Preventive Medicine 51 (2010) 157–163

Contents lists available at ScienceDirect

Preventive Medicine

j ourna l homepage: www.e lsev ie r.com/ locate /ypmed

Relative importance of physical and social aspects of perceived neighbourhoodenvironment for self-reported health

Christopher Gidlow a,⁎, Thomas Cochrane a, Rachel C. Davey b, Graham Smith c, Jon Fairburn c

a Centre for Sport Health and Exercise Research, Faculty of Health, Staffordshire University, Stoke-on-Trent, England, ST4 2DF, UKb Centre for Research and Action in Public Health, Faculty of Health, University of Canberra, ACT 2601, Australiac Institute for Environment, Sustainability and Regeneration, Faculty of Sciences, Staffordshire University, Stoke-on-Trent, England, ST4 2DE, UK

⁎ Corresponding author. Centre for Sport, Health andHealth, Staffordshire University, Brindley Building, LeekUK.

E-mail address: c.gidlow@staffs.ac.uk (C. Gidlow).

0091-7435/$ – see front matter © 2010 Elsevier Inc. Adoi:10.1016/j.ypmed.2010.05.006

a b s t r a c t

a r t i c l e i n f o

Available online 24 May 2010

Keywords:HealthNeighbourhoodPhysical environmentSocial environment

Objective. To explore the relative importance of the perceived physical and social neighbourhoodenvironment for physical and mental health.

Methods. A representative random sample of adults was recruited from 10 areas across Stoke-on-Trent,UK (June–September 2007). Interview-administered surveys were used to record data on the perceivedneighbourhood environment (physical and social), self-reported health, and socio-demographics. Multipleregression analysis was used to explore independent associations between environmental factors and

physical and mental health.

Results. Independent physical and social environmental factors respectively explained 6.0% and 3.2% ofvariability in physical health and 2.8% and 4.4% of variability in mental health. Diversity of land use was thestrongest and only physical environmental predictor of physical health (Beta=0.27, pb0.001), explainingmore variability than social environmental factors combined. Conversely, social support was the mostimportant (and only) social environmental factor for mental health (Beta=−0.20, pb0.001); again, thisexplained more variability in mental health than the combined effect of four physical environmentalpredictors.

Conclusion. Perceived physical and social environmental characteristics were important for physical andmental health, independent of socio-demographic factors. Living in neighbourhoods with greater land usediversity appears particularly important for physical health, whereas social support appears more closelylinked to mental health.

© 2010 Elsevier Inc. All rights reserved.

Introduction

The importance of ‘place’ for health is consistently demonstratedthrough geographical health inequalities (de Hollander and Staatsen,2003; Ellaway et al., 2001; Pickett and Pearl, 2001). Socio-economiccircumstance has often been used to characterise neighbourhoodeffects (O'Campo et al., 2009), but there is growing recognition of theneed to consider the health implications of other environmentalfactors (Jones et al., 2007). The social ecologymodel, which has gainedprominence in health promotion, asserts that health is influenced byvarious facets of both the physical and the social environment, inaddition to various personal attributes (Stokols, 1992). This notionthat physical, social, and emotional (or mental) well-being is theresult of ‘people's transactions with their physical and socio-cultural

Exercise Research, Faculty ofRoad, Stoke-on-Trent, ST4 2DF,

ll rights reserved.

environments’ (Sallis et al., 1998, p. 380) provides the rationale forresearch that explores the links between health and various attributesof the neighbourhood environment.

As an environment with which people interact on a daily basis, the‘neighbourhood’ is of growing public health interest as a potentialmeans of contextually sensitive intervention (Cummins et al., 2007).Studies of the associations between environment and health ofteninvolve combining existing area-level environmental data withpopulation health data (Cummins et al., 2005; Latkin et al., 2009;Riva et al., 2007), such as those exploring health effects of exposure togreen space (Maas et al., 2006; Mitchell and Popham, 2007; Takano etal., 2002) or urban sprawl (Ewing et al., 2003).

Using individual-level data on perceptions, we can begin tounderstand the health implications of peoples' experiences of livingin their neighbourhood. Consistent with the social ecology model,features of the neighbourhood environment can be broadly classifiedinto physical (e.g., residential density, land use, green space) and socialcharacteristics (e.g., crime, trust, reciprocity). There is some evidencelinking individual residents' perceptions of their physical neighbour-hood environment with health. Using data from Australian adults,

158 C. Gidlow et al. / Preventive Medicine 51 (2010) 157–163

Leslie and Cerin (2008) observed associations between numerousperceived physical environmental factors (e.g., aesthetics and green-ery, land use mix diversity, street connectivity) and neighbourhoodsatisfaction, which were in turn linked to mental health. Data from alargeWelsh cohort similarly found independent associations betweenperceived physical environmental characteristics and self-reportedgeneral health (Poortinga et al., 2007). Yet in both cases, the socialenvironment also emerged as important. Although the physicalenvironment can be increasingly considered as a potential supportor deterrent for health behaviour (NICE, 2008; Jones et al., 2007), thesocial environment has often been the focus of investigations ofperceived environment and health, especially mental health.

