Open Access

Does exposure to the food environment differ by socioeconomic position? Comparing area-based and person-centred metrics in the Fenland Study, UK

International Journal of Health Geographics201716:33

https://doi.org/10.1186/s12942-017-0106-8

Received: 13 May 2017

Accepted: 23 August 2017

Published: 6 September 2017

Abstract

Background

Retail food environments (foodscapes) are a recognised determinant of eating behaviours and may contribute to inequalities in diet. However, findings from studies measuring socioeconomic inequality in the foodscape have been mixed, which may be due to methodological differences. The aim of this cross-sectional study was to compare exposure to the foodscape by socioeconomic position using different measures, to test whether the presence, direction or amplitude of differences was sensitive to the choice of foodscape metric or socioeconomic indicator.

Methods

A sample of 10,429 adults aged 30–64 years with valid home address data were obtained from the Fenland Study, UK. Of this sample, 7270 participants also had valid work location data. The sample was linked to data on food outlets obtained from local government records. Foodscape metrics included count, density and proximity of takeaway outlets and supermarkets, and the percentage of takeaway outlets relative to all food outlets. Exposure metrics were area-based (lower super output areas), and person-centred (proximity to nearest; Euclidean and Network buffers at 800 m, 1 km, and 1 mile). Person-centred buffers were constructed using home and work locations. Socioeconomic status was measured at the area-level (2010 Index of Multiple Deprivation) and the individual-level (highest educational attainment; equivalised household income). Participants were classified into socioeconomic groups and average exposures estimated. Results were analysed using the statistical and percent differences between the highest and lowest socioeconomic groups.

Results

In area-based measures, the most deprived areas contained higher takeaway outlet densities (p < 0.001). However, in person-centred metrics lower socioeconomic status was associated with lower exposure to takeaway outlets and supermarkets (all home-based exposures p < 0.001) and socioeconomic differences were greatest at the smallest buffer sizes. Socioeconomic differences in exposure was similar for home and combined home and work measures. Measuring takeaway exposure as a percentage of all outlets reversed the socioeconomic differences; the lowest socioeconomic groups had a higher percentage of takeaway outlets compared to the middle and highest groups (p < 0.001).

Conclusions

We compared approaches to measuring socioeconomic variation in the foodscape and found that the association was sensitive to the metric used. In particular, the direction of association varied between area- and person-centred measures and between absolute and relative outlet measures. Studies need to consider the most appropriate measure for the research question, and may need to consider multiple measures as a single measure may be context dependent.

Keywords

Diet inequalities Food environment Fast food Supermarkets Socioeconomic status Area deprivation Geographic information systems Density Proximity Food access

Background

Socioeconomic disparities in diet quality have been identified [13] and likely contribute to inequalities in health outcomes including obesity, type 2 diabetes and cardiovascular disease [47]. Pervasive inequalities in diet have led researchers to seek modifiable determinants as potential targets of public health policies and interventions. The built environment, including the local retail food environment, or ‘foodscape’, may influence dietary behaviours and as a consequence contribute to rising levels of obesity, through shaping the context in which people make their food decisions [811]. The foodscape may be important for explaining socioeconomic disparities in diet, with some evidence of foodscapes varying by socioeconomic status (SES) [11, 12].

Much of the evidence on disparities in food retail access has centred on supermarkets and takeaway food (‘fast-food’) outlets. Specifically, supermarkets are seen to offer a range of affordable fresh produce for preparation at home [13, 14]. The ready-prepared hot foods typically served in takeaway outlets have been positively associated with energy and fat intakes [15], with frequent consumption associated with weight gain over time [16, 17]. In the United States (US) systematic reviews have identified consistent evidence of limited supermarket access and high takeaway food outlet access in low-income and racially-segregated neighbourhoods [1820]. However, outside of the US, the picture has been more mixed. In the United Kingdom (UK), Canada and Australia, a limited number of studies have found more takeaway outlets in deprived areas [2125], with one exception from Glasgow, Scotland [26]. No UK studies have identified socioeconomic disparities in the geographic accessibility of supermarkets [23, 27, 28]. In New Zealand, findings have shown that all food outlets are more prevalent in deprived areas [29]. These studies tend to use area-based measures of SES, limiting further association of these exposures to individual-level dietary behaviours and health outcomes. Socioeconomic differences in foodscape exposure based on individual level measures are therefore important to understand.

Furthermore, a number of studies have recognised the importance of considering a wider range of food outlets when assessing potential neighbourhood impacts on diet [3034]. It may be that outlets offering both ‘healthy’ and ‘unhealthy’ options are commonly co-located, given food outlets tend to cluster in commercial areas [30]. Few previous studies have examined the socioeconomic differences in foodscape exposure using a relative measure of the retail food landscape [35, 36]. A recent Canadian study reported how more deprived neighbourhoods in Waterloo were more likely to have a less healthy food retail mix [36], however these findings were in contrast to previous Canadian research [35], may not reflect the wider international context, and were not based on person-centred estimates of either food access or indicators of SES.

The majority of foodscape studies use Geographic Information Systems (GIS), which is now a common tool in public health research [8]. However, a number of review articles have noted that methodological differences limit comparisons between food environment studies [8, 9, 12]. As a result, such reviews have called for greater consistency across methods used, as well as consideration of specific aspects of the methods, when assessing associations between measures of the foodscape and both SES and other outcomes including diet and obesity.

The purpose of this study was to compare socioeconomic differences in foodscape exposure using a number of commonly-used GIS-based metrics to better understand the implications of selecting different metrics. Using a cohort sample of UK adults, we compared area-based (‘ecological’) and person-centred (‘egocentric’) methods. Moreover, as much of the person-centred methods published in the literature is based on exposures around the residence only, we analysed person-centred metrics derived from both residential exposures and combined home- and work-based exposures. We also examined both absolute measures of takeaway outlet and supermarket exposure and relative measures of takeaway outlet exposure across three socioeconomic indicators.

Methods

Study sample

The Fenland Study is a population-based, observational cross-sectional cohort study, which aims to understand the genetic, behavioural and environmental factors relating to obesity and diabetes in adults. Eligible participants were men and women born between the years 1950 and 1975, aged 30–62 years at recruitment, who were registered at participating general practices across Fenland and East Cambridgeshire. Exclusion criteria were diagnosed diabetes, psychotic or terminal illness, pregnancy and being unable to walk unaided. Recruitment ran from 2005 to 2014 [37].

At the time of requesting data for this study, data were available for 11,857 Fenland Study participants. Participant home and work postcodes were geocoded using GeoConvert [38], and mapped using a GIS software package (ArcMAP 10, ESRI). UK postcodes contain 15 addresses on average, and therefore represent addresses with relative precision [39]. Participants whose home postcodes were missing (n = 1276), incomplete or invalid (n = 121), or outside the study area (n = 31) were removed, leaving an analytic sample of 10,429 with complete home address data. For work-based exposure assessments, we further excluded those not in work (n = 963) or where work status data was missing (n = 94), as well as those missing work postcodes, who provided incomplete work postcodes, or whose work address was outside the study area (n = 2102), resulting in a sub-sample of 7270.

Food outlet data

Food outlet data collection and classification for use alongside Fenland Study data has been previously reported [40]. Briefly, food outlet data were collected in November and December 2011 from ten local councils covering the study area. All food outlet owners in the UK are required to register their premises with local Environmental Health departments, and to notify the department if they are closing. Local councils therefore hold records of food outlets that are regularly updated [41]. These are considered the most accurate source of food outlet data in the UK [42]. For this study we focussed on the locations of takeaways, supermarkets, convenience stores, restaurants, and cafés, which together account for the majority of household food shopping and out-of-home eating [40].

Foodscape exposure metrics

Two broad and commonly employed approaches to assessing the foodscape were compared [4345]: area-based metrics and person-centred (individual-level) metrics. Area-based metrics define neighbourhoods using administrative boundaries, which in this study were lower super output areas (LSOAs) designed to contain an average of 1500 residents and 650 households [46]. There were 801 LSOAs within the Fenland Study area. When using area-based metrics, outlet counts are generally standardised by resident population to give a density measure of exposure. In this study, takeaway and supermarket outlet counts were standardised against mid-2011 LSOA population estimates [47].

Person-centred metrics capture a neighbourhood specific to the individual, typically centred on a study participants’ home location. In addition to home, we considered the workplace an important daily anchor point and therefore also assessed the inclusion of workplace exposure to capture wider exposure to the foodscape [48]. We counted numbers of food outlets within Euclidean (straight-line) buffers and street network buffers, which measure distance along the road network and thus account for land use, for the reported home and work postcodes of Fenland Study participants. Buffers were constructed at three distances: 800, 1000, and 1609 m (1 mile), based on common precedent for their use in the published literature, wherein they have been theorised as ‘walkable’ for an average adult in 10–15 min [4952]. In addition, person-centred exposure was also measured as proximity to the nearest takeaway outlet and supermarket, calculated as the shortest street network distance using ArcGIS Network Analyst.

