Introduction

It is well-established that sufficient levels of physical activity aid in the reduction of chronic disease, and mortality [1–4]. The United Kingdom (UK) government currently recommends that adults (19–64 years old) are active daily and should accumulate either ≥150 minutes of moderate activity (MPA; e.g. cycling), 75 minutes of vigorous activity (VPA; e.g. running), or a combination, over the week [5]. Additionally, walking 10,000 steps per day is recommended to obtain health benefits [6,7].

However, adults in the UK have low levels of physical activity, even when compared with similar European countries [8,9]. For example, accelerometer data from the Health Survey for England 2008 showed that only 6% of males and 4% of females met national guidelines [10]. Levels of inactivity (i.e. <30 minutes/week of MPA) are particularly high in deprived groups: self-reported data from 163,099 adults in England revealed that levels of inactivity were almost 10% higher in local authorities with the highest levels of socio-economic deprivation compared with the lowest [9]. Even small increases in MPA may be beneficial for inactive groups. A European cohort study including 334,161 adults estimated that moving individuals from inactivity to moderate activity (equivalent to a daily 20-minute walk) produced a 7.35% reduction in all-cause mortality [3].

Understanding the determinants of physical activity, particularly in populations at higher risk of inactivity, could contribute to the development of interventions to increase physical activity. Socioecological approaches to physical activity posit that individual characteristics, the social environment, the physical environment and policies are all key determinants of activity, which are interrelated and embedded in a complex system [11–13]. In addition to having simultaneous independent effects, it is postulated that these influences can also have interactive effects. Growing evidence suggests independent effects of the objectively-measured and perceived aspects of the neighbourhood physical and social environment on physical activity (e.g. [14–17]). However, as Gubbels et al. highlight, while there is a move towards the integration of influences through the use of multivariate models, there is still very limited research examining the interaction of factors [18].

Addressing this gap in the literature is an important next step for physical activity research. Firstly, accounting for potential variance arising from hitherto unmeasured interactive effects could help to explain inconsistent results in the literature examining neighbourhood effects. For example, inconsistent effects of crime and aesthetics on physical activity could arise from interactive effects of other aspects of the social and physical environment suppressing or heightening measured effects within certain environmental conditions [19–24]. Secondly, understanding pathways of influence between environmental variables and physical activity could inform development of interventions to create activity-supportive environments in specific contexts, embracing the complexity of environmental influences and ensuring that unmeasured features of the environment do not suppress effects [25]. Examining interactive effects is required to address these points and to test a central tenant of socioecological models of physical activity.

Recent research examining the health effect of micro-scale features of the environment (e.g. disorder, aesthetics, street lighting) in deprived neighbourhoods suggests that disparities in these features between deprived and non-deprived neighbourhoods could contribute to inequalities in health and behaviours [26,27]. Geographic Information Systems (GIS) measures and neighbourhood audits of streetscapes in New York City revealed that more deprived neighbourhoods had poorer aesthetics and safety conditions (e.g. fewer clean streets, fewer landmarked buildings, more felony complaints) than non-deprived neighbourhoods, potentially offsetting macro-scale elements of deprived urban neighbourhoods that ostensibly create walkable environments (e.g. increased land use mix or density) [27]. Results were corroborated by further research in USA demonstrating poorer aesthetics and worse maintenance of the environment (e.g. litter, vandalism) in deprived neighbourhoods with larger populations of racial and ethnic minorities than in higher-income neighbourhoods [26]. Variation in neighbourhood quality between 3 cities (Seattle, San Diego and Baltimore) underscored the necessity of context-specific investigation [26]. Both studies recommended further research examining how features of social and physical environments impact physical activity behaviours, and whether they interact to produce effects, particularly in deprived communities [26,27].

Emerging research in the USA explores interactive associations. Bracy et al. presented few significant interactive effects of perceived safety and perceived physical environment features on objectively-measured moderate-to-vigorous physical activity (MVPA) in a sample from Washington metropolitan areas [28]. However, there was an interaction between perceived safety and walkability: adults who felt unsafe (i.e. reported low levels of perceived safety) and lived in a neighbourhood with low walkability achieved 91.2 fewer minutes of MVPA per week than adults who felt unsafe but lived in a highly-walkable neighbourhood with low walkability. The difference in MVPA between adults living in neighbourhoods with high and low walkability was markedly less (38.8 minutes) among adults reporting high levels of perceived safety, suggesting feelings of safety might have mitigated the effects of walkability. To the best of our knowledge, there is no research examining interactive environmental effects on physical activity in a deprived context in the UK. It is possible that within this context, the social environment and micro-scale features of the physical environment interact in creating an environment which supports physical activity by altering the way in which the space is used or perceived. As such, environmental features could operate synergistically or a feature could modify the effect of the other through mediation or moderation. For example, an individual with higher levels of social interaction in the neighbourhood may be feel more inclined to be active with friends or walk to a neighbour’s house when the local physical environment is attractive and aesthetically-pleasing than when it is not, while the quality of local physical environment has less influence on individuals with fewer social contacts with which to be active. Alternatively, an environment that is clean and orderly might suggest residents abide by social norms, generating feelings of safety and creating an environment in which individuals feel comfortable walking (e.g. [29]).