The social environment is multi-faceted, variously defined (e.g.,social support, social capital, social cohesion), and consequently,much debated both in concept and measurement (Almedom, 2005;De Silva et al., 2005). Nevertheless, it is quite reasonable that a personwill experience better health if they live in an area in which they aresocially active, feel supported, safe, and trust their neighbours. Someresearchers have demonstrated such links, but more commonlylinking social environment with mental health (Baum et al., 2009;Veenstra et al., 2005). Indeed, most analyses of associations betweenperceived neighbourhood environment and self-reported health havetended to focus on mental health (Almedom, 2005; Kruger et al.,2007; Leslie and Cerin, 2008; O'Campo et al., 2009) or use a measureof general/global health (Poortinga et al., 2007; Sundquist and Yang,2007; Wen et al., 2006).

Using data collected from a representative random sample ofadults residing in Stoke-on-Trent, UK, we explored relationshipsbetween aspects of the perceived neighbourhood environment andhealth. In keeping with the social ecology model (Stokols, 1992), (i)neighbourhood environment was defined using a range of physicaland social characteristics, in addition to socio-demographics to reflectrelevant personal attributes, and (ii) their relative associations withhealth were explored for both physical and mental health outcomes.

Methods

Study area

This study was conducted in Stoke-on-Trent, UK, a medium sizedconurbation with a population approaching 240,000. Many areas of the cityare considered deprived. Approximately one-third of neighbourhood areasfall within the most deprived decile of national rankings (Communities andLocal Government, 2007). Stoke-on-Trent has correspondingly high rates ofmorbidity and premature mortality (National Health Service, 2008) and pooroutcomes for lifestyle (National Health Service, 2008; Sport England, 2007).Despite greater than average green space provision, Stoke-on-Trent ispredominantly urban, characterised in many places by dense networks ofterraced (or row) houses; it has a pervasive road network, and has lowlevels of active transport as highlighted by previous work (Cochrane et al.,2009).

Sample

Detailed accounts of sample recruitment and data collection protocolshave been reported elsewhere (Cochrane et al., 2009; Davey et al., 2008).Briefly, Stoke-on-Trent comprises 160 Lower Level Super Output Areas(LSOA), with a mean population of approximately 1500 per area. Our adultsample was randomly selected from the 25 LSOAs which were not earmarkedfor regeneration, were not adjacent to one another or the city boundaries, andhad low population churn rates (i.e., would be appropriate for intervention).Our sample size estimates indicated a requirement for 10 clusters. Thus, 10LSOAs were chosen in 5 matched (by deprivation decile) non-adjacent pairsfrom the most deprived six deciles of deprivation (Cochrane et al., 2009). Foreach LSOA, 170 addresses were selected at random from the potential 1700using the Postcode Address File for England, excluding non-residentialaddresses. For those comprised of several dwelling units (e.g., blocks of flats),one dwelling unit was selected randomly using a Kish grid. At each eligible

address, one adult (aged 16 years or older) was selected as a respondent tothe survey, again using a Kish grid.

Procedures

Data collection was undertaken between June and September 2007.Willing participants were asked to complete an interview-administeredsurvey, which was carried out by the National Centre for Social Research(NatCen) using a computer-assisted personal interviewing (CAPI) approach.Interviews typically lasted 45 min, which included height and weightmeasurements. Full details of the interview procedures are available in thesurvey technical report (Speight et al., 2007). The study was approved by theStaffordshire University Research Ethics Committee.

Measures

Self-reported healthThe Short Form 12 (Quality Metric, 2006) generic health questionnaire

comprised 12 individual items on functional health and well-being. Scoringsoftware was used to create two composite scores for physical and mentalhealth: the physical component score (PCS) and the mental component score(MCS).

Socio-demographicsIndividual socio-demographic variables collected included gender, age,

ethnicity, occupational socio-economic classification (Office of NationalStatistics, 2000), employment status, education level, household tenure,and income. Question formats were those used in the decennial populationcensus (England) (Office for National Statistics, 2001). Neighbourhooddeprivation was ranked by decile from 1 (most deprived) to 10 (leastdeprived) according to individuals' resident LSOA using the Index of MultipleDeprivation 2004 (latest version not available at the time of study)(Communities and Local Government, 2004).

Perceived neighbourhood environment

(i) The Abbreviated Neighbourhood Environment Walkability Scale(ANEWS) (Cerin et al., 2006) was used with minor adaptations foruse in a UK context. This comprised 40 items: 17 scored on a 5-pointordinal scale and 23 scored on a 4-point ordinal scale expressingdegree of agreement with a given statement. Individual items wereprocessed creating scores for various domains of the physical andsocial environment. Only those considered conceptually relevant tohealth were included in analysis (Leslie and Cerin, 2008): physicalenvironment—residential density, land use mix, access to services,safety (traffic), street connectivity, physical barriers to walking,aesthetics, infrastructure for walking and cycling, and hilliness(excluded—lack of cul-de-sacs and lack of parking); social environ-ment—safety (crime).

