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.