Socioeconomic indicators

Area-level SES was defined using the 2010 Index of Multiple Deprivation (IMD) [53]. IMD is a composite measure of deprivation calculated from indicators in seven domains: income; employment; health and disability; education skills and training; barriers to housing and other services; crime; living environment. IMD assigns relative scores to LSOAs, with higher scores reflecting greater deprivation. Fenland Study LSOAs ranged in score from 1.01 (least deprived) to 61.39 (most deprived), spanning the first and ninety-eighth centiles of all scores across England. LSOAs were divided into tertiles of deprivation based on their IMD score.

Two common indicators of individual-level SES were used: highest educational attainment and equivalised income. These data were provided by Fenland Study participants in a self-completed general questionnaire. Participants’ highest educational attainment was collapsed into three categories (‘Low’, includes compulsory school education and equivalent qualifications, typically completed at 16 years of age; ‘Medium’, includes academic or vocational qualifications gained during further education, such as those that allow university entry; ‘High’, degree-level or equivalent qualifications). To adjust for household size, annual household income was equivalised using the OECD’s modified equivalence scale [54]. Equivalised income was calculated for each participant from the midpoint of the available household income categories and collapsed into three bands (<£23,000; £23,000–£42,999; ≥£43,000).

As this was a complete case analysis, missing socioeconomic data reduced sample sizes in person-centred models that included educational attainment and equivalised income to 10,276 and 9617 respectively for home-based assessments, and to 7169 and 6774 respectively for combined home and work assessments.

Data analysis

Average counts of, and distances to the nearest of each supermarket and takeaway were calculated across socioeconomic groups (highest educational attainment and household income groups for person-centred exposure metrics; tertiles of IMD scores for area-based metrics). Average proximity to the nearest supermarket and takeaway outlet from home was calculated for each socioeconomic group. Relative differences in exposures between the highest and lowest socioeconomic groups were also calculated. Takeaway density as a percentage of the sum of all food outlet types (takeaways, supermarkets, convenience stores, restaurants, cafés) was calculated. This calculation resulted in missing values where there were both no takeaway outlets and no other outlets, that is, both the numerator and denominator were zero. This was a complete case analysis, such that missing values resulted in list wise exclusion from respective analyses. For example, where person-based residential neighbourhood food environment exposures were derived in relation to participant educational attainment, we included all participants with data for education, home address, and associated covariates. Participants included in these analyses could have been lacking income data and work addresses. The number of missing values also varied according to the size of spatial buffer used, as detailed in Appendix Table 6. Linear tests for trends were used to test differences in means across socioeconomic groups, with p < 0.05 considered statistically significant. All analyses were conducted in Stata 13 (StataCorp., 2013).

Sensitivity analyses

As every LSOA for the total Fenland Study geographic area was included in main area-based analyses (n = 801 LSOAs)—including both LSOAs where Fenland participants lived and those where they did not—there was a potential for bias. Thus, we conducted a sensitivity analysis of only the subsample of LSOAs where Fenland Study participants lived (n = 285 LSOAs).

Results

Sample characteristics

Characteristics of the analytic sample are provided in Table 1. Of the home-based sample, 89.9% reported being in work. The home and the combined home and work samples were similar across key age, sex, education and income profiles (Table 1).
Table 1

Descriptive statistics for the Fenland Study analytic sample (n = 10,429)

Variable

n

%

Missing n (%)

Female

5392

51.7

0

Age (years)

10,429

100

0

 29–39

1652

15.8

 

 40–49

4338

41.6

 

 50–64

4439

42.6

 

Educational attainmenta

10,276

98.5

153 (1.5)

 Low

2091

20.1

 

 Medium

4698

45.1

 

 High

3487

33.4

 

Equivalised incomeb

9617

92.2

812 (7.8)

 <£23,000

2372

22.7

 

 £23,000–£42,999

3991

38.3

 

 ≥£43,000

3254

31.2

 

Work address reported

7927

76.0

2502 (24.0)

Work status

10,335

99.1

94 (0.9)

 In workc

9372

89.9

 

 Not in workd

963

9.2

 

aEducational attainment: ‘Low’ indicates compulsory school education and corresponding qualifications; ‘Medium’ indicates further education academic or vocational qualifications; ‘High’ represents degree or higher qualifications

bTotal household income equivalised using a version of the OECD’s modified equivalence scale

cListed as working ‘full time’, ‘part time’, ‘obtained new job’

dListed as ‘retired’, ‘keeping house’, ‘unemployed’, ‘sick leave’

Area-based exposures

The population-adjusted density of takeaway outlets was greatest in the most deprived LSOAs (Table 2), with 76% fewer outlets in the least deprived tertile of areas compared to the most, and with a significant gradient across the deprivation tertiles (p < 0.001). There was no significant difference in area-based supermarket density by deprivation (p = 0.126). When measured as a percentage of all outlets, takeaway outlets made up 21.5% in the most deprived tertile compared to 12.4% in the least deprived tertile.
Table 2

Area-based takeaway outlet and supermarket densities (mean, 95% CI) per 10,000 population

LSOAsa

Takeaway outlets (n)

Supermarkets (n)

Takeaway outlets (% all outlets)d

Most deprived

9.4 (7.4, 11.5)

1.4 (1.0, 1.8)

21.5 (18.6, 24.4)

Medium

4.9 (3.6, 6.1)

1.0 (0.7, 1.3)

15.0 (12.1, 17.9)

Least deprived

2.3 (1.7, 2.8)

1.0 (0.7, 1.3)

12.4 (9.7, 15.2)

p-Trendb

<0.001

0.126

<0.001

% Differencec

-76

-29

-42

aLower super output areas (n = 801) in tertiles of deprivation, defined using 2010 Index of Multiple Deprivation

bTest for trend based on linear regression with tertiles of deprivation treated as a continuous variable

cPercent difference computed as 100*((least deprived-most deprived)/most deprived)

dAll food outlets = takeaways, supermarkets, convenience stores, restaurants, cafes

Person-centred exposures

Comparing geographic boundaries for home-based exposures

Around the home, mean takeaway exposure was significantly and positively associated with both equivalised income and level of education at each scale of Euclidean and Network buffers (Table 3). Of the two socioeconomic indicators, educational attainment revealed larger differences in takeaway exposure than equivalised income. For example, within 1 mile Euclidean buffers, the percentage increase in takeaway outlets between the lowest and highest socioeconomic group was 56% for level of education and 17% for equivalised income. While absolute counts of outlets were greater in the larger buffers, relative socioeconomic differentials were greater at the smallest buffer sizes. Within Network buffers across level of education, for example, the difference in takeaway outlets at 800 m was 69% and at 1 mile was 49%. Outlet counts were greater in Euclidean than Network buffers. Proximity to the nearest takeaway was 7% shorter for the most educated participants than the least educated, which was significantly different (p = 0.012). There was no significant difference between equivalised income groups in proximity to the nearest takeaway outlet.
Table 3

Home-based takeaway exposure by groups of educational attainment and equivalised income

 

Euclidean buffers (no. of outlets)

Network buffers (no. of outlets)

Proximity to nearest (km)

800 m

1 km

1 mile

800 m

1 km

1 mile

Mean

Educational attainment a (n = 10,276)

Low

2.6 (2.4, 2.8)

3.8 (3.6, 4.1)

7.8 (7.4, 8.2)

1.3 (1.2, 1.4)

1.9 (1.8, 2.13)

4.7 (4.5, 5.0)

1.55 (1.50, 1.60)

Medium

2.5 (2.4, 2.6)

3.6 (3.4, 3.7)

7.0 (6.7, 7.3)

1.24 (1.2, 1.3)

1.9 (1.8, 2.0)

4.3 (4.1, 4.5)

1.64 (1.59, 1.69)

High

4.3 (4.1, 4.5)

6.1 (5.9, 6.4)

12.2 (11.7, 12.6)

2.2 (2.0, 2.3)

3.2 (3.1, 3.4)

7.0 (6.7, 7.3)

1.45 (1.39, 1.51)

p (Highest to lowest)

<0.001

<0.001

<0.001

<0.001

<0.001

<0.001

0.012

% Differencec

65

61

56

69

68

49

-7

Equivalised income b (n = 9617)

<£23,000

2.9 (2.7, 3.1)

4.2 (4.0, 4.5)

8.6 (8.2, 9.0)

1.5 (1.4, 1.6)

2.3 (2.1, 2.4)

5.1 (4.8, 5.4)

1.55 (1.48, 1.62)

£23,000–£42,999

3.0 (2.8, 3.1)

4.3 (4.1, 4.5)

8.3 (7.9, 8.6)

1.5 (1.4, 1.6)

2.3 (2.1, 2.4)

5.1 (4.8, 5.3)

1.58 (1.52, 1.63)

≥£43,000

3.6 (3.4, 3.8)

5.1 (4.8, 5.3)

10.1 (9.7, 10.6)

1.8 (1.6, 1.9)