This study aimed to explore independent and interactive effects of social and physical environmental factors on neighbourhood walking and MPA in adults in income-deprived communities. Focus was on the quality of micro-scale features of the physical environment (disorder, maintenance and aesthetics) and social environment (cohesion, trust, social interaction, social support, participation and safety), furthering previously discussed research exploring the independent effects of these components in deprived settings elsewhere [26,27]. It was hypothesised that higher quality social and physical environments would be independently and interactively associated with increased physical activity.

Methods

Measures

Neighbourhood walking was assessed in the GoWell community survey using the item: ‘In a typical week, on how many days do you go for a walk around your neighbourhood?’ This item did not distinguish between walking for recreational or utilitarian purposes and captured frequency rather than duration of walking periods. Participation in MPA was measured using the item: ‘In a typical week, on how many days do you do 30 minutes of moderate physical exercise such as brisk walking, cleaning the house–it doesn’t have to be 30 minutes all at once?’. Responses were collapsed into two binary variables using 5 days as a cut-off (walking/participating in MPA on at least 5 days/week; equivalent to >150 minutes MPA/week), in order to assess whether participants were meeting national recommended guidelines for physical activity [5]. Although participation in vigorous physical activity (VPA) was also measured in the survey, levels of VPA were very low (5% participating in ≥5 days/week). Therefore, only MPA was examined in this study.

The quality of the neighbourhood social environment was self-reported using the GoWell community survey. Items were drawn or adapted from previous surveys including the Home Office Citizenship Survey [34], Scottish Social Attitudes Survey [35], British Social Attitudes Survey [36], Office for National Statistics Measuring Social Capital in the UK [37] and the SHARP Questionnaire [38]. Responses to items assessing diverse aspects of the quality of the social environment (e.g. social support: ‘how many people could you ask to give you advice and support in a crisis?’; community cohesion: ‘to what extent do you agree that this neighbourhood is a place where people from different backgrounds get on well together?’; social interaction: ‘On how many days a week do you speak to your neighbours?’) were scored on 4-6-point Likert scales. The neighbourhood was described to participants as the area within 5–10 minutes’ walk from their home.

A neighbourhood audit collected data on the quality of the physical environment. The audit was conducted by two trained surveyors across 95 randomly-selected postcodes within the GoWell study neighbourhoods. Evaluations pertained to the 100-metre area surrounding the central point of the postcode (comprising the ‘audit site’), encompassing streets, buildings, gardens, paths, fences, outdoor communal areas and public spaces. Items selected from the audit assessed the quality of the physical environment in terms of aesthetics, maintenance and disorder (e.g. ‘Are buildings marked with graffiti or other signs of vandalism?’ and ‘Buildings are clean and fresh looking’). Items were scored on 4-point Likert scales. Distance to the central point of the nearest audit site was calculated for each participant to permit data linkage (median distance: 151 metres). Audit data were aggregated at the level of the audit site, creating 95 data points.

A principal components analysis was previously conducted on items measuring the social environment and physical environment using this sample. Items measuring the social environment were drawn from the GoWell community survey and items measuring the physical environment were drawn from the neighbourhood audit. Items were retained on a factor if the loading in the pattern matrix was >0.4. Reliability of the scale was deemed satisfactory if ≥0.5, consistent with recommended levels for scales with few items [39]. The analysis obtained 4 factors assessing the quality of the neighbourhood social environment: ‘social support’ (Cronbach’s alpha = 0.9); ‘social interaction’ (Cronbach’s alpha = 0.7); ‘trust and empowerment’ (Cronbach’s alpha = 0.6); ‘cohesion and safety’ (Cronbach’s alpha = 0.5) and 3 factors assessing the quality of the neighbourhood physical environment: ‘aesthetics of built form’ (Cronbach’s alpha = 0.8); ‘cues of physical disorder’ (Cronbach’s alpha = 0.8) and ‘aesthetics and maintenance of open space’ (Cronbach’s alpha = 0.6). Items for each factor are presented in Table 1. Each factor was scored from 0.0 to 1.0. These factors were included in analyses and are collectively referred to as environmental factors.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Items comprising environmental factors.