(ii) Social Capital and Social Exclusion Condensed module developed forthe Health Survey for England (Bajekal and Purdon, 2001) was used toprovide further measures of the social environment. This comprised12 items for social capital (8 scored on a 4-point ordinal scaleexpressing degree of agreement with a given statement and 4dichotomous items) and seven items for social support, all scoredon a 3-point ordinal scale expressing the degree of agreement with agiven statement.

Statistical analyses

The purpose of analysis was to explore the relative importance ofindividuals' perceptions of their physical and social neighbourhood environ-ment for self-reported physical and mental health, independent of socio-demographics. Inspection of the two dependent variables derived from theSF12 (v2) health questions revealed significant negative skew for PCS andMCS. Therefore, both were inverse transformed; i.e., reflected (1−P/MCS+sample max. for P/MCS) and square root transformed. Inverse transformeddependent variables were used in all analyses. Intra-class correlationcoefficients calculated from individual and area-level variance for bothdependent variables (ICC for PCS=0.0152; ICC for MCS=0.0357) demon-strated a lack of clustering of dependent variables in the 10 study areas (i.e.,within-area variancewasmuch greater than the between-area variance). Thisprovided no justification for multi-level analysis. To explore further the

Table 2Sample versus normative health scores (n=761 Stoke-on-Trent, UK; June–September2007).

Age group(years)

n Sample Englanda

PCS MCS PCS MCS

16–29 134 54.03⁎⁎ 50.80⁎⁎ 52.14 49.2130–49 274 51.11 48.94 50.87 49.6250–64 179 46.41 49.35⁎ 46.21 51.5365–74 102 40.33 50.80 — —

≥75 72 34.95 49.13 — —

PCS, physical component score (self-reported physical health); MCS, mentalcomponent score (self-reported mental health).

a Population norms take from Jenkinson et al. (2001).⁎ pb0.05, represents significance level from one-sample test comparing sample with

available normative data.⁎⁎ pb0.01, represents significance level from one-sample test comparing sample withavailable normative data.

159C. Gidlow et al. / Preventive Medicine 51 (2010) 157–163

possibility that any relationships in analysis were affected by spatialdependencies (i.e., spatial autocorrelation) (Browning et al., 2003), the PCSscores were tested for spatial dependency using a Moran's I test. The Moran I-statistic was not significant (I=0.007, p=0.862), suggesting minimal spatialautocorrelation among neighbourhoods with respect to self-rated health.

For both PCS and MCS, multivariate linear regression models thatexplored the explanatory power of the three categories of independentvariable (physical environment, social environment, and socio-demograph-ics) were developed. First, individual item responses were examinedseparately and considered for inclusion in the model if they demonstrated alinear relationship with the outcome measure (inspection of scatter plots—continuous variables, error bar plots—ordinal variables) and significantcorrelation (pb0.05). Second, separate multiple linear regression modelswere developed for each of the three independent variable categories. Whereapplicable, summary scores and individual items from constituent ques-tionnaires were explored and those demonstrating the strongest relation-ships were included (e.g., the land use diversity summary score from ANEWSwas included as it demonstrated a stronger relationship with dependentvariables than any of its constituent items). A variable was only included if itadded significant explanatory power to the model (significant adjusted R2

change). Multi-collinearity between independent variables retained inmodels was not observed (regression diagnostics showed that the ConditionIndices were acceptable, b30). Finally, individual items retained in theseseparate sub-models were then entered as three blocks into the final stagelinear regression model (for each dependent variable). Analyses wereundertaken using SPSS 16.0.

Results

Sample

As previously reported (Cochrane et al., 2009), the 761 productiveinterviews obtained represented a 49% response rate. The samplecomprised slightly more women than men (55% vs. 45%), with goodcoverage of ages and relatively low ethnic diversity typical of the city(Table 1). The majority of the population was overweight (65%), oflow educational attainment (72%), and with modest income consis-tent with the relative deprivation in much of Stoke-on-Trent. Self-reported health data for the sample were comparable with availablenormative data for the English adult population (Jenkinson et al.,2001) (Table 2).

Table 1Total sample characteristics (n=761 Stoke-on-Trent, UK; June–September 2007).

Characteristic Category n %

Gender Male 343 45.1Female 418 54.9

Age group 15–24 76 10.025–44 276 36.345–64 235 30.965+ 174 22.9

Ethnicity Caucasian 709 93.2Other ethnic background 52 6.9

Weight (BMI) category Underweight 10 1.3Healthy range 256 33.6Overweight/obese 495 65.1

Education None 3 0.4Secondary school (≤age 16 years) 547 71.9Further or higher education 211 27.8

Employment status Paid/unpaid work outside home 435 57.2No work outside the home 326 42.8

Annual household income b £10,000 178 23.5£10,000 –b£20,000 167 22.0£20,000–b£30,000 199 26.0≥£40,000 72 9.5Not answered 145 19.0

Housing tenurea Own outright 160 21.0Mortgage 82 10.8Private rented 291 38.2Social rented/rent free 227 29.8

a Data not complete for all cases (nb761).