2.6 (2.5, 2.8)

5.9 (5.6, 6.1)

1.52 (1.46, 1.58)

p (Highest to lowest)

<0.001

<0.001

<0.001

<0.001

<0.001

<0.001

0.915

% Differencec

24

21

17

20

13

16

-2

Mean values (95% confidence intervals), and statistical and percent difference between the lowest and highest socioeconomic groups

aEducational attainment: ‘Low’ indicates compulsory school education and corresponding qualifications; ‘Medium’ indicates further education academic or vocational qualifications; ‘High’ represents degree or higher qualifications

bTotal household income equivalised using a version of the OECD’s modified equivalence scale

cPercent difference computed as 100*((Higher SES-Lower SES)/lower SES)

Results for supermarkets were similar (Table 4). Supermarket exposures were significantly and positively associated with both socioeconomic indicators, with educational attainment showing larger differentials than equivalised income. Within 1 mile Euclidean buffers, for example, between the lowest and highest socioeconomic group, the difference was 136% for level of education and 37% for equivalised income. The absolute outlet counts were greater in larger Euclidean and Network buffer sizes, and the relative differences were greater for smaller buffer sizes. Distances to the nearest supermarket were significantly shorter for highly-educated participants (23% shorter) compared to those least educated, and for higher income participants (9% shorter) compared to those in the lowest income group.
Table 4

Home-based supermarket exposure by groups of educational attainment and equivalised income

 

Euclidean buffer counts (no. of outlets)

Network buffer counts (no. of outlets)

Proximity to nearest (km)

800 m

1 km

1 mile

800 m

1 km

1 mile

Mean

Educational attainment a (n = 10,276)

Low

0.4 (0.4, 0.5)

0.7 (0.6, 0.7)

1.4 (1.4, 1.5)

0.2 (0.2, 0.2)

0.3 (0.2, 0.4)

1.7 (1.6, 1.8)

3.73 (3.59, 3.88)

Medium

0.5 (0.5, 0.5)

0.7 (0.7, 0.7)

1.5 (1.4, 1.5)

0.2 (0.2, 0.2)

0.3 (0.3, 0.4)

1.7 (1.6, 1.8)

3.90 (3.80, 3.99)

High

1.1 (1.0, 1.1)

1.6 (1.5, 1.6)

3.3 (3.2, 3.4)

0.5 (0.4, 0.5)

0.7 (0.7, 0.8)

3.8 (3.7, 4.0)

2.87 (2.76, 2.98)

p (Highest to lowest)

<0.001

<0.001

<0.001

<0.001

<0.001

<0.001

<0.001

% Differencec

1175

129

136

150

133

124

-23

Equivalised income b (n = 9617)

<£23,000

0.6 (0.5, 0.6)

0.8 (0.8, 0.9)

1.9 (1.8, 2.0)

0.3 (0.2, 0.3)

0.4 (0.4, 0.4)

2.2 (2.0, 2.3)

3.67 (3.54, 3.81)

£23,000–£42,999

0.6 (0.6, 0.7)

0.9 (0.9, 1.0)

2.0 (1.9, 2.1)

0.3 (0.3, 0.3)

0.5 (0.5, 0.5)

2.3 (2.2, 2.4)

3.60 (3.49, 3.71)

≥£43,000

0.8 (0.8, 0.9)

1.2 (1.1, 1.3)

2.6 (2.4, 2.7)

0.4 (0.3, 0.4)

0.6 (0.5, 0.6)

2.9 (2.8, 3.1)

3.34 (3.22, 3.46)

p (Highest to lowest)

<0.001

<0.001

<0.001

<0.001

<0.001

<0.001

<0.001

% Differencec

33

50

37

33

50

32

-9

Mean values (95% confidence intervals), and statistical and percent difference between the lowest and highest socioeconomic groups

aEducational attainment: ‘Low’ indicates compulsory school education and corresponding qualifications; ‘Medium’ indicates further education academic or vocational qualifications; ‘High’ represents degree or higher qualifications

bTotal household income equivalised using a version of the OECD’s modified equivalence scale

cPercent difference computed as 100*((Higher SES-Lower SES)/lower SES)

Geographic boundaries and combined home and work exposures

With combined home and work exposures, there was a consistent association between level of education and takeaway exposure, which was positive and significant for all Euclidean buffers and for 1 mile Network buffers (Table 5). For example, within the 1 mile Network buffer highest educated participants had 16% greater takeaway exposure than lowest educated participants (p < 0.001). The association between equivalised income and takeaway exposure was positive and significant within 1 mile Euclidean and Network buffers, with a 11% and 7% difference between the highest and lowest income groups, respectively.
Table 5

Combined home and work exposure to takeaway outlets and supermarkets, by groups of educational attainment and equivalised income

 

Euclidean buffer counts (no. of outlets)

Network buffer counts (no. of outlets)

800 m

1 km

1 mile

800 m

1 km

1 mile

Takeaway outlets

Educational attainmenta (n = 7169)

 Low

7.86 (7.39, 8.32)

10.73 (10.16, 11.29)

20.22 (19.29, 21.14)

4.55 (4.23, 4.87)

6.29 (5.89, 6.69)

12.74 (12.09, 13.38)

 Medium

7.01 (6.72, 7.29)

9.53 (9.18, 9.89)

18.05 (17.46, 18.64)

4.00 (3.80, 4.20)

5.54 (5.29, 5.79)

11.34 (10.93, 11.76)

 High

8.49 (8.15, 8.83)

12.35 (11.90, 12.81)

25.14 (24.32, 25.95)

4.48 (4.27, 4.70)

6.62 (6.32, 6.91)

14.82 (14.27, 15.37)

 p (Highest to lowest)

0.026

<0.001

<0.001

0.725

0.191

<0.001

 % Differencec

8

15

24

-2

5

16

Equivalised incomeb (n = 6774)

 <£23,000

7.73 (7.28, 8.17)

10.59 (10.04, 11.14)

20.31 (19.38, 21.24)

4.47 (4.17, 4.77)

6.23 (5.84, 6.61)

12.62 (11.98, 13.26)

 £23,000−£42,999

7.51 (7.20, 7.82)

10.47 (10.06, 10.88)

20.18 (19.48, 20.87)

4.16 (3.95, 4.37)

5.89 (5.62, 6.16)

12.50 (12.02, 12.99)

 ≥£43,000

7.91 (7.57, 8.25)

11.28 (10.83, 11.72)

22.49 (21.71, 23.27)

4.25 (4.03, 4.47)

6.13 (5.84, 6.42)

13.46 (12.93, 13.98)

 p (Highest to lowest)

0.510

0.057

<0.001

0.236

0.686

0.048

 % Differencec

2

7

11

-5

-2

7

Supermarkets

Educational attainmenta (n = 7169)

 Low

1.38 (1.30, 1.46)

1.94 (1.84, 2.05)

3.74 (3.54, 3.94)

0.75 (0.70, 0.81)

1.05 (0.98, 1.12)

3.22 (3.04, 3.39)

 Medium

1.32 (1.27, 1.37)

1.86 (1.79, 1.94)

3.65 (3.51, 3.79)

0.71 (0.68, 0.75)

1.00 (0.96, 1.05)

3.10 (2.98, 3.23)

 High

2.20 (2.11, 2.28)

3.23 (3.12, 3.35)

6.66 (6.42, 6.90)

1.07 (1.02, 1.12)

1.60 (1.53, 1.66)

5.90 (5.67, 6.12)

 p (Highest to lowest)

<0.001

<0.001

<0.001

<0.001

<0.001

<0.001

 % Differencec

59

67

78

45

52

83

Equivalised incomeb (n = 6774)

 <£23,000

1.43 (1.35, 1.51)

2.08 (1.96, 2.19)

4.15 (3.92, 4.38)

0.77 (0.71, 0.82)

1.08 (1.01, 1.14)

3.63 (3.43, 3.83)

 £23,000−£42,999

1.61 (1.55, 1.68)

2.32 (2.23, 2.42)

4.56 (4.37, 4.74)

0.84 (0.80, 0.89)

1.21 (1.15, 1.26)

3.96 (3.79, 4.13)

 ≥£43,000

1.87 (1.79, 1.94)

2.68 (2.57, 2.79)

5.47 (5.25, 5.69)

0.93 (0.88, 0.98)

1.37 (1.30, 1.43)

4.75 (4.55, 4.95)

 p (Highest to lowest)

<0.001

<0.001

<0.001

<0.001

<0.001

<0.001

 % Differencec

31

29

32

21

27

31

Mean values (95% confidence intervals), trend statistics and percent difference between the lowest and highest socioeconomic groups

aEducational attainment: ‘Low’ indicates compulsory school education and corresponding qualifications; ‘Medium’ indicates further education academic or vocational qualifications; ‘High’ represents degree or higher qualifications

bTotal household income equivalised using a version of the OECD’s modified equivalence scale

cPercent difference computed as 100*((Higher SES-Lower SES)/lower SES)

Absolute takeaway outlet counts for the combined home and work exposures were greater than home-only exposures. For example, within the 800 m Euclidean buffer, combined home and work exposures were two to three times greater than home-based exposures alone. However, the relative differences between socioeconomic groups were smaller than for home-based exposures. In addition, the relative differences were greater in larger than smaller buffer sizes. Highest educated participants had an additional 8% of takeaway outlets within 800 m buffers and 24% within 1 mile Euclidean buffers. In common with home-based exposures, outlet counts were greater in Euclidean than Network buffers.

