Accessibility of fast food outlets is associated with fast food intake. A study in the Capital Region of Denmark

Health & Place 48 (2017) 102–110

Contents lists available at ScienceDirect

Health & Place journal homepage: www.elsevier.com/locate/healthplace

Accessibility of fast food outlets is associated with fast food intake. A study in the Capital Region of Denmark

MARK

⁎

Kamille Almer Bernsdorf a, , Cathrine Juel Laua, Anne Helms Andreasena, Ulla Tofta, Maja Lykkea, Charlotte Glümera,b,1 a Research Centre for Prevention and Health, Capital Region of Denmark, Rigshospitalet – Glostrup, Nordre Ringvej 57, Section 84/85, 2600 Glostrup, Denmark b Department of Health Sciences and Technology, Aalborg University, Fredrik Bayers vej 7D2, DK-9220 Aalborg, Denmark

A R T I C L E I N F O

A BS T RAC T

Keywords: Fast food Accessibility GIS Density Proximity SES

Literature suggests that people living in areas with a wealth of unhealthy fast food options may show higher levels of fast food intake. Multilevel logistic regression analyses were applied to examine the association between GIS-located fast food outlets (FFOs) and self-reported fast food intake among adults (+ 16 years) in the Capital Region of Denmark (N = 48,305). Accessibility of FFOs was measured both as proximity (distance to nearest FFO) and density (number of FFOs within a 1 km network buffer around home). Odds of fast food intake ≥ 1/ week increased significantly with increasing FFO density and decreased significantly with increasing distance to the nearest FFO for distances ≤ 4 km. For long distances (> 4 km), odds increased with increasing distance, although this applied only for car owners. Results suggest that Danish health promotion strategies need to consider the contribution of the built environment to unhealthy eating.

1. Background Food environments are built environments described by the location of food outlets (FOs), and access to these environments is theorized to influence individual dietary patterns and, ultimately, risk of obesity and chronic diseases (Caspi et al., 2012). The accessibility of food is often defined by geographical measures from home to FOs. Specific measures hypothesized to be important contributors to eating patterns are proximity and density of different types of FOs (BooneHeinonen et al., 2011b; Cobb et al., 2015; Dunn et al., 2012; Fraser et al., 2010; Gamba et al., 2014; Longacre et al., 2012; Moore et al., 2009; Oexle et al., 2015; Richardson et al., 2011; Thornton et al., 2009; Turrell and Giskes, 2008). Fast food outlets (FFOs) generally tend to serve foods with a higher energy density and poorer nutritional quality than foods prepared at home (Moore et al., 2009; Powell et al., 2012). Eating fast food has been associated with poor dietary habits, such as higher intakes of energy, fat, sodium, added sugars and sugar-sweetened beverages, and lower intakes of fruit, vegetables, fibre and milk (Bowman and Vinyard, 2004; Fraser et al., 2010; Lachat et al., 2012; Orfanos et al., 2009; Richardson et al., 2015). Furthermore, eating fast food has been associated with an

increased risk of obesity and other health-related factors such as insulin resistance (Laxy et al., 2015; Pereira et al., 2005; Richardson et al., 2015). Consequently, there is increasing interest to assess the influence of accessibility of FFOs on health-related parameters. High accessibility of FFOs in a neighbourhood has been associated with high fast food intake (Boone-Heinonen et al., 2011b; Longacre et al., 2012; Moore et al., 2009; Richardson et al., 2015; Thornton et al., 2009), unhealthy dietary habits (He et al., 2012; Moore et al., 2009; Richardson et al., 2015) and a higher prevalence of obesity (Burgoine et al., 2016; Cobb et al., 2015; Dunn et al., 2012; Gamba et al., 2014). However, research has reported conflicting findings. Several reviews have highlighted considerable heterogeneity in measures and techniques within the geographical information system (GIS) research, which is commonly used to describe accessibility (Caspi et al., 2012; Charreire et al., 2010; Cobb et al., 2015; Gamba et al., 2014; McKinnon et al., 2009; Wilkins et al., 2017). According to Wilkins et al. (2017), there are commonly five dimensions of methodological diversity: the choice of FO data, the methods used to extract FO data of interest, the ways that FOs are defined, the geocoding methods used and the ways that FO access is operationalized. Wilkins et al. (2017) state that “while most authors acknowledge these limitations, an

