999).
- Vlahov, D. & Galea, S. Urbanization, urbanicity, and health. J. Urban Health Bull. N. Y. Acad. Med. 79, S1–S12 (2002).
- Allender, S., Foster, C., Hutchinson, L. & Arambepola, C. Quantification of Urbanization in Relation to Chronic Diseases in Developing Countries: A Systematic Review. J. Urban Health Bull. N. Y. Acad. Med. 85, 938–951 (2008).
- Galea, S. & Vlahov, D. URBAN HEALTH: Evidence, Challenges, and Directions. Annu. Rev. Public Health 26, 341–365 (2005).
- Allender, S., Wickramasinghe, K., Goldacre, M., Matthews, D. & Katulanda, P. Quantifying urbanization as a risk factor for noncommunicable disease. J. Urban Health Bull. N. Y. Acad. Med. 88, 906–918 (2011).
- Cleland, C., Ferguson, S., Ellis, G. & Hunter, R. F. Validity of the International Physical Activity Questionnaire (IPAQ) for assessing moderate-to-vigorous physical activity and sedentary behaviour of older adults in the United Kingdom. BMC Med. Res. Methodol. 18, (2018).
- Corsi, D. J. et al. Prospective Urban Rural Epidemiology (PURE) study: Baseline characteristics of the household sample and comparative analyses with national data in 17 countries. Am. Heart J. 166, 636–646.e4 (2013).
- Teo, K., Chow, C. K., Vaz, M., Rangarajan, S. & Yusuf, S. The Prospective Urban Rural Epidemiology (PURE) study: Examining the impact of societal influences on chronic noncommunicable diseases in low-, middle-, and high-income countries. Am. Heart J. <bvertical-align:super;>158</bvertical-align:super;>, 1–7.e1 (2009).
- Arku, R. E. et al. Characterizing exposure to household air pollution within the Prospective Urban Rural Epidemiology (PURE) study. Environ. Int. 114, 307–317 (2018).
- World Bank. World Bank Country and Lending Groups – World Bank Data Help Desk. World Bank Country and Lending Groups https://datahelpdesk.worldbank.org/knowledgebase/articles/906519-world-bank-country-and-lending-groups.
- Yusuf, S. et al. Effect of potentially modifiable risk factors associated with myocardial infarction in 52 countries (the INTERHEART study): case-control study. Lancet Lond. Engl. 364, 937–952 (2004).
- Ara, A. Guidelines for Data Processing and Analysis of the International Physical Activity Questionnaire (IPAQ) – Short and Long Forms Contents.
- Ritchie, H. & Roser, M. Urbanization. Our World Data (2018).
- Haustein, M. D. The urban-rural environment: effects of impervious surface land cover on lake ecosystems. (2010).
- Scalenghe, R. & Marsan, F. A. The anthropogenic sealing of soils in urban areas. Landsc. Urban Plan. 90, 1–10 (2009).
- Socioeconomic Data and Applications Center (sedac). Gridded Population of the World (GPW), v4 | SEDAC. https://sedac.ciesin.columbia.edu/data/collection/gpw-v4.
- Chen, J., Chang, K., Karacsonyi, D. & Zhang, X. Comparing urban land expansion and its driving factors in Shenzhen and Dongguan, China. Habitat Int. 43, 61–71 (2014).
- Ma, Y. & Xu, R. Remote sensing monitoring and driving force analysis of urban expansion in Guangzhou City, China. Habitat Int. 34, 228–235 (2010).
- Xu, X. & Min, X. Quantifying spatiotemporal patterns of urban expansion in China using remote sensing data. Cities 35, 104–113 (2013).
- Greiner, P. T., Shtob, D. A. & Besek, J. F. Is Urbanization Good for the Climate? A Cross-County Analysis of Impervious Surface, Affluence, and the Carbon Intensity of Well-Being. Socius 6, 2378023119896896 (2020).
- Balk, D., Leyk, S., Jones, B., Montgomery, M. R. & Clark, A. Understanding urbanization: A study of census and satellite-derived urban classes in the United States, 1990–2010. PLoS ONE 13, (2018).
- Gong, P. et al. Annual maps of global artificial impervious area (GAIA) between 1985 and 2018. Remote Sens. Environ. 236, 111510 (2020).
- SAS Institute Inc 2013. SAS/ACCESS® 9.4 Interface to ADABAS: Reference. Cary, NC: SAS Institute Inc. https://www.sas.com/en_us/solutions.html.
No competing interests reported.