Results for combined home and work exposures were similar for supermarkets, although unlike takeaway outlets, there were significant and positive associations with both socioeconomic indicators at every Euclidean and Network buffer scale. Educational attainment showed larger differentials than equivalised income, with a relative difference at the 1 mile Euclidean buffer of 78% between lowest and highest level of education and 32% between lowest and highest equivalised income. Similar to takeaway outlets, absolute counts were two to three times greater compared to home-based exposures only. Further, the relative differences in supermarket exposures were greater at larger buffer sizes. For example, compared to lowest educated participants, highest educated participants had 59% greater numbers of supermarkets within 800 m Euclidean buffers and 78% more supermarkets within 1 mile.

Takeaway outlets as a percentage of all food outlets

Takeaway exposure as a percentage of all outlets was negatively and significantly associated with both indicators of individual SES. This association was found for both home and combined home and work exposures. For example, in the 1 mile home-based Euclidean buffer (Fig. 1), takeaway outlets constituted 21.4% of all outlets for the least educated and 16.5% of all outlets for the most educated. There was a smaller but significant difference between low income (20.3%) and high income (18.0%) groups. Results were similar for combined home and work exposure showing that takeaway outlets comprised 21.7% of all outlets for the least educated participants compared to 16.9% for the most educated, and 20.9% for the lowest compared to 18.4% for the highest income participants.
Fig. 1

Takeaway outlets as a percentage of all outlets, 1 mile Euclidean buffers, home and combined home and work exposures. Mean values and error bars showing 95% confidence intervals. ***Linear test for trend across socioeconomic groups significant at the <0.001 level. a Educational attainment (‘education’): low, compulsory schooling; medium, further academic or vocational qualifications; high, degree or higher; Equivalised household income (‘income’): low, <£23,000; medium, £23,000–£42,999; high, ≥£43,000

Similar results were found for all other Euclidean and Network buffers (Appendix Table 6). Average home-based and combined home and work-based exposures by SES to convenience stores, restaurants, cafés and all outlets are also shown in Appendix Table 7.
Table 6

Home and combined home and work exposures (mean, 95% CI) to takeaway outlets as a percentage of all food outlets

 

Home

Home and Work

800 m

1 km

1 mile

800 m

1 km

1 mile

Educational attainment a

Euclidean

N = 9276d

N = 9568d

N = 9929d

N = 6890d

N = 6961d

N = 7067d

 Low

22.2 (21.2, 23.2)

23 (22.1, 23.9)

21.4 (20.8, 22.1)

21.1 (20.4, 21.8)

21.7 (21.1, 22.3)

21.7 (21.2, 22.2)

 Medium

20.0 (19.4, 20.7)

19.9 (19.4, 20.5)

19.7 (19.2, 20.1)

20.2 (19.8, 20.7)

20.5 (20.1, 20.9)

20.7 (20.4, 21)

 High

18.1 (17.5, 18.7)

17.9 (17.4, 18.4)

16.5 (16, 16.9)

17.1 (16.6, 17.5)

17.0 (16.6, 17.3)

16.9 (16.6, 17.2)

 p-trendc

<0.001

<0.001

<0.001

<0.001

<0.001

<0.001

Network

N = 7850d

N = 8717d

N = 9592d

N = 6479d

N = 6769d

N = 7002d

 Low

20.4 (19.1, 21.8)

22.2 (21.3, 23)

21.9 (21.4, 22.5)

20.5 (19.6, 21.4)

20.8 (20.1, 21.6)

21.3 (20.8, 21.8)

 Medium

18.4 (17.5, 19.2)

20.3 (19.7, 20.8)

20.5 (20.1, 20.9)

19.1 (18.5, 19.7)

19.7 (19.2, 20.2)

20.2 (19.9, 20.6)

 High

16.8 (16.1, 17.5)

17 (16.5, 17.5)

16.6 (16.2, 16.9)

15.9 (15.4, 16.5)

16.5 (16, 17)

16.4 (16.1, 16.7)

 p-trendc

<0.001

<0.001

<0.001

<0.001

<0.001

<0.001

Equivalised income b

Euclidean

N = 8,609d

N = 8,856d

N = 9,189d

N = 6,524d

N = 6,588d

N = 6,686d

 <£23,000

21.3 (20.4, 22.2)

21.7 (20.9, 22.5)

20.3 (19.7, 21)

20.3 (19.6, 20.9)

20.8 (20.3, 21.4)

20.9 (20.4, 21.3)

 £23,000–£42,999

19.5 (18.9, 20.2)

19.6 (19, 20.2)

18.7 (18.3, 19.2)

19.5 (19, 20)

19.6 (19.2, 20)

19.7 (19.3, 20)

 ≥£43,000

19.1 (18.4, 19.8)

18.9 (18.3, 19.5)

18.0 (17.5, 18.5)

18.4 (17.9, 18.9)

18.3 (17.9, 18.8)

18.4 (18, 18.7)

 p-trendc

<0.001

<0.001

<0.001

<0.001

<0.001

<0.001

Network

N = 7498d

N = 8281d

N = 9014d

N = 6153d

N = 6424d

N = 6627d

 <£23,000

18.7 (17.5, 19.9)

21.0 (20.2, 21.7)

21 (20.5, 21.5)

19.3 (18.4, 20.1)

19.8 (19.1, 20.6)

20.4 (19.9, 20.9)

 £23,000−£42,999

18.2 (17.4, 19.1)

19.6 (19, 20.1)

19.4 (19, 19.8)

18.5 (17.9, 19.1)

19 (18.4, 19.5)

19.2 (18.8, 19.5)

 ≥£43,000

17.8 (16.9, 18.7)

18.1 (17.6, 18.7)

18.2 (17.8, 18.6)

17.2 (16.5, 17.8)

17.7 (17.1, 18.2)

17.9 (17.5, 18.3)

 p-trendc

0.071

<0.001

<0.001

<0.001

<0.001

<0.001

All food outlets = takeaways, supermarkets, convenience stores, restaurants, cafes

aEducational attainment: ‘Low’ indicates compulsory school education and corresponding qualifications; ‘Medium’ indicates further education academic or vocational qualifications; ‘High’ represents degree or higher qualifications

bTotal household income equivalised using a version of the OECD’s modified equivalence scale

cTest for trend based on linear regression with groups of education/income treated as a continuous variable

dAnalytic sample size restricted to person-centred neighbourhoods containing a minimum of one none takeaway food outlet

Table 7

Home-based exposures (mean, 95% CI) to convenience stores, restaurants, cafés and all outlets, Euclidean and Network 1 mile buffers

Home

Convenience stores

Restaurants

Cafés

All outlets

1 mile euclidean

1 mile network

1 mile euclidean

1 mile network

1 mile euclidean

1 mile network

1 mile euclidean

1 mile network

Educational attainmenta (n = 10,276)

 Low

9.0 (8.5, 9.5)

5.3 (5.0, 5.6)

10.2 (9.5, 11.0)

5.6 (5.2, 6.0)

5.8 (5.3, 6.3)

3.0 (2.6, 3.3)

34.3 (32.3, 36.3)

20.2 (19.0, 21.5)

 Medium

8.0 (7.7, 8.2)

4.8 (4.6, 4.9)

10.7 (10.1, 11.3)

5.9 (5.6, 6.3)

6.4 (6.0, 6.8)

3.3 (3.0, 3.5)

33.5 (32.0, 35.0)

19.9 (19.0, 20.8)

 High

13.3 (12.8, 13.7)

7.5 (7.2, 7.8)

30.1 (28.7, 31.5)

15.7 (14.9, 16.6)

20.0 (19.0, 20.9)

9.8 (9.2, 10.4)

78.8 (75.4, 82.2)

43.8 (41.7, 45.9)

 p-trendc

<0.001

<0.001

<0.001

<0.001

<0.001

<0.001

<0.001

<0.001

 % Differenced

48

42

195

180

245

227

129

117

Equivalised incomeb (n = 9617)

 <£23,000

9.9 (9.5, 10.4)

5.8 (5.5, 6.1)

14.2 (13.1, 15.2)

7.6 (7.0, 8.2)

8.5 (7.8, 9.2)

4.2 (3.8, 4.7)

43.0 (40.4, 45.7)

24.9 (23.2, 26.5)

 £23,000−£42,999

9.4 (9.0, 9.7)

5.6 (5.3, 5.8)

15.7 (14.8, 16.6)

8.5 (8.0, 9.1)