Physical health

Regression analysis for each category of independent variableidentified 10 variables with significant independent predictive powerfor PCS: 3 physical environment, 4 social environment, and 3 socio-demographic (Table 3). Strongest predictors were land use mix(measured as perceived walking time to various services and ameni-ties), age, employment status, and years of residence in the local area.

Table 5 shows the final regression model. Six independentvariables remained, accounting for 35.6% of variance in physicalhealth (26.4% socio-demographic, 6.0% physical environment, and3.2% social environment). When the direction of the association, thenature of the independent variables (i.e., whether high scores indicatebetter or worse outcomes), and the reflected health variable wereconsidered, lower self-reported physical health was associated with

- physical environment (1): lower land use diversity (measured asperceived walking time to various services and amenities)

- social environment (3): lack of trust in local people, notparticipating in any organisations (e.g., residents associations,parent-teacher), and having lived longer in the area

- socio-demographics (2): being older and not being in work.

Mental health

Nine variableswere identified as individual predictors of self-reportedmental health: 4physical environment, 2 social environment, and3socio-demographic (Table 4). Total household income and social supportnetworks were the strongest predictors. Table 6 shows the finalregression analysis in which seven independent variables were retained,explaining 13.9% of variance (6.7% socio-demographic, 2.8% physicalenvironment, and 4.4% social environment). Relationships in Table 6indicate that lower self-reported mental health was associated with

- physical environment (4): higher traffic hazards, lower streetconnectivity, lower land use diversity, and higher residentialdensity

- social environment (1): lower social support- socio-demographics (2): being younger and lower householdincome.

Discussion

Data from a representative random sample of adults from deprivedurban neighbourhoods indicated that self-reported physical andmental health were associated with individuals' perceptions of boththe physical and social neighbourhood environment, independent ofwell-established socio-demographic effects (Acheson, 1998; Depart-ment of Health, 2008).

Table 3Summary of univariate linear regression of individual variables and reflected log of SF12 physical component score (PSC) (n=761 Stoke-on-Trent, UK; June–September 2007;significant predictors only).

Variable category Independent variable Ba SEb Betac pd 95% CIe AdjustedR2f

Physical environment (Constant) 0.62 0.04 b0.001 0.53 — 0.70 0.24Land use mix diversity 0.15 0.01 0.41 b0.001 0.13 — 0.18Hilliness 0.04 0.01 0.15 b0.001 0.02 — 0.05Traffic hazards 0.04 0.01 0.11 0.001 0.02 — 0.06

Social environment (Constant) 0.99 0.04 b0.001 0.92 — 1.07 0.16Crime 0.06 0.01 0.16 b0.001 0.03 0.08Organisational participation −0.46 0.11 −0.15 b0.001 −0.66 — −0.25Total years lived in local area 0.00 0.00 0.31 b0.001 0.00 — 0.01Whether thinks people can be trusted 0.04 0.02 0.08 0.019 0.01 — 0.08

Socio-demographics (Constant) 0.93 0.07 b0.001 0.80 — 1.07 0.26Age of respondent 0.00 0.00 0.30 b0.001 0.00 — 0.01Whether has paid/unpaid work outside home 0.14 0.02 0.27 0.000 0.10 — 0.19Area by IMD −0.01 0.00 −0.08 0.032 −0.01 — −0.00

a Regression coefficients.b Standard error of regression coefficient.c Standardised regression coefficient.d Significance level.e 95% Confidence intervals for Beta coefficient.f Adjusted R2 for the model indicating relative strength of the contribution of the variables.

160 C. Gidlow et al. / Preventive Medicine 51 (2010) 157–163

The relative contribution of physical and social environmentindicators to physical and mental health outcomes was somewhatintuitive; i.e., more variability in physical health was explained byphysical environment indicators, whereas social environmentexplained a greater proportion of variability in mental health.Therefore, despite intrinsic links between physical and mental health,we might need to rethink the tendency to focus on general or mentalhealth outcomes in environmental perception studies.

When comparing strength of relationships with physical health,land use diversity (measured as perceived walking time to variouslocal services and amenities) was the only independent physicalenvironmental predictor but the strongest overall, accounting for 6%of variability. With cross-sectional data, it is not possible to determinethe direction of relationships; is greater perceived distance to localdestinations the result of poor physical health and immobility, or is alack of ‘walkable’ services detrimental to physical health, perhapsthrough reduced opportunities for walking and active transport? Thewide-ranging physical and mental health benefits of physical activityand the concurrent relationship linking physical activity with theproximity of local destinations (Cochrane et al., 2009) or neighbour-hood ‘walkability’ (Cleland et al., 2008; Owen et al., 2007) present a

Table 4Summary of linear regression of each variable category with reflected log of SF12 mental copredictors only).

Variable category Independent variable Ba

Physical environment (Constant) 1.30Traffic hazards 0.03Residential density 0.00Land use mix diversity 0.03Street connectivity −0.03

Social environment (Constant) 1.56Social support −0.10Good neighbourhood (lower scores better) 0.05

Socio-demographics (Constant) 1.51Total Annual Household Income −0.01Age −0.00Housing tenure −0.02

a Regression coefficients.b Standard error of regression coefficient.c Standardised regression coefficient.d Significance level.e 95% Confidence intervals for Beta coefficient.f Adjusted R2 for the model indicating relative strength of the contribution of the variabl

possible mechanism through which physical (and mental) healtheffects could result from differences in land use diversity. To explicatesuch a causal link, however, would require longitudinal data, perhapsfrom ‘natural experiments’ (Petticrew et al., 2005) that seek tomodifyenvironments or associated perceptions, and monitor changes inphysical activity and health.

