This study examined the association between residential density and physical activity among urban adults in a rapidly-urbanizing region in China. A negative gradient association between residential density and physical activity was identified within the study participants in that the odds ratio of achieving sufficiently physical activity decreased by gradient from lower, middle to higher tertile of residential density. Our findings were inconsistent with the majority of literature from Western countries where residential density was positively associated with physical activity [11,12,13,14,15,16,17,18], but in line with our previous study conducted among urban Chinese adolescents in the same city [20].
In this study, the proportion of participants engaging in sufficient physical activity was 28.5% (442/1551), which was higher than the national average figure (22.2%) among Chinese urban adults officially reported by China Committee of Health and Family Planning in 2015 [34]. Therefore, low level of sufficient physical activity is a public health problem for Chinese people and physical activity promotion is in urgent need for China. Among those 422 participants who met the physical activity recommendations, the smallest number of persons within the categories of sufficient physical activity was 56 for women within the upper tertile of residential density. In the basis of statistical principle, such a minimal number of participants can warrant an acceptable statistical power in the data analysis. So there would be no marked adverse impact on statistical power caused by the number of participants with sufficient physical activity in our study.
Interestingly, the selected individual-level potential confounders (socio-demographic characteristics and anthropometric measures) exerted little influence on the RD-PA relationship, but the community-level factors (green space and potential neighborhood-level clustering effects) substantially mediated the RD-PA association in our study. This suggested that neighborhood-level factors might exert more important impact on local residents’ physical activity than participants’ personal socio-demographic and anthropometric characteristics. Moreover, those neighborhood-level factors markedly mediated the RD-PA relationship for women but not for men, which implied that women’ engagement in physical activity might be more easily influenced by local neighborhood-level factors relative to men. Thus, neighborhood-level covariates should be put into consideration in future studies regarding neighborhood built environment and physical activity.
The mechanisms behind RD-PA link are complicated. The most important potential explanation widely used is that an area with higher residential density, usually meaning more diverse land use and closer destinations, has more recreational facilities (stores, parks, transport stops, sport facilities, etc.) available for local residents. Thus, local residents tend to physically visit those neighborhood-around facilities or destinations by walk or bicycle [11,12,13,14,15,16,17,18], which results in that residents within higher densed neighborhoods are more likely to be physically active than their counterparts from lower densed areas. This mechanism is able to easily explain the positive association between RD and PA in Western countries, but can not explain the negative RD-PA link observed in this study.
It needs to note that the mean value of residential density within Nanjing urban areas was 9267 persons/km2 in 2016 [19, 27], which is unbelievably higher than that in typical urban areas of American or Australian cities where PA-related built environment attributes have been identified [35, 36]. For example, the mean residential density was 1210 persons/km2 in Atlanta, USA in 2010 [37]. In epidemiological studies regarding residential density and health conditions in Western communities, an area with ≥500 persons/km2 was usually classified as densely populated [38]. Based on this cutoff (500 persons/km2), RD was dichotomized as “high density” or “low density” in studies examining RD-PA relationship in Western societies, while RD was tertiled into three subgroups in our study. Furthermore, such an area with population of 500+ persons/km2 in Western societies might be a sparsely populated community in China. In our study, the cutoffs were 29.8 thousands and 56.5 thousands residents/km2 for middle and upper tertile, respectively. Even the population density within the lower tertile in our study was still much higher than 500 persons/km2, the cutoff for high residential density in Western countries. Thus, these may be the important factors contributing to the different scenarios of RD-PA relationship in China and Western countries.
The RD-PA relationship among adults in this study was in line with that observed within adolescents in the same urban areas in Nanjing [20]. For either adults or adolescents, RD was negatively associated with PA in urban areas of Nanjing. In those two studies, RD-PA association was examined using RD as tertiles and PA as binary variable (“sufficient” or “insufficient”). The consistent findings from those two studies suggested that RD might have similar influence on residents’ physical activity irrespective of their age in urban areas of regional China.
There was a possible explanation on such an inverse association between residential density and physical activity observed in China. One important potential inhibitor to outdoor physical activity may be roads/streets and traffic volume around neighborhoods. For those neighborhoods with high-enough population density in mega-cities like our study city, the roads/streets around neighborhoods are typically narrow and/or crowded with traffic which implying not sufficiently safe, even if not unsafe, for outdoor activities. Therefore, local residents within such highly-densed neighborhoods might not be willing to take the potential risk to do outdoor physical activities. People might choose to stay at home instead to visit parks or some places with physical activity facilities by walking a little bit long distance under such unsafe traffic-related circumstance. This might, at least in part, explain that the very population-densed urban environment may inhibit physical activity observed in our city. However the mechanisms behind RD-PA relationship are really complex and need to be further investigated with well-designed programs in future.
Regarding the strengths of this study, it is the first study reporting RD-PA relationship among urban adults from a rapidly urbanizing region in China. In this study RD was used as tertiles not dichotomous variables, which was more informative and allowed us to look at a gradient RD-PA association. The second strength is that, in addition to conventional confounders (including green public space), community-level potential clustering-effects were also considered using mixed-effects regression models. Finally, an interesting finding, a negative RD-PA association, was examined among urban adults, which was consistent with that in our previous study among Chinese urban adolescents, but not in line with those reported from Western communities, implying complex mechanisms behind the RD-PA link and suggesting further well-designed studies needed in different countries.
Several limitations also need to be noted. First, the RD-PA link did not imply any causality as it was observed from a cross-sectional study. Second, information on PA was self-reported by participants, which might cause potential recall bias, although the PA questionnaire has been validated and widely used to collect PA data in epidemiological surveys [39, 40]. Finally, although socio-demographic characteristics, public green space and community-level cluster-effects were controlled for in our analysis, some other potential confounders could not be considered due to data limitation. Therefore, considering these limitations of this study, the findings shall be prudently interpreted, and well-designed longitudinal studies are in need to demonstrate the influence of residential density on residents’ physical activity in future.