9.6 (9.0, 10.3)

4.8 (4.5, 5.2)

44.9 (42.6, 47.2)

26.3 (24.9, 27.7)

 ≥£43,000

11.2 (10.7, 11.6)

6.3 (6.1, 6.6)

22.0 (20.8, 23.3)

11.4 (10.7, 12.2)

14.3 (13.4, 15.1)

6.9 (6.4, 7.5)

60.2 (57.2, 63.2)

33.5 (31.7, 35.3)

 p-trendc

<0.001

<0.001

<0.001

<0.001

<0.001

<0.001

<0.001

<0.001

 % Differenced

13

9

55

50

68

64

50

35

Home and work e

Educational attainmenta (n = 7169)

 Low

23.0 (21.9, 24.1)

14.2 (13.5, 15.0)

32.6 (30.4, 34.8)

21.0 (19.4, 22.6)

19.3 (17.8, 20.8)

12.7 (11.5, 13.8)

98.8 (93.4, 104.2)

63.8 (59.9, 67.7)

 Medium

20.3 (19.6, 21.0)

12.6 (12.1, 13.1)

33.2 (31.6, 34.8)

21.4 (20.3, 22.5)

20.4 (19.3, 21.4)

13.2 (12.4, 14.0)

95.5 (91.7, 99.3)

61.7 (59.0, 64.4)

 High

27.7 (26.8, 28.6)

16.2 (15.6, 16.7)

69.1 (66.2, 72.0)

42.7 (40.6, 44.7)

45.3 (43.3, 47.2)

27.9 (26.5, 29.3)

173.9 (167.3, 180.5)

107.4 (102.8, 112.1)

 p-trendc

<0.001

<0.001

<0.001

<0.001

<0.001

<0.001

<0.001

<0.001

 % Differenced

20

14

112

103

135

120

76

68

Equivalised incomeb (n = 6774)

 <£23,000

23.0 (21.9, 24.1)

14.2 (13.5, 14.9)

36.6 (34.1, 39.2)

22.9 (21.1, 24.6)

22.3 (20.6, 24.0)

14.0 (12.8, 15.2)

106.4 (100.3, 112.5)

67.3 (63.1, 71.4)

 £23,000−£42,999

22.7 (21.9, 23.5)

13.9 (13.3, 14.4)

43.2 (41.1, 45.4)

27.5 (26.0, 29.1)

26.9 (25.4, 28.3)

17.2 (16.1, 18.3)

117.5 (112.6, 122.5)

75.1 (71.5, 78.6)

 ≥£43,000

25.1 (24.2, 26.0)

14.9 (14.3, 15.5)

55.2 (52.6, 57.9)

34.4 (32.6, 36.2)

35.5 (33.8, 37.3)

22.1 (20.9, 23.4)

143.8 (137.8, 149.9)

89.6 (85.4, 93.8)

 p-trendc

0.001

0.084

<0.001

<0.001

<0.001

<0.001

<0.001

<0.001

 % Differenced

9

5

51

50

59

58

35

33

aEducational attainment: ‘Low’ indicates compulsory school education and corresponding qualifications; ‘Medium’ indicates further education academic or vocational qualifications; ‘High’ represents degree or higher qualifications

bTotal household income equivalised using a version of the OECD’s modified equivalence scale

cTest for trend based on linear regression with groups of education/income treated as a continuous variable

dPercent difference is of the additional exposure in the highest socioeconomic group compared to the lowest

eParticipants included if they had a valid work postcode and were in work

All food outlets = takeaways, supermarkets, convenience stores, restaurants, cafes

Sensitivity analysis

In sensitivity analyses, we compared these results, which were derived from analyses of all LSOAs (n = 801) within the extent of the Fenland Study area, to those obtained from analyses which only included LSOAs (n = 285) containing Fenland Study participants. These alternate results (not shown) were not materially different from those presented here.

Discussion

Summary of findings

We compared the socioeconomic differences in foodscape exposure across a number of GIS-based metrics, and found that the direction and magnitude of socioeconomic differences was metric dependent. Using an area-based metric, takeaway outlets were more concentrated in more deprived areas. This is consistent with the majority of existing area-based work across international contexts [21, 22, 24, 25, 55, 56]. Area-based analysis of supermarkets showed no association between density and area deprivation, consistent with findings from previous UK-based research [23, 27, 28].

The most notable finding in this study is the contrast between the area-based and person-centred measures, when person-centred exposure was characterised as counts of outlets. The area-based results provided evidence that deprived areas have more takeaway outlets, and numbers of supermarkets equivalent to those found in less deprived areas, whereas using person-centred measures, takeaway and supermarket counts were greater for those with higher levels of education and income. There are multiple potential explanations for this discordance. Firstly, an individual’s own SES (based on income or educational attainment) may not be aligned with the SES of their assigned ‘neighbourhood’ (i.e. the deprivation level of their LSOA of residence) [57], or even their specific location within their assigned neighbourhood. There will inevitably exist unmeasured variation in levels of deprivation especially within and across larger administrative areas, such that these estimates cannot be attributed confidently to all residents [58]. Secondly, administrative boundaries such as LSOAs were not designed to capture the extent of their resident’s movements [44, 59], in the way that our person-based metrics of exposure are a better (although imperfect) approximation [60]. For smaller administrative boundaries in particular, individuals’ ‘activity spaces’ (defined as “the sub-set of all locations within which an individual has direct contact as a result of his or her day-to-day activities” [61]) are likely to extend beyond their limits [62], potentially resulting in foodscape exposure that is not reflected in estimates based solely on their assigned residential LSOA (Fig 2). These activity spaces would also likely be heterogeneous in socioeconomic conditions, with the individual exposed to neighborhoods socioeconomically disimilar to their own. This uncertain geographic context problem [63] may partly explain why socioeconomic differentials in absolute person-based food exposure are attenuated when accounting for food outlets around the work place.
Fig. 2

Comparison of lower super output area (area-based) boundaries and 1 mile Euclidean (person-centred) buffer. ©Crown Copyright and Database Right 2015. Ordnance Survey (Digimap Licence)

By contrast, relative takeaway outlet exposure (takeaways as a percentage of all outlets), which better accounts for wider food environment context, showed the same (negative) socioeconomic gradient in both person- and area-level analyses, and this relationship was not attenuated by the addition of work-based exposure estimates. This relative approach has precedent [31, 3336, 6468], and has the potential to minimise residual confounding in analyses relating food outlet exposure to health [55]. In addition, relative access may be a more important conceptualisation of the impact of food outlet exposure on dietary behaviours than absolute numbers [12]. For example, while a high concentration of takeaway outlets offering a range of takeaway options might encourage more frequent consumption, the salience of this exposure may be reduced in the presence of retailers who are offering healthier food options [33]. Prior studies have associated higher relative takeaway outlet densities with low neighbourhood SES [36, 65], higher weight status [33, 34, 64, 67], and lower diet quality [68]. Our findings support the hypothesis that absolute counts of individual outlet types do not provide a complete picture of foodscape exposures.

Our neighbourhood foodscape exposure estimates also showed sensitivity to buffer scale and type specifications, emerging from what has been described in the geographical literature the modifiable areal unit problem’s (MAUP’s) ‘scale’ and ‘zonation’ effects [69]. As illustrated in Fig. 3, irrespective of buffer type larger buffers tend to contain greater numbers of food outlets, resulting in greater estimates of exposure to the foodscape. Similarly, at any given scale, network buffers (that account for land use and may be considered more realistic delineators of neighbourhood than Euclidean buffers [70] generally have a smaller footprint than Euclidean equivalents (Fig. 3), and are therefore likely to contain fewer outlets, resulting in reduced estimates of foodscape exposure. The MAUP is a common consideration in the statistical analysis of geographical data, but has received limited attention in the food environments literature, wherein exposure estimates driven by methodological choices may reveal differential associations with individual-level outcomes of interest. Importantly however, the relative differences between socioeconomic groups in outlet counts remained present across all combinations of buffer scale and type, and were particularly similar across buffers of equal size irrespective of type (i.e., little evidence of MAUP zonation effect was observed when determining differences in outlet densities across socioeconomic groups).
Fig. 3

Comparison of person-centred euclidean and network buffers (800 m, 1 km, 1 mile). ©Crown Copyright and Database Right 2015. Ordnance Survey (Digimap Licence)

There was also some sensitivity in our findings across the two indicators of individual SES tested, with larger differences between the groups highest and lowest educated, than between those on highest and lowest incomes. This suggests that the foodscape is structured according to local residents’ education more so than their income, and may reflect a greater spatial segregation among participants on the basis of their educational background. Alternatively or in addition, it may also reflect the targeting of stores according to neighbourhood sociodemographics.

Finally, the food outlet proximity measures tested also showed evidence of socioeconomic difference. The observed consistency with density measures is consistent with the findings of previous work, which has shown a strong correlation in this regard [71]. However, while proximity measures are useful for assessing geographic access to the closest outlet, they do not address the wider issues of the concentration or mix of outlets in the neighbourhood, which may be important to consider.

