Associations between access to recreational physical activity facilities and body mass index in Scottish adults

The prevalence of obesity, which is high in developed countries and rising worldwide, is associated with a significant burden of ill-health including cardiovascular disease, certain cancers and diabetes [1‐5], particularly in lower socio-economic groups [6]. Efforts to address key determinants of obesity, such as initiatives to promote physical activity directed at individuals, have had limited success to date [7, 8]. This has led to increasing attention on other potential determinants not specific to the individual [9], including efforts to understand environmental influences, particularly those related to the built environment [10, 11]. However, most studies to date have tended to use relatively simple metrics of accessibility such as proximity to the nearest facility or the density of facilities in the neighbourhood of residence. Few studies have taken account of the fact that individuals may travel further afield to reach facilities, or that access of this kind may be inequitably patterned in that low income groups may be less likely to have a private motor vehicle available to take them to more distant destinations. In addition, few studies have focused on the specific roles of different forms of transport such as cars and public transport on access to physical activity opportunities, which may in turn promote participation and affect subsequent obesity levels [12]. This is despite a longstanding literature on the importance of the availability of public transport for access to health care [13, 14].

Some studies, mainly from the USA and Australia, have shown that distance to physical activity opportunities and perceived availability of transport options appears to influence use [15‐17]. Moreover, residents of rural areas may be particularly constrained in their access to a range of goods and services if public transport is limited [18]. In our Scotland-wide study [19] we found that those without car access but living in more affluent neighbourhoods [20] or rural areas where public transport provision is low [21] might be particularly vulnerable to having difficulties in accessing health promoting facilities.

In this paper, we extend our previous work by linking individual health and behavioural data with area-level data on accessibility to examine whether access to recreational physical activity facilities by different modes of transport was associated with the body mass index of adults in Scotland, and the extent to which any association was mediated by physical activity participation.

The sample included in the analysis comprised 2,860 men and 3,505 women with a mean age of 49.1 years, of whom 77 % had access to a car and 63 % lived in urban areas (Table 1). The mean BMI of the sample was 27.3 kg/m² (Table 2). There was strong evidence of an association between age and BMI (p < 0.001) and there was also evidence that participants with no longstanding illness had a lower mean BMI than those with a limiting or non-limiting longstanding illness; that lower social class was associated with a higher mean BMI; and that participants with the worst quintile of dietary quality had a lower mean BMI than those in the middle quintile. There was no evidence of a difference in mean BMI between males and females, or between those with and without access to a car, but there was evidence of an association between BMI and the PA measures (data now added to Table 2).

Associations between 24 separate accessibility measures and BMI were tested. There was evidence of an association between almost a third (7/24) of these measures and BMI after adjusting for age, gender, longstanding illness, car availability, social class, dietary quality and urban/rural classification in separate multilevel multivariable models, all in the expected (negative) direction indicating a reduction in mean BMI with increasing accessibility of facilities (Table 3). The seven significant measures of accessibility were those for all facilities within a 20 or 30 min walk or a 20 min cycle; and for private facilities within a 20 or 30 min walk, a 20 min cycle or a 20 min bus journey. There was no evidence of an association between BMI and any of the accessibility measures relating to bus travel to public facilities, or any of the measures of accessibility by car, in the multivariable models.

To address the first criterion for mediation by participation in physical activity, each of these seven significant accessibility measures was considered as a predictor of physical activity participation. Although some significant associations were observed in univariable multinomial regression models, none of the seven accessibility measures remained significantly associated with participation in any physical activity or sport or exercise activity in multivariable models after adjustment for age, gender, social class, longstanding illness, dietary quality, car availability or urban/rural classification (data available on request). It therefore did not appear that the measures of physical activity participation in the Scottish Health Survey were capable of meeting the criteria to be identified as mediators of the relationship between accessibility of recreational physical activity facilities and BMI. Formal tests of the remaining Baron and Kenny criteria were therefore not performed.

In our Scotland-wide study, we found some evidence of an association between proximity to physical activity facilities and individual BMI, with almost a third of the accessibility measures we examined exhibiting a significant negative association with BMI. There were no significant associations between BMI and access to publicly owned physical activity facilities by any of the travel accessibility measures when we explored these separately by facility ownership. Although we found small effect sizes, these would nevertheless be important if adopted across a whole population [36]. We have previously reported that there are more public physical activity facilities in deprived areas across mainland Scotland [19] but this apparent advantage does not seem to be reflected in BMI levels, since obesity is more prevalent among people (particularly women) living in deprived areas [37]. It may be that other factors beyond transport access per se are more important; for example, public transport can be viewed as of inferior status or less comfortable [38, 39]. Some studies have found that residents’ perceptions of the pleasantness or safety of their neighbourhood influences their use of different types of local facilities that might promote physical activity [40]. The cost of using facilities [41] and their perceived availability may be more important than transport access which is objectively measured [42‐44] and the extent to which different groups perceive physical activity facilities as suitable for their needs may be a contributory factor [45]. It may be that in designing interventions to encourage participation in physical activity, other factors such as facilities for childcare may be needed [46]. Psychosocial factors such as lack of confidence or anxieties over one’s appearance, or having someone to go with, have also been shown to be important [47, 48]. The role of transport access in relation to physical activity opportunities may be more important for some groups, e.g. the elderly [49‐51], among whom a more tailored approach to promoting access may be required.