https://doi.org/10.1371/journal.pone.0188962.t001

Socio-demographics previously found to be associated with environmental factors were included as covariates: sex, tenure, age, citizenship, employment status and neighbourhood income-deprivation (% residents in receipt of income-related benefits). Mobility-limiting illness and vehicle ownership were also included as covariates owing to their possible relationship with physical activity [40,41]. All socio-demographics were self-reported in the GoWell community survey. Neighbourhood income-deprivation was previously calculated at neighbourhood level using the Scottish Index of Multiple Deprivation (SIMD) income deprivation domain [42].

Results

Complete data were available for 5,923 participants. Numbers of missing values were low because data were collected face-to-face (<2% of the original sample of 6,008 participants were excluded because of incomplete data). Imputation of missing data was deemed inappropriate owing to the very small number of excluded participants which would have an inconsequential effect on statistical inferences [44]. Participant characteristics are presented in Table 2. There were slightly more females than males, British citizens than non-British citizens, unemployed individuals than employed or retired individuals, and more individuals living in social- or private-rented accommodation than those in owner-occupied accommodation.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Participant characteristics and differences in walking and MPA by socio-demographics (n = 5,923).

https://doi.org/10.1371/journal.pone.0188962.t002

Only 29.4% of participants reported walking in the neighbourhood on at least 5 days/week and 23.5% reported participating in MPA on at least 5 days/week. Frequent walking was associated with participation in MPA (X²(1) = 955.49, p<0.001). Younger participants were more likely to report frequent walking and MPA, with a linear trend across the age groups. Participants in and out of employment were significantly more likely to walk or perform MPA on at least 5 days/week compared with retired participants and those with vehicles, were more likely to engage in frequent walking and MPA than others were. Those in family households were more likely to report frequent walking and MPA compared with those living in adult or older adult households, as were participants residing in owned accommodation compared with rented accommodation. Female participants were also significantly more likely to perform MPA on ≥5 days/week, as were British participants.

Associations between social environmental factors and physical activity

Table 3 presents findings for model 1 (containing main effects of environmental factors separately and adjusting for confounders) and model 2 (multivariate model containing main effects for all environmental factors, but no interaction terms, and adjusting for confounders) with walking as the outcome. Models were also conducted using continuous physical activity outcomes; associations were in the same direction and therefore are not reported here. In model 1, independent effects of all social factors were obtained, in the direction expected, i.e. stronger social factors associated with more walking. In model 2 (the multivariate model), three social factors retained significant positive associations with walking, they were: ‘social support’ (OR:1.22, 95%CI = 1.06–1.41, p<0.01), ‘social interaction’ (OR:1.25, 95%CI = 1.10–1.42, p<0.01) and ‘cohesion and safety’ (OR:1.78, 95%CI = 1.56–2.03, p<0.001). There was no effect of ‘trust and empowerment’ on walking in the multivariate model.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Independent effects of social and physical environment factors on neighbourhood walking on at least 5 days/week (n = 5,923).

https://doi.org/10.1371/journal.pone.0188962.t003

Table 4 presents findings for participation in MPA on ≥5 days/week. In model 1, there was a significant effect of ‘social interaction’ and ‘cohesion and safety’, only, both in the direction expected, i.e. stronger social factors associated with more physical activity. In model 2, there was an independent effect of three social factors: two were in the direction expected, namely ‘social interaction’ (OR:6.16, 95%CI = 5.14–7.37, p<0.001) and ‘cohesion and safety’ (OR:1.93, 95%CI = 1.65–2.27, p<0.001); one association was negative, namely ‘social support’ (OR:0.79, 95%CI = 0.67–0.94, p<0.01). There was no effect of ‘trust and empowerment’ in the multivariate model.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Independent effects of social and physical environment factors on moderate physical activity on at least 5 days/week (n = 5,923).