Implications for further research

Given that the choice of GIS metric and specifications thereof can influence results, it is important that studies using GIS-based measures of the foodscape are aware of whether and to what extent methodological choices matter, and to conduct sensitivity analyses accordingly. For example, the null associations observed among smaller Network buffers for combined home and work exposure to takeaway outlets suggests that at smaller scales, Type-II errors may be introduced, where an association that exists is not identified, an issue previously identified by Thornton et al. [52]. The choice of metric should be informed by theories of the expected mechanisms by which the foodscape influences diet, and which features of the foodscape are seen as important. For example, takeaway outlets are increasingly seen as particularly relevant within the context of the work ‘neighbourhood’, with behavioural evidence from a US study showing that trips to fast-food outlets occur frequently on workdays [72]. However, given the time constraints of breaks at work, it may be that smaller buffer sizes are more appropriate in this context, in order to capture where workers are able to travel for food purchases, when examining the potential influence of takeaway outlets around the workplace [40]. On the other hand, the appropriate scale for assessing supermarket access may be larger given evidence that shoppers travel longer distances for grocery shopping trips [72], and that they are less likely to use their local supermarket if it is not also economically accessible to them [73]. Proximity measures could inform the size of spatial buffers to capture specific outlet types. Here, the median distance from residents to takeaway outlets was less than 1 km and to supermarkets was up to 2 km. Studies may use such distances to specify exposures by outlet type, as seen in a recent paper examining restaurant and supermarket exposures in relation to dietary behaviours and body mass index [32].

Methodological considerations and limitations

This study is not without limitations. Some degree of temporal mismatch is possible, as the Fenland Study data collection ran from 2005 to 2014 and the food outlet data were collected in 2011. As relative SES is understood as a fairly stable construct [74], the level of error from changes in SES should be minimal. However inaccuracies may still result from any moves in participants’ home or work location. Further, addressing food environment exposure around the workplace does not provide a full assessment of exposure within broader activity spaces. While GIS techniques allow proxy measures, recording the extent of people’s actual activity spaces is possible with global positioning systems (GPS) technologies, an approach that is gradually becoming more common [62, 75, 76]. In addition, participants who did not report being in work were excluded from this wider assessment, and their exposure restricted to the residential neighbourhood. This has potential implications for understanding socioeconomic differences in exposure, as those not in work for certain reasons (e.g. unemployed or on sick leave) may be more likely to be disadvantaged more generally, thus skewing the sample towards higher SES participants. However, there is some evidence that being out-of-work is predictive of having a smaller daily activity space [75], potentially resulting in more accurate approximations of activity space exposure to food outlets when only addressing residential neighbourhood exposure for this group. While 24% of our sample did not have valid work addresses, we found that this percentage did not vary by area deprivation (data not shown), suggesting the lack of this information did not bias the socioeconomic profile of the sample.

The Fenland Study area includes urban and rural areas, and as such is fairly typical of many regions of the United Kingdom. The spread of LSOAs in this sample across the first and ninety-eighth percentile of IMD scores for England, further suggests that our findings are generalisable. With regard to the individual SES measures, the proportions of study participants in the highest education and income groups were also similar to the national average. For example, in 2012/13, 40% of non-retired households in England had an equivalised income upwards of £39,000 [77], while in 2011, 30% of the population held a degree qualification [78]. However, our sample contained a smaller representation of individuals with lower income and less education. For example, in 2011, 13% of residents in the same region had no qualifications [78]. However, the underrepresentation of lower socioeconomic groups among study cohorts is a general concern not unique to our study [79]. Moreover, underrepresentation of low SES groups may only have served to underestimate the scale of inequalities in foodscape exposures between the highest and lowest income groups.

The study also has a number of strengths. Most importantly, we systematically assessed multiple commonly-used foodscape exposure metrics using a large study cohort with detailed sociodemographic information, allowing two indicators of individual-level SES to be used. In addition, the inclusion of workplace location in the cohort allows an important additional anchor point in people’s regular activity space to be included in the measured exposure. In measuring the foodscape, food outlet data were collected from a validated source [41], the accuracy of which is comparable to other secondary food outlet location datasets commonly utilised in food environments research [80].

Conclusions

This study showed associations between socioeconomic status and absolute exposure to takeaway outlets and supermarkets, and relative exposure to takeaway outlets, around home and work locations in a large UK sample of working age adults. The study provides a reference to health researchers interested in measuring the socioeconomic differences in exposure to the food environment and the impact that the choice of metric can have on results. As such, the results presented here can inform further investigations of associations between SES and foodscape exposure , to explore the role that the neighbourhood retail food environment plays in dietary inequalities.

Abbreviations

IMD: 

Index of Multiple Deprivation

GIS: 

geographical information systems

GPS: 

global positioning systems

LSOA: 

lower super output area

MAUP: 

modifiable areal unit problem

OECD: 

Organisation for Economic Co-operation and Development

SES: 

socioeconomic status

UK: 

United Kingdom

US: 

United States

Declarations

Authors’ contributions

ERM, TB and PM conceived of the study. ERM conducted the spatial and statistical analyses and drafted the manuscript. TLP contributed to the statistical analysis. TB collected and categorised the food outlet data, and contributed to the spatial and statistical analysis. PM provided input on the statistical analysis and interpretation of results. NGF is a principal investigator of the Fenland Study. All authors contributed considerably to critical revisions and have read and approved the final manuscript.

Acknowledgements

We thank all the volunteers who have participated in the Fenland Study, the Fenland PIs and the MRC Epidemiology Unit Study Coordination, Field Epidemiology, and Data Management teams. The Fenland study is funded by the MRC (MC_UU-12,015/1).

Competing interests

The authors declare that they have no competing interests and no financial disclosures to make.

Availability of data and materials

The dataset supporting the conclusions of this article is not available; however data requests for the Fenland Study can be submitted to the Fenland Steering Committee for approval. A data request form is available at: http://epi-meta.medschl.cam.ac.uk/includes/fenland/fenland_metadata.html.

Consent for publication

Not applicable.

Ethics approval and consent to participate

All Fenland Study volunteers gave their written informed consent, and the study procedures were approved by the Cambridge Regional Ethics Committee (Ref 04/q0108/19).

Funding

This work was undertaken by the Centre for Diet and Activity Research (CEDAR), a UKCRC Public Health Research Centre of Excellence. Funding from Cancer Research UK, the British Heart Foundation, the Economic and Social Research Council, the Medical Research Council, the National Institute for Health Research [ES/G007462/1], and the Wellcome Trust [087636/Z/08/Z], under the auspices of the UK Clinical Research Collaboration, is gratefully acknowledged. TLP’s Ph.D. studentship is generously supported by the Cambridge International Scholarship, a scheme funded by the Cambridge Commonwealth, European & International Trust. Core Medical Research Council Epidemiology Unit for NGF (MC_UU_12015/5) is also acknowledged. The Fenland Study is funded by the Wellcome Trust and the Medical Research Council.

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Authors’ Affiliations

(1)
UKCRC Centre for Diet and Activity Research (CEDAR), MRC Epidemiology Unit, University of Cambridge School of Clinical Medicine, Institute of Metabolic Science
(2)
MRC Epidemiology Unit, University of Cambridge School of Clinical Medicine, Institute of Metabolic Science
(3)
Department of Nutrition and Exercise Physiology, Washington State University Elson S Floyd College of Medicine