Social capital indicators of trust and social participation alsoemerged as predictors of physical health. Despite some debate aroundconceptual and measurement issues (Almedom, 2005; De Silva et al.,2005), social capital tends to be measured in terms of participation,social trust, and reciprocity (Cooper et al., 1999). Compared withsocial support, which relates more to perceived support of family andfriends, acquiring ‘social capital’ arguably necessitates greater mobil-ity and physical independence (e.g., to attend groups, clubs, ororganisations). This difference might explain the association withphysical but not mental health reported here. Much of the relatedsocial capital literature, however, explores associations with mentalhealth (Almedom, 2005; Baum et al., 2009; Kruger et al., 2007;Veenstra et al., 2005). For example, O'Campo et al. (2009) reportedconnections between perceived neighbourhood friendliness, sense ofcommunity, interaction with neighbours, and good mental health.

mponent score (MCS) (n=761 Stoke-on-Trent, UK; June–September 2007; significant

SEb Betac pd 95% CIe AdjustedR2f

0.06 b0.001 1.19 — 1.41 0.060.01 0.13 b0.001 0.01 — 0.050.00 0.12 0.001 0.00 — 0.000.01 0.12 0.003 0.01 — 0.050.01 −0.11 0.006 −0.05 — −0.010.06 b0.001 1.44 — 1.69 0.070.02 −0.21 b0.001 −0.13 — −0.060.01 0.15 b0.001 0.02 — 0.070.05 b0.001 1.41 — 1.62 0.070.00 −0.18 0.001 −0.01 — −0.000.00 −0.11 0.025 −0.00 — 0.000.01 −0.11 0.022 −0.03 — −0.00

es.

Table 5Summary of final multivariate linear regression for reflected log of SF12 Physical Component Score (PCS) (n=761 Stoke-on-Trent, UK; June–September 2007; significant predictorsonly).

Ba SEb Betac pd 95% CIe Correlationsf AdjustedR2g

R2 Partial Part

PCS (Constant) 0.75 0.08 b0.001 0.59 — 0.91Socio-demographic Age of respondent 0.00 0.00 0.19 b0.001 0.00 — 0.00 0.17 0.13 0.26

Whether has paid/unpaid work outside home 0.09 0.02 0.17 b0.001 0.04 — 0.13 0.15 0.12Social Organisational participation −0.31 0.10 −0.10 0.002 −0.51 — −0.11 −0.13 −0.10 0.30

Total years lived in local area 0.00 0.00 0.09 0.029 0.00 — 0.00 0.09 0.07Whether thinks people can be trusted −0.04 0.02 −0.08 0.029 −0.08 — −0.00 −0.09 −0.07

Physical Land use mix diversity 0.10 0.01 0.27 b0.001 0.08 — 0.13 0.29 0.24 0.36

a Regression coefficients.b Standard error of regression coefficient.c Standardised regression coefficient.d Significance level.e 95% Confidence intervals for Beta coefficient.f Partial and part correlations.g Adjusted R2 for the model indicating relative strength of the contribution of the variables.

161C. Gidlow et al. / Preventive Medicine 51 (2010) 157–163

Even where associations between social capital and physical healthhave been observed, it is in addition to those for mental health(Fujiwara and Kawachi, 2008). Reasons why we did not find such anassociation are unclear but could relate to differences in population,measurement, or the combination of personal and environmentfactors considered.

In relation tomental health, lower land use diversity was again thestrongest (negative) of several physical environmental factors (streetconnectivity, residential density, and traffic hazards), which collec-tively confirm the negative mental health consequences of living inareas with poor access to local shops/services, that are less wellconnected and have busier roads (e.g., Leslie and Cerin, 2008). Socialsupport (not capital) exhibited the strongest relationshipwithmentalhealth (positive). This accords with evidence of direct health benefits(Berkman, 1984; Kessler et al., 1985) or those conferred indirectlythrough buffering or moderating against the negative healthconsequences of life stressors (Cooper et al., 1999; Stansfield, 2006)thought to be experienced disproportionately in ‘worse’, moredisadvantaged neighbourhoods.

The ‘stress response’ is an often purported and well-evidencedmechanism that suggests health inequalities arise as a result of chronicdaily stressors. Brunner and Marmot (2006) suggest that exposure tosuch stressors can depend on relative socio-economic circumstance(e.g., greater stressors associated with financial strain, unrewardingmonotonous low paid work, and lack of employment control). This

Table 6Summary of final multivariate linear regression for reflected log of SF12Mental Component Sonly).