https://doi.org/10.1371/journal.pone.0188962.t004

Discussion

In an adult population living in deprived communities in Glasgow, UK, independent and interactive effects of the social and physical environment on neighbourhood walking and participation in MPA were revealed. In models including a single environmental factor and adjusting for covariates and nesting in sub-area, all environment factors (‘social support’, ‘trust and empowerment’, ‘social interaction’, ‘cohesion and safety’, ‘aesthetics of built form’, ‘physical disorder’ and ‘aesthetics and maintenance of open space’) were significantly associated with increased likelihood of walking in the neighbourhood on ≥5 days/week. ‘Cohesion and safety’ had the largest effect, with participants reporting higher levels of cohesion being nearly twice as likely to regularly walk around their neighbourhood. In models with MPA as the outcome, ‘social interaction’, ‘cohesion and safety’, ‘physical disorder’ and ‘aesthetics and maintenance of open space’ were significant. ‘Social interaction’ had the largest independent effect on MPA, with participants reporting higher levels of ‘social interaction’ being more than 6 times more likely to meet national guidelines by participating in >30 minutes of MPA on ≥5 days/week. Interactive effects on walking and MPA were reported between i) ‘trust and empowerment’ and ‘aesthetics and maintenance of open space’ and ii) ‘cohesion and safety’ and ‘physical disorder’. There was also an interactive effect on MPA between ‘social interaction’ and ‘aesthetics or built form’.

Most independent effects were in the expected direction, with better quality social and physical environments related to an increased likelihood of frequent walking or MPA. The only negative relationship was between ‘social support’ and MPA which was non-significant in a univariate model (while also controlling for covariates) but attained significance in the multivariate model. This could be a spurious result (despite an alpha level p<0.01) or could be indicative of reverse causality whereby individuals with a greater need for social support (and therefore higher self-reported levels) are less able or likely to participate in higher-intensity physical activity. Results replicate previous findings from the GoWell sample where individual measures of safety after dark, informal social control (intervening in harassment), belonging to the neighbourhood and cohesion between residents of different backgrounds had significant positive independent effects on neighbourhood walking [45]. Findings also support those from other populations. For example, Shelton et al. found in a sample of 1,112 adults in low-income communities in the USA that participants who reported stronger social networks (and thereby higher levels of social interaction) also performed higher levels of pedometer-assessed physical activity [46]. Positive associations between objectively-measured aesthetics of the built form and open space and self-reported walking reflect previous findings from the IPEN study across 17 cities in 12 countries which found an effect of self-reported perceived aesthetics and self-reported walking for transport for ≥150 minutes per week [47]. As noted in Kerr et al., an effect of micro-scale features of the physical environment such as aesthetics presents a potentially low-cost intervention strategy [47].

There were some notable differences between independent environmental effects on neighbourhood-based walking and MPA. In models including all environmental factors and covariates, there was an effect of all social environmental factors except ‘trust and empowerment’ on both activity outcomes. However, effects on MPA, compared with neighbourhood-based walking, were much stronger for ‘social interaction’, slighter stronger for ‘cohesion and safety’ and in the opposite direction for ‘social support’. This is somewhat surprising as it might be expected that if there are neighbourhood effects on activity, the local neighbourhood environment would be more closely associated with activity performed in the neighbourhood. However, it is likely that although the measure of MPA was not context-specific, the majority of activity reported by participants was in fact performed in the neighbourhood and encompassed walking and additional activities such as household chores, gardening or using local physical activity facilities [48]. There were also differences in the effect of the built environment on activity outcomes:–‘physical disorder’ only had an independent effect on MPA while ‘aesthetics of built form’ and ‘aesthetics and maintenance of open space’ only had an effect on walking. It is possible these differences can be attributed to the activities included in MPA other than walking, such as use of physical activity facilities, and the comparative importance of a pleasant and attractive public realm for walking-based activities. Differences might be explained by the extent to which the physical environment interacted with the social environment, as discussed below.

Interactions between i) ‘trust and empowerment’ and ‘aesthetics and maintenance of open space’ and ii) ‘cohesion and safety’ and ‘physical disorder’ were found for both walking and MPA. An interactive effect of ‘social interaction’ and ‘aesthetics of built form’ was also reported for MPA. Post hoc analyses suggested that the effect of ‘physical disorder’ appeared to operate through ‘cohesion and safety’, with fewer cues of disorder supporting higher levels of activity only when participants viewed their neighbourhoods as socially-cohesive and safe. Post-hoc tests showed that the largest effect was observed for participants reporting high levels of ‘cohesion and safety’ and living in areas with fewer cues of ‘physical disorder’. These participants were 2.2 times more likely to walk or perform MPA on at least 5 days/week than participants living in areas with fewer cues of ‘physical disorder’ but reporting lower levels of ‘cohesion and safety’. There was no independent effect of ‘physical disorder’.

The effect of ‘aesthetics of built form’ on MPA was similarly constrained by ‘social interaction’: only when participants had higher levels of social interaction did the positive influence of aesthetics of the built form occur–there was no effect of ‘aesthetics of built form’ when ‘social interaction’ levels were lower. The largest effect was apparent when both ‘social interaction’ and ‘aesthetics of built form’ were high, in this case, participants were 10 times more likely to perform MPA on 5 days/week, than participants reporting low levels of ‘social interaction‘ in areas with better ‘aesthetics of built form’.