References

  1. James WP, Nelson M, Ralph A, Leather S. Socioeconomic determinants of health. The contribution of nutrition to inequalities in health. BMJ. 1997;314:1545–9.View ArticlePubMedPubMed CentralGoogle Scholar
  2. Dowler E. Inequalities in diet and physical activity in Europe. Public Health Nutr. 2001;4:701–9.View ArticlePubMedGoogle Scholar
  3. Ball K, Mishra GD, Thane CW, Hodge A. How well do Australian women comply with dietary guidelines? Public Health Nutr. 2004;7:443–52.PubMedGoogle Scholar
  4. Giskes K, Avendano M, Brug J, Kunst AE. A systematic review of studies on socioeconomic inequalities in dietary intakes associated with weight gain and overweight/obesity conducted among European adults. Obes Rev. 2010;11:413–29.View ArticlePubMedGoogle Scholar
  5. Drewnowski A. Obesity, diets, and social inequalities. Nutr Rev. 2009;67(Suppl 1):S36–9.View ArticlePubMedGoogle Scholar
  6. Mejean C, Droomers M, van der Schouw YT, Sluijs I, Czernichow S, Grobbee DE, Bueno-de-Mesquita HB, Beulens JW. The contribution of diet and lifestyle to socioeconomic inequalities in cardiovascular morbidity and mortality. Int J Cardiol. 2013;168:5190–5.View ArticlePubMedGoogle Scholar
  7. Stringhini S, Tabak AG, Akbaraly TN, Sabia S, Shipley MJ, Marmot MG, Brunner EJ, Batty GD, Bovet P, Kivimaki M. Contribution of modifiable risk factors to social inequalities in type 2 diabetes: prospective Whitehall II cohort study. BMJ. 2012;345:e5452.View ArticlePubMedPubMed CentralGoogle Scholar
  8. Caspi CE, Sorensen G, Subramanian SV, Kawachi I. The local food environment and diet: a systematic review. Health Place. 2012;18:1172–87.View ArticlePubMedPubMed CentralGoogle Scholar
  9. Feng J, Glass TA, Curriero FC, Stewart WF, Schwartz BS. The built environment and obesity: a systematic review of the epidemiologic evidence. Health Place. 2010;16:175–90.View ArticlePubMedGoogle Scholar
  10. Swinburn B, Vandevijvere S, Kraak V, Sacks G, Snowdon W, Hawkes C, Barquera S, Friel S, Kelly B, Kumanyika S, et al. Monitoring and benchmarking government policies and actions to improve the healthiness of food environments: a proposed Government Healthy Food Environment Policy Index. Obes Rev. 2013;14:24–37.View ArticlePubMedGoogle Scholar
  11. Popkin BM, Duffey K, Gordon-Larsen P. Environmental influences on food choice, physical activity and energy balance. Physiol Behav. 2005;86:603–13.View ArticlePubMedGoogle Scholar
  12. Black C, Moon G, Baird J. Dietary inequalities: what is the evidence for the effect of the neighbourhood food environment? Health Place. 2014;27:229–42.View ArticlePubMedGoogle Scholar
  13. Burns CM, Inglis AD. Measuring food access in Melbourne: access to healthy and fast foods by car, bus and foot in an urban municipality in Melbourne. Health Place. 2007;13:877–85.View ArticlePubMedGoogle Scholar
  14. Morland K, Diez Roux AV, Wing S. Supermarkets, other food stores, and obesity: the atherosclerosis risk in communities study. Am J Prev Med. 2006;30:333–9.View ArticlePubMedGoogle Scholar
  15. Lachat C, Nago E, Verstraeten R, Roberfroid D, Van Camp J, Kolsteren P. Eating out of home and its association with dietary intake: a systematic review of the evidence. Obes Rev. 2012;13:329–46.View ArticlePubMedGoogle Scholar
  16. Prentice AM, Jebb SA. Fast foods, energy density and obesity: a possible mechanistic link. Obes Rev. 2003;4:187–94.View ArticlePubMedGoogle Scholar
  17. Rosenheck R. Fast food consumption and increased caloric intake: a systematic review of a trajectory towards weight gain and obesity risk. Obes Rev. 2008;9:535–47.View ArticlePubMedGoogle Scholar
  18. Fleischhacker SE, Evenson KR, Rodriguez DA, Ammerman AS. A systematic review of fast food access studies. Obes Rev. 2011;12:e460–71.View ArticlePubMedGoogle Scholar
  19. Morland K, Wing S, Diez Roux A, Poole C. Neighborhood characteristics associated with the location of food stores and food service places. Am J Prev Med. 2002;22:23–9.View ArticlePubMedGoogle Scholar
  20. Filomena S, Scanlin K, Morland KB. Brooklyn, New York foodscape 2007–2011: a five-year analysis of stability in food retail environments. Int J Behav Nutr Phys Act. 2013;10:46.View ArticlePubMedPubMed CentralGoogle Scholar
  21. Cummins SC, McKay L, MacIntyre S. McDonald’s restaurants and neighborhood deprivation in Scotland and England. Am J Prev Med. 2005;29:308–10.View ArticlePubMedGoogle Scholar
  22. Macdonald L, Cummins S, Macintyre S. Neighbourhood fast food environment and area deprivation-substitution or concentration? Appetite. 2007;49:251–4.View ArticlePubMedGoogle Scholar
  23. Maguire ER, Burgoine T, Monsivais P. Area deprivation and the food environment over time: a repeated cross-sectional study on takeaway outlet density and supermarket presence in Norfolk, UK, 1990–2008. Health Place. 2015;33:142–7.View ArticlePubMedPubMed CentralGoogle Scholar
  24. Smoyer-Tomic KE, Spence JC, Raine KD, Amrhein C, Cameron N, Yasenovskiy V, Cutumisu N, Hemphill E, Healy J. The association between neighborhood socioeconomic status and exposure to supermarkets and fast food outlets. Health Place. 2008;14:740–54.View ArticlePubMedGoogle Scholar
  25. Reidpath DD, Burns C, Garrard J, Mahoney M, Townsend M. An ecological study of the relationship between social and environmental determinants of obesity. Health Place. 2002;8:141–5.View ArticlePubMedGoogle Scholar
  26. Macintyre S, McKay L, Cummins S, Burns C. Out-of-home food outlets and area deprivation: case study in Glasgow, UK. Int J Behav Nutr Phys Act. 2005;2:16.View ArticlePubMedPubMed CentralGoogle Scholar
  27. White M, Bunting J, Mathers J, Williams L, Adamson A. Do ‘food deserts’ exist A multi-level, geographical analysis of the relationship between retail food access, socio-economic position and dietary intake. Newcastle: University of Newcastle; 2004.Google Scholar
  28. Cummins S, Macintyre S. A systematic study of an urban foodscape: the price and availability of food in Greater Glasgow. Urban Stud. 2002;39:2115–30.View ArticleGoogle Scholar
  29. Pearce J, Day P, Witten K. Neighbourhood Provision of food and alcohol retailing and social deprivation in Urban New Zealand. Urban Policy Res. 2008;26:213–27.View ArticleGoogle Scholar
  30. Kestens Y, Daniel M. Social inequalities in food exposure around schools in an urban area. Am J Prev Med. 2010;39:33–40.View ArticlePubMedGoogle Scholar
  31. Clary CM, Ramos Y, Shareck M, Kestens Y. Should we use absolute or relative measures when assessing foodscape exposure in relation to fruit and vegetable intake? Evidence from a wide-scale Canadian study. Prev Med. 2015;71:83–7.View ArticlePubMedGoogle Scholar
  32. Richardson AS, Meyer KA, Howard AG, Boone-Heinonen J, Popkin BM, Evenson KR, Shikany JM, Lewis CE, Gordon-Larsen P. Multiple pathways from the neighborhood food environment to increased body mass index through dietary behaviors: a structural equation-based analysis in the CARDIA study. Health Place. 2015;36:74–87.View ArticlePubMedPubMed CentralGoogle Scholar
  33. Polsky J, Moineddin R, Dunn JR, Glazier RH, Booth GL. Absolute and relative densities of fast-food versus other restaurants in relation to weight status: Does restaurant mix matter? Prev Med (Baltim) 2015;82:28–34Google Scholar
  34. Stark JH, Neckerman K, Lovasi GS, Konty K, Quinn J, Arno P, Viola D, Harris TG, Weiss CC, Bader MD, Rundle A. Neighbourhood food environments and body mass index among New York City adults. J Epidemiol Commun Health. 2013;67:736–42.View ArticleGoogle Scholar
  35. Mercille G, Richard L, Gauvin L, Kestens Y, Payette H, Daniel M. Comparison of two indices of availability of fruits/vegetable and fast food outlets. J Urban Health. 2012;90:240–5.View ArticlePubMed CentralGoogle Scholar
  36. Luan H, Minaker LM, Law J. Do marginalized neigbourhoods have less healthy retail food environments? An analysis using Bayesian spatial latent factor and hurdle models. Int J Health Geographys. 2016;15:1–16.View ArticleGoogle Scholar
  37. Fenland Study. http://www.mrc-epid.cam.ac.uk/research/studies/fenland.
  38. UK Data Service Census Support. GeoConvert. In Book GeoConvert (Editor ed.^eds.); 2015.Google Scholar
  39. Smith D, Cummins S, Clark C, Stansfeld S. Does the local food environment around schools affect diet? Longitudinal associations in adolescents attending secondary schools in East London. BMC Public Health. 2013;13:70.View ArticlePubMedPubMed CentralGoogle Scholar
  40. Burgoine T, Monsivais P. Characterising food environment exposure at home, at work, and along commuting journeys using data on adults in the UK. Int J Behav Nut Phys Act. 2013;10:85.View ArticleGoogle Scholar