Ba SEb

MCS (Constant) 1.63 0.09Socio-demographic Age of respondent −0.00 0.00

Total Annual Household Income (gross) −0.00 0.00Social Social support network (higher scores better) −0.09 0.02Physical Land use mix diversity 0.03 0.01

Traffic hazards 0.02 0.01Street connectivity −0.02 0.01Residential density 0.00 0.00

a Regression coefficients.b Standard error of regression coefficient.c Standardised regression coefficient.d Significance level.e 95% Confidence intervals for Beta coefficient.f Partial and part correlations.g Adjusted R2 for the model indicating relative strength of the contribution of the variab

principle can be applied to help understand the detrimental healthconsequences for residents who live in neighbourhoods which theyperceive to be poorly connected, with inadequate facilities/services,untrustworthy neighbours, and a lack of social opportunities andsupport. The apparent associations between individual environmentperceptions and health, but concomitant lack of neighbourhood-levelvariability across the10areas in this study, also raises thepossibility thatdetrimental health consequences of perceived environmental stressors(physical or social) are more pronounced where there is greaterenvironmental inequality (real or perceived) within local communitiesthan between local communities. Again, our understanding of theserelationships could be enhanced through larger studies that involvegreater environmental diversity.

Our findings add to the existing cross-sectional evidence base byexploring the relative associations between perceptions of the neighbour-hoodenvironment andphysical andmental health. Yet, theneed for robustlongitudinal data remains. There are substantial challenges in resourcingand evaluating ecological interventions, but arguably, the evidence isstrong enough to justify intervention as part of rigorously evaluated‘natural experiments’ (Petticrewet al., 2005). Thesemight include physicalmodifications or activities to build community social capital and supportnetworks. There are now examples in which experimental rigour is beingapplied to substantial environmental interventions to promote physicalactivity, such as the Australian RESIDential Environment (RESIDE) study(http://www.sph.uwa.edu.au/research/cbeh/projects/reside) and the

cores (MCS) (n=761 Stoke-on-Trent, UK; June–September 2007; significant predictors

Betac pd 95% CIe Correlationsf AdjustedR2g

Part Partial

b0.001 1.46 — 1.80−0.11 0.022 −0.00 — 0.00 −0.09 −0.09 0.08−0.10 0.044 −0.01 — 0.00 −0.08 −0.08−0.20 b0.001 −0.13 — −0.06 −0.20 −0.19 0.11

0.12 0.008 0.01 — 0.05 0.11 0.10 0.140.10 0.012 0.01 — 0.04 0.10 0.09

−0.08 0.043 −0.04 — −0.00 −0.08 −0.080.08 0.045 0.00 — 0.00 0.08 0.08

les.

162 C. Gidlow et al. / Preventive Medicine 51 (2010) 157–163

iConnect evaluation of Connect2, a programme of cycling/walkinginfrastructure improvements across the UK (http://hsc.uwe.ac.uk/net/research/Default.aspx?pageid=505).

On the basis of our finding that links perceived land use mix withhealth, efforts to improve health through increasing the perceived‘walkability’ of local destinations and to test the possible mediatingeffect of physical activity would be a logical next step. Given theapparent role of perceived environment, this could, for example, includea campaign to promote active travel to local destinations, perhapscombined with or compared to changes arising from improvements totraffic-free routes in similar neighbourhood areas. Further local scopingand consultation would be required to determine the nature ofintervention (Bryant, 2002), but consistentwith relatedphysical activitystudies, the ‘walkability’ of local services and amenities appears to be apromising area to explore through such natural experiments.

Study strengths and limitations

The studyhad strength in rigorous recruitment anddata collection toobtain a representative random sample. Complete data on a range ofphysical and social environmental indicators collected through estab-lished and validated questionnaires have enabled a unique explorationof the differing relationships with physical and mental health.

A number of study limitations are recognised. A larger sample wouldhave enabledmore powerful analysis of relationships. All datawere self-report, albeit using the CAPI approach. The focus on perceivedenvironmental variables was deemed conceptually relevant to self-reported health and in keeping with evidence that negative neighbour-hood perceptions are perhaps more important to health than theobjective ‘reality’ (Wen et al., 2006). A logical next step would be tocombine these with objective environmental measures to explore theirrelative importance for self-reported health. Participants were residentacross 10 areas, all of which were urban and could, therefore, havelimited the possible variability in some environmental measures. Thesampling frame used here was intended to obtain areal representationfor inner city areas (i.e., definably urban, not rural or fringe), with arepresentative sample of individuals within each.Wewere successful inthis regard but recognise that greater environmental heterogeneitycould be achieved through sampling areas based on environmentalattributes objectively determined using GIS (e.g., residential density ordensity of local amenities). Finally the cross-sectional data presentedpreclude inferences about causality.

Conclusions

Our findings confirm that socio-economic factors alone cannotexplain the associations between neighbourhood perceptions andhealth. At the individual-level, personal health was associated with anumber of other physical and social neighbourhood characteristics.Living in neighbourhoods with ‘walkable’ local destinations appearsparticularly important, especially for physical health, whereas socialsupport is more pertinent to mental health. Differences in the specificenvironment factors that might promote physical and mental healthhighlight the potential for further exploration in larger samples, moreenvironmentally diverse areas, and through robustly evaluatednaturalexperiments. Our findings make a novel contribution to an expandingbody of research that makes the case for improving neighbourhoodenvironments and associatedperceptions to benefit populationhealth.