Previous research has documented an association between physical disorder and anti-social behaviour [49,50], speculating that cues of neglect and ‘physical incivilities’ suggest residents do not abide by social norms and are therefore more likely to engage in anti-social or criminal behaviour [29]. It is therefore feasible that physical disorder acts as a cue to harmonious relationships between neighbours, or that harmonious relationships encourage individuals to take care of their physical environment. Findings from this study suggests that when individuals perceived low levels of safety and cohesion there was no effect of physical disorder, possibly because cues for safety and cohesion that were usually elicited from physical disorder were overridden. However, physical disorder had an effect on physical activity in individuals reporting high levels of safety and cohesion; possibly disorder was neutralised as a cue for safety but still created an uninviting, smelly or dirty environment.

Additionally, it could be speculated from results that aesthetically-pleasing built form only supported MPA when individuals had many friends and neighbours because visually interesting and attractive environments and buildings were only important when they could be used for recreational team sports or joint activities (e.g. an attractive green space or leisure centre). Because assessments of the built form were objectively reported by auditors, it suggests that an interactive effect is not due to individuals with higher levels of social interaction feeling more attached to the neighbourhood and therefore perceiving the local physical environment in a more positive light. However, it is possible that an attractive built form might be more readily felt by individuals with higher levels of social interaction, who then are more motivated to use, and be active in, their local environment.

These interactive effects support a key tenet of socioecological models of physical activity and are substantiated by emerging evidence from the USA. For example, perceived safety appeared to interact with walkability, mitigating the effect of neighbourhood walkability whereby walkability exerted a smaller effect on MVPA when levels of perceived safety were held constant (e.g. constantly high) than did the effect of safety when levels of walkability were held constant [28]. Such findings suggest that aspects of the social environment may be more important than physical aspects in encouraging individuals to be active within a deprived context. Intervening to increase physical activity through reduction of physical disorder or creation of more attractive and interesting built form might therefore be successful only when the social environment is also supportive of physical activity.

In the current study, ‘trust and empowerment’ and ‘aesthetics and maintenance of open space’ appeared to act synergistically upon walking and MPA: ‘aesthetics and maintenance of open space’ was only important when there was a high level of ‘trust and empowerment’ and vice versa. It is feasible that these aspects of the social and physical environment reinforce one another. For example, residents who trust each other and feel empowered may have the capability and motivation to maintain and/or advocate for attractive open space, resulting in attractive open spaces which prompt further mutual trust and empowerment. Similarly, in Kaczynski and Glover’s study of 380 adults in Canada, the highest levels of walking for transport or recreation were reported in neighbourhoods with high walkability (a composite measure including macro- and micro-scale physical features) and high levels of cohesion and trust [51]. To our knowledge, this is the first time an interactive effect of community trust and empowerment and quality of open space on physical activity has been reported in a deprived setting in the UK.

Findings should be replicated and further explored within deprived contexts and using longitudinal or quasi-experimental study designs in order to establish the reliability of associations and unpick the direction of causality. If supported by future research, interactive environmental effects on physical activity may have important implications for future research and policy within the field of active design. Firstly, if unmeasured factors moderate or act synergistically upon measured factors, variance arising from the influence of unmeasured factors might underpin some of the reported inconsistency within the physical activity literature examining neighbourhood effects. This is highlighted by the independent effects reported in the multivariate models that did not include interactive effects. Insight into these relationships can contribute to the development of theoretical frameworks of neighbourhood influence on physical activity.

Secondly, as discussed, it is possible that simultaneously targeting aspects of social and physical environment that work together to influence activity will harness the largest effects on walking and MPA. For example, interventions may be most successful when they simultaneously target aspects of the environment or deploy specific strategies in neighbourhoods with existing social/physical environmental characteristics conducive to activity. Findings are therefore supportive of a holistic approach to regeneration which includes strategies for both social and physical regeneration or leverages the effect of existing supportive social contexts.

Conclusions

Findings reveal independent and interactive associations between the quality of the social and physical environment and neighbourhood walking and MPA. Results demonstrate the importance of simultaneous consideration of multiple aspects of the social and physical environment in both research and active design policy. To our knowledge, this is the first examination of interactive effects of the social and physical environment in a sample of adults living in income-deprived neighbourhoods in the UK.

Supporting information

Acknowledgments

The views expressed are those of the authors and not necessarily the sponsors. The authors would like to thank the individuals participating in GoWell.