  41. Lake AA, Burgoine T, Greenhalgh F, Stamp E, Tyrrell R. The foodscape: classification and field validation of secondary data sources. Health Place. 2010;16:666–73.View ArticlePubMedGoogle Scholar
  42. Lake AA, Burgoine T, Stamp E, Grieve R. The foodscape: classification and field validation of secondary data sources across urban/rural and socio-economic classifications in England. Int J Behav Nut Phys Act. 2012;9:37.View ArticleGoogle Scholar
  43. Charreire H, Casey R, Salze P, Simon C, Chaix B, Banos A, Badariotti D, Weber C, Oppert J-M. Measuring the food environment using geographical information systems: a methodological review. Public Health Nutr. 2010;13:1773–85.View ArticlePubMedGoogle Scholar
  44. Thornton LE, Pearce JR, Kavanagh AM. Using geographic information systems (GIS) to assess the role of the built environment in influencing obesity: a glossary. Int J Behav Nutr Phys Activ. 2011;8:71.View ArticleGoogle Scholar
  45. Burgoine T, Alvanides S, Lake AA. Creating ‘obesogenic realities’; Do our methodological choices make a difference when measuring the food environment? Int J Health Geograph. 2013;12:1–9.View ArticleGoogle Scholar
  46. Super Output Area. http://www.ons.gov.uk/ons/guide-method/geography/beginner-s-guide/census/super-output-areas--soas-/index.html.
  47. Office for National Statistics. Super output Area mid-year population estimates for England and Wales, Mid-2011 (Census based); 2013. https://www.ons.gov.uk/peoplepopulationandcommunity/populationandmigration/populationestimates/datasets/lowersuperoutputareamidyearpopulationestimates. Accessed 28 Apr 2017.
  48. Chaix B, Merlo J, Evans D, Leal C, Havard S. Neighbourhoods in eco-epidemiologic research: delimiting personal exposure areas. A response to Riva, Gauvin, Apparicio and Brodeur. Soc Sci Med. 2009;69:1306–10.View ArticlePubMedGoogle Scholar
  49. Jeffery RW, Baxter J, McGuire M, Linde J. Are fast food restaurants an environmental risk factor for obesity? Int J Behav Nutr Phys Act. 2006;3:2.View ArticlePubMedPubMed CentralGoogle Scholar
  50. An R, Sturm R. School and residential neighborhood food environment and diet among California youth. Am J Prev Med. 2012;42:129–35.View ArticlePubMedPubMed CentralGoogle Scholar
  51. Bodor JN, Rose D, Farley TA, Swalm C, Scott SK. Neighbourhood fruit and vegetable availability and consumption: the role of small food stores in an urban environment. Public Health Nutr. 2008;11:413–20.View ArticlePubMedGoogle Scholar
  52. Thornton LE, Pearce JR, Macdonald L, Lamb KE, Ellaway A. Does the choice of neighbourhood supermarket access measure influence associations with individual-level fruit and vegetable consumption? A case study from Glasgow. Int J Health Geogr. 2012;11:29.View ArticlePubMedPubMed CentralGoogle Scholar
  53. English Indices of Deprivation 2010. http://data.gov.uk/dataset/index-of-multiple-deprivation.
  54. Anyaegbu G. Using the OECD equivalence scale in taxes and benefits analysis. Econ Labour Mark Rev. 2010;4:49–54.View ArticleGoogle Scholar
  55. Pearce J, Blakely T, Witten K, Bartie P. Neighborhood deprivation and access to fast-food retailing: a national study. Am J Prev Med. 2007;32:375–82.View ArticlePubMedGoogle Scholar
  56. Richardson AS, Boone-Heinonen J, Popkin BM, Gordon-Larsen P. Are neighbourhood food resources distributed inequitably by income and race in the USA? Epidemiological findings across the urban spectrum. BMJ Open. 2012;2:e000698.View ArticlePubMedPubMed CentralGoogle Scholar
  57. Blakely T, Woodward AJ. Ecological effects in multi-level studies. J Epidemiol Commun Health. 2000;54:367–74.View ArticleGoogle Scholar
  58. Riva M, Gauvin L, Apparicio P, Brodeur J-M. Disentangling the relative influence of built and socioeconomic environments on walking: the contribution of areas homogenous along exposures of interest. Soc Sci Med. 2009;69:1296–305.View ArticlePubMedPubMed CentralGoogle Scholar
  59. Boyle MH, Willms JD. Place effects for areas defined by administrative boundaries. Am J Epidemiol. 1999;149:577–85.View ArticlePubMedGoogle Scholar
  60. Smith G, Gidlow C, Davey R, Foster C. What is my walking neighbourhood? A pilot study of English adults’ definitions of their local walking neighbourhoods. Int J Behav Nutr Phys Act. 2010;. doi:10.1186/1479-5868-1187-1134.Google Scholar
  61. Golledge RG, Stimson RJ. Spatial behavior: a geographic perspective. New York: The Guilford Press; 1997.Google Scholar
  62. Kestens Y, Lebel A, Daniel M, Thériault M, Pampalon R. Using experienced activity spaces to measure foodscape exposure. Health Place. 2010;16:1094–103.View ArticlePubMedGoogle Scholar
  63. Kwan M-P. The uncertain geographic context problem. Ann Assoc Am Geogr. 2012;102:37–41.Google Scholar
  64. Mehta NK, Chang VW. Weight status and restaurant availability a multilevel analysis. Am J Prev Med. 2008;34:127–33.View ArticlePubMedPubMed CentralGoogle Scholar
  65. Powell LM, Chaloupka FJ, Bao Y. The availability of fast-food and full-service restaurants in the United States: associations with neighborhood characteristics. Am J Prev Med. 2007;33:S240–5.View ArticlePubMedGoogle Scholar
  66. Mason KE, Bentley RJ, Kavanagh AM. Fruit and vegetable purchasing and the relative density of healthy and unhealthy food stores: evidence from an Australian multilevel study. J Epidemiol Commun Health. 2013;67:231–6.View ArticleGoogle Scholar
  67. Kestens Y, Lebel A, Chaix B, Clary C, Daniel M, Pampalon R, Theriault M. SV PS: association between activity space exposure to food establishments and individual risk of overweight. PLoS One. 2012;7:e41418.View ArticlePubMedPubMed CentralGoogle Scholar
  68. Mercille G, Richard L, Gauvin L, Kestens Y, Shatenstein B, Daniel M, Payette H. Associations between residential food environment and dietary patterns in urban-dwelling older adults: results from the VoisiNuAge study. Public Health Nutr. 2012;15:2026–39.View ArticlePubMedGoogle Scholar
  69. Flowerdew R, Manley DJ, Sabel CE. Neighbourhood effects on health: does it matter where you draw the boundaries? Soc Sci Med. 2008,66:1241–55.Google Scholar
  70. Oliver LN, Schuurman N, Hall AW. Comparing circular and network buffers to examine the influence of land use on walking for leisure and errands. Int J Health Geogr 2007;6:1–11.Google Scholar
  71. Burgoine T, Alvanides S, Lake AA. Creating ‘obesogenic realities’; Do our methodological choices make a difference when measuring the food environment? Int J Health Geogr. 2013;12:33.View ArticlePubMedPubMed CentralGoogle Scholar
  72. Kerr J, Frank L, Sallis JF, Saelens B, Glanz K, Chapman J. Predictors of trips to food destinations. Int J Behav Nutr Phys Act. 2012;9:58.View ArticlePubMedPubMed CentralGoogle Scholar
  73. Drewnowski A, Aggarwal A, Hurvitz PM, Monsivais P, Moudon AV. Obesity and supermarket access: proximity or price? Am J Public Health. 2012;102:e74–80.View ArticlePubMedPubMed CentralGoogle Scholar
  74. Sampson R, Morenoff J. Durable inequality: spatial dynamics, social processes and the persistence of poverty in Chicago neighborhoods. In: Bowles S, Durlauf S, Hoff K, editors. Poverty traps. Princeton: Princeton University Press; 2006. p. 176–203.Google Scholar
  75. Zenk SN, Schulz AJ, Matthews SA, Odoms-Young A, Wilbur J, Wegrzyn L, Gibbs K, Braunschweig C, Stokes C. Activity space environment and dietary and physical activity behaviors: a pilot study. Health Place. 2011;17:1150–61.View ArticlePubMedPubMed CentralGoogle Scholar
  76. Christian WJ. Using geospatial technologies to explore activity-based retail food environments. Spat Spatiotemporal Epidemiol. 2012;3:287–95.View ArticlePubMedGoogle Scholar
  77. The effects of taxes and benefits on household income. https://www.ons.gov.uk/peoplepopulationandcommunity/personalandhouseholdfinances/incomeandwealth/datasets/theeffectsoftaxesandbenefitsonhouseholdincomefinancialyearending2014.
  78. Local Area Analysis of Qualifications Across England and Wales. http://webarchive.nationalarchives.gov.uk/20160105160709/http://www.ons.gov.uk/ons/dcp171776_355401.pdf.
  79. Choudhury Y, Hussain I, Parsons S, Rahman A, Eldridge S, Underwood M. Methodological challenges and approaches to improving response rates in population surveys in areas of extreme deprivation. Prim Health Care Res Dev. 2012;13:211–8.View ArticlePubMedGoogle Scholar
  80. Lebel A, Daepp MIG, Block JP, Walker R, Lalonde B, Kestens Y, Subramanian SV. Quantifying the foodscape: a systematic review and meta-analysis of the validity of commercially available business data. PLoS ONE. 2017;12:1–17.View ArticleGoogle Scholar

Copyright

© The Author(s) 2017