A strength of our study is that we developed a novel transport network across a whole country and were able to match this to data on the body size of individual participants in a large population survey. To our knowledge this has never been done before. A limitation of our study is that 18 % of the sample had missing or improbable values for BMI. Although there is increasing use of multiple imputation of missing data [52], as much of our missing data was in the outcome variable and we had no auxiliary variables in the survey data which we felt would predict missing objectively measured height and weight we chose to conduct a complete case analysis. Although complete case analysis assumes the data are Missing Completely At Random (MCAR), complete case analysis with covariate adjustment has been found to yield unbiased estimates of associations between exposures and outcomes when the data are Missing At Random (MAR) [53]. The Scottish Health Survey PA measure is limited in that it relies on self-report, but it would be challenging and costly to obtain objectively measured PA (e.g. through accelerometry) on over 8,000 respondents. Moreover, the self-report module covers a wide range of activities and a similar module has been validated against objectively measured PA [54]. The use of general rather than specific measures of PA such as walking and sports participation reflects the need for breadth (rather than depth) in general national population surveys. A further limitation of our study is that the Scottish Health Survey does not contain information on whether individuals actually used any of the facilities we examined. Therefore, we cannot assume that a given individual actually benefited from the variety or availability of facilities accessible from his or her local area, or that this explains the association with BMI. Other factors, unmeasured in our study, may also be relevant. Firstly, in relation to the physical environment, people may undertake physical activity in different settings. For example, some groups may prefer to exercise in open space [55] or may prefer to travel further afield to use particular facilities which are not in their area. Indeed, it may be more convenient for some people to use facilities near their place of work [56] rather than their immediate area of residence, whereas our analysis is predicated on respondents’ area of residence and we did not have information on where they worked. It is interesting to note here that in our study there was no association between BMI and access to facilities within 20 min by car, suggesting that a comparative lack of knowledge of the locations and characteristics of facilities further from respondents’ homes may be a contributory factor [55]. Secondly, we cannot discount the possibility that individuals may self-select into areas with a range of opportunities to be physically active [11]. However, this may be a more salient issue for those who have more choice over where they live, e.g. home owners rather than those who are bound by the availability and location of public sector housing (the predominant housing tenure in deprived areas in Scotland). Thirdly, social and cultural factors may contribute; for example, some studies have found that low income groups are more likely than higher income groups to perceive that there are fewer local facilities than actually exist on the ground [40, 49, 57]; in addition, social norms for health related behaviours may vary across different neighbourhoods and networks [58]. Fourthly, it may be that the food environment is more important for obesity than the physical activity environment [9]. However, we were unable to investigate the role of the local food environment on BMI as we know of no readily available single source dataset on the food environment across Scotland comparable to the sportscotland dataset. We have, however, controlled for dietary quality in our study. Finally, we lack information on any differences in accessibility of facilities by bus (or indeed by car) between 2003, when the physical activity data were collected, and 2007, when the facilities were ascertained. It is unlikely that there would have been significant changes in the locations of facilities, given that land use change tends to happen relatively slowly, and smaller local changes are likely to have balanced out across the country and in our analysis. Changes in the provision of public transport could also have taken place. The annual vehicle-kilometres (km) provided by local bus services in Scotland increased by 7.6 % between 2003/04 (369 million km) and 2007/08 (397 million km) [59]. However, this proportionate increase overestimates any change in accessibility as conceived in our analysis, which is based on the travel time between data zone centroids and sports facilities. More services (and more vehicle-kilometres) do not necessarily mean faster journey times (and hence increased accessibility) overall. In a deregulated public transport system, such data are more likely to reflect increases in service frequency on more popular routes, which would have only a marginal effect on our measures of accessibility.

Our study has found some evidence that access to recreational facilities by different forms of transport is associated with obesity, and whilst we did not find this for all facilities, this may reflect the limitations of our data. Improving access to, and knowledge of, health-promoting facilities for those on low incomes may offer an important contribution to reducing the population prevalence of obesity.

BMI, body mass index; DQI, Dietary Quality Index; DZs, Data zones; GIS, Geographic Information System; Km, kilometres; NS-SEC, National Statistics Socio-economic Classification; PA, physical activity