Conflict of interest statementThe authors declare there are no conflicts of interest.

Acknowledgments

This research was funded under the National Prevention ResearchInitiative, managed by the Medical Research Council, UK (MRC Grant

ID 75376). The opinions expressed are those of the authors and notnecessarily those of the funding bodies.

References

Acheson, D., 1998. Independent inquiry into inequalities in health. The StationeryOffice, London.

Almedom, A.M., 2005. Social capital and mental health: an interdisciplinary review ofprimary evidence. Soc. Sci. Med. 61, 943–964.

Bajekal, M., Purdon, S., 2001. Social capital and social exclusion: development of acondensed module for the Health Survey for England. National Centre for SocialResearch, Department of Health, London.

Baum FE, Ziersch AM, Zhang G, Osborne K, 2009. Do perceived neighbourhood cohesionand safety contribute to neighbourhood differences in health? Journal ofEpidemiology and Community Health 15, 925–934.

Berkman, L.F., 1984. Assessing the physical health effects of social networks and socialsupport. Annu. Rev. Publ. Health 5, 413–432.

Browning, C.R., Cagney, K.A., Wen, M., 2003. Explaining variation in health status acrossspace and time: implications for racial and ethnic disparities in self-rated health.Soc. Sci. Med. 57, 1221–1235.

Brunner, E., Marmot, M., 2006. Social organisation, stress and health. In: Marmot, M.,Wilkinson, R.G. (Eds.), Social Determinants of Health, second ed. Oxford UniversityPress, Oxford, pp. 6–30.

Bryant, T., 2002. Role of knowledge in public health and health promotion policychange. Health Promot. Int. 17, 89–98.

Cerin, E., Saelens, B.E., Sallis, J.F., Frank, L.D., 2006. Neighborhood environmentwalkability scale: validity and development of a short form. Med. Sci. Sport Exer.38, 1682–1691.

Cleland, V.J., Timperio, A., Crawford, D., 2008. Are perceptions of the physical and socialenvironment associated with mothers' walking for leisure and for transport? Alongitudinal study. Prev. Med. 47, 188–193.

Cochrane, T., Davey, R., Gidlow, C., Smith, G., Fairburn, J., Armitage, C.J., Stephansen, H.,Speight, S., 2009. Small area and individual level predictors of physical activity inurban communities: a multi-level study in Stoke on Trent, England. Int. J. Environ.Res. Public Health 6, 654–677.

Communities and Local Government, 2004. The English Indices of Deprivation 2004(revised), available at http://www.communities.gov.uk/documents/communi-ties/pdf/131209.

Communities and Local Government, 2007. The English Indices of Deprivation 2007.Department for Communities and Local Government, London.

Cooper, H., Arber, S., Fee, L., Ginn, J., 1999. Influence of social support and social capitalon health. Health Education Authority, London.

Cummins, S., Curtis, S., Diez-Roux, A.V., Macintyre, S., 2007. Understanding andrepresenting `place' in health research: a relational approach. Soc. Sci. Med. 65,1825–1838.

Cummins, S., Macintyre, S., Davidson, S., Ellaway, A., 2005. Measuring neighbourhoodsocial and material context: generation and interpretation of ecological data fromroutine and non-routine sources. Health Place 11, 249–260.

Davey, R.C., Cochrane, T., Gidlow, C., Fairburn, J., Smith, G., 2008. Design of a pragmaticcluster randomised controlled trial: ecological approach to increasing physicalactivity in an urban community. Contemp. Clin. Trial 29, 774–782.

de Hollander, A.E.M., Staatsen, B.A.M., 2003. Health, environment and quality of life: anepidemiological perspective onurbandevelopment. LandscapeUrban Plan. 65, 53–62.

De Silva, M.J., McKenzie, K., Harpham, T., Huttly, S.R.A., 2005. Social capital and mentalillness: a systematic review. J. Epidemiol. Commun. Health 59, 619–627.

Department of Health, 2008. Health Inequalities: Progress and Next Steps. DH, London.Ellaway, A., Macintyre, S., Kearns, A., 2001. Perceptions of place and health in socially

contrasting neighbourhoods. Urban Stud. 38, 2299–2316.Ewing, R., Schmid, T., Killingsworth, R., Zlot, A., Raudenbush, S., 2003. Relationship

between urban sprawl and physical activity, obesity, and morbidity. Am. J. HealthProm. 18, 47–57.

Fujiwara, T., Kawachi, I., 2008. Social capital and health: a study of adult twins in the U.S.Am. J. Prev. Med. 35, 139–144.

Jenkinson, C., Chandola, T., Coulter, A., Bruster, S., 2001. An assessment of the constructvalidity of the SF-12 summary scores across ethnic groups. J. Public Health 23,187–194.

Jones, A., Bentham, G., Foster, C., Hillsdon, M., Panter, J., 2007. Obesogenic EnvironmentsEvidence Review. Foresight Tackling Obesities: Future Choices Project. Office ofScience and Innovation.

Kessler, R.C., Price, R.H., Wortman, C.B., 1985. Social factors in psychopathology: stress,social support, and coping processes. Annu. Rev. Psych. 36, 531–572.

Kruger, D.J., Reischl, T.M., Gee, G.C., 2007. Neighborhood social conditions mediate theassociation between physical deterioration and mental health. Am. J. Commun.Psych. 40, 261–271.

Latkin, C.A., German, D., Hua, W., Curry, A.D., 2009. Individual-level influences onperceptions of neighborhood disorder: a multilevel analysis. J. Commun. Psych. 37,122–133.

Leslie, E., Cerin, E., 2008. Are perceptions of the local environment related toneighbourhood satisfaction and mental health in adults? Prev. Med. 47, 273–278.

Maas, J., Verheij, R.A., Groenewegen, P.P., de Vries, S., Spreeuwenberg, P., 2006. Greenspace, urbanity, and health: how strong is the relation? J. Epidemiol. Commun.Health 60, 587–592.

Mitchell, R., Popham, F., 2007. Greenspace, urbanity and health: relationships inEngland. J. Epidemiol. Commun. Health 61, 681–683.

163C. Gidlow et al. / Preventive Medicine 51 (2010) 157–163

National Health Service, 2008. Director of Public Health Annual Report 07/08. NHS,Stoke-on-Trent.

National Institute for Health and Clinical Excellence, 2008. Promoting and creating builtor natural environments that encourage and support physical activity. : NICE publichealth guidance, 8. NICE, London.

O'Campo, P., Salmon, C., Burke, J., 2009. Neighbourhoods and mental well-being: whatare the pathways? Health Place 15, 56–68.

Office for National Statistics, 2001. Census 2001 forms. ONS, London. Available athttp://www.statistics.gov.uk/census2001/census_form.asp.

Office of National Statistics, 2000. National Statistics Socio-economic Classification (NS-SEC). ONS, London. Available at http://www.statistics.gov.uk/methods_quality/ns_sec/default.asp.

Owen, N., Cerin, E., Leslie, E., duToit, L., Coffee, N., Frank, L.D., Bauman, A.E., Hugo, G.,Saelens, B.E., Sallis, J.F., 2007. Neighborhood walkability and the walking behaviorof australian adults. Am. J. Prev. Med. 33, 387–395.

Petticrew, M., Cummins, S., Ferrell, C., Findlay, A., Higgins, C., Hoy, C., Kearns, A., Sparks,L., 2005. Natural experiments: an underused tool for public health? Public Health119, 751–757.

Pickett, K.E., Pearl, M., 2001. Multilevel analyses of neighbourhood socioeconomic contextand health outcomes: a critical review. J. Epidemiol. Commun. Health 55, 111–122.

Poortinga, W., Dunstan, F., Fone, D., 2007. Perceptions of the neighbourhood environ-ment and self rated health: a multilevel analysis of the Caerphilly Health and SocialNeeds Study. BMC Pub. Health 7, 285.

Quality Metric, 2006. The SF-12®: An Even Shorter Health Survey, Available at: http://www.sf-36.org/tools/sf12.shtml.

Riva, M., Gauvin, L., Barnett, T.A., 2007. Toward the next generation of research intosmall area effects on health: a synthesis of multilevel investigations published sinceJuly 1998. J. Epidemiol. Commun. Health 61, 853–861.

Sallis JF, Bauman A, Pratt M, 1998. Environmental and policy interventions to promotephysical activity. American Journal of Preventive Medicine 15, 379–397.

Speight, S., Stephansen, H., Clemens, S., 2007. Socio-ecological mapping of physicalactivity behaviours and health outcomes in deprived inner-city communities.Technical Report. National Centre for Social Research, London.

Sport England, 2007. The Active People Survey. Available at http://www.sportengland.org/research/active_people_survey.aspx.

Stansfield, S.A., 2006. Social support and social cohesion, In: Marmot, M., Wilkinson, R.G.(Eds.), Social determinants of health:, second edition. OxfordUniversity Press, Oxford,pp. 148–171.

Stokols, D., 1992. Establishing and maintaining health environments: toward a socialecology of health promotion. Am. Psychol. 47, 6–22.

Sundquist, K., Yang, M., 2007. Linking social capital and self-rated health: a multilevelanalysis of 11, 175 men and women in Sweden. Health Place 13, 324–334.

Takano, T., Nakamura, K., Watanabe, M., 2002. Urban residential environment andsenior citizens' longevity in megacity areas: the importance of walkable greenspace. J. Epidemiol. Commun. Health 56, 913–918.

Veenstra, G., Luginaah, I., Wakefield, S., Birch, S., Eyles, J., Elliott, S., 2005. Who you know,where you live: social capital, neighbourhood and health. Soc. Sci. Med. 60, 2799–2818.

Wen, M., Hawkley, L.C., Cacioppo, J.T., 2006. Objective and perceived neighborhoodenvironment, individual SES and psychosocial factors, and self-rated health: Ananalysis of older adults in Cook County, Illinois. Soc. Sci. Med. 63, 2575–2590.