⁎

Corresponding author. E-mail addresses: [email protected] (K.A. Bernsdorf), [email protected] (C.J. Lau), [email protected] (A.H. Andreasen), [email protected] (U. Toft), [email protected] (M. Lykke), [email protected] (C. Glümer). 1 Present address: Center for diabetes, Municipality of Copenhagen, Vesterbrogade 121, 3rd floor, 1620 Copenhagen V, Denmark. http://dx.doi.org/10.1016/j.healthplace.2017.10.003 Received 29 November 2016; Received in revised form 22 September 2017; Accepted 3 October 2017 1353-8292/ © 2017 Elsevier Ltd. All rights reserved.

Health & Place 48 (2017) 102–110

K.A. Bernsdorf et al.

To ensure that preventive efforts and Danish public health policies can focus on the built environments, geographical areas and population sub-groups where most benefit can be expected, it is essential to identify which characteristics may promote health-related behaviour (e.g. fast food intake) among the Danish population. GIS is widely used by researchers for measuring food environments and by town planners and local authorities in developing policy and making planning decisions (Glanz et al., 2016). Consequently, this makes GIS-based research particularly relevant to policy development. Thus, in order to facilitate an effective translation of research into practice, the present study seeks to be transparent regarding choices within the GIS technology used to analyse the two aims of this study: 1) to examine the association between FFO accessibility and fast food intake, and 2) to examine whether this association is modified by area SES and urbanicity.

absence of best practices means the problems look set to persist. With such diversity in methods, accurate and transparent reporting is essential”. Another issue potentially contributing to the conflicting findings is that evidence mainly stems from the US and Australia. European and specifically Danish studies are sparse. While Denmark may share some similarities with the US and Australia, there are distinct geographical, political, economic, commercial, social and cultural differences between the continents in relation to planning, distribution and usage of FOs. In the US, for example, the food environment often consists of rural or urban low-income areas with limited access to affordable and nutritious food, i.e. ‘food deserts’ (Beaulac et al., 2009). Such differences make it difficult to translate international findings into a Danish context. The food environment can be associated with the socioeconomic status (SES) of an area in several ways, e.g. by grouping of specific types of people, FOs and facilities. Furthermore, the social position of an individual may influence the choice of residential location and as such the potential environment and its built characteristics, such as accessibility of food (Sushil et al., 2017; Voigtländer et al., 2013) Studies in the US, Canada, Australia and the UK have shown that the accessibility of FFOs is higher in deprived areas than in non-deprived areas (Black et al., 2014a, 2014b; Fraser et al., 2010; Laxy et al., 2015; Mozaffarian et al., 2012; Richardson et al., 2015; Rummo et al., 2017). Access to food may also vary with urban-rural residence in Denmark. Particularly in rural and frontier areas FFOs may be limited, while greater access may be found in urban areas (Pearce et al., 2007; Powell et al., 2007; Thornton et al., 2016). In Denmark, access to convenience stores has been positively associated with unhealthy diet only in nonmetropolitan areas, suggesting a role for urbanicity in the potential association between access and intake of fast food (Lind et al., 2016). Whether these conditions are mirrored in Denmark has not been examined, but such associations could have great importance for the persistence of health inequalities that we see across place of residence in Denmark (Christensen et al., 2014a; Lau et al., 2015; Macintyre et al., 2002; Mendis and Banerjee, 2010; Robinson et al., 2014).

2. Methods 2.1. The Danish Capital Region Health Survey The present study is based on data from the Danish Capital Region Health Survey, a cross-sectional survey conducted in the 29 municipalities of the Capital Region of Denmark (Christensen et al., 2012; HammerHelmich et al., 2011). The survey was conducted from February to May 2010. A random sample of individuals was drawn from the Danish Civil Registration System (CRS). CRS identifies all inhabitants in Denmark by a unique 10-digit personal identification (CPR) number that allows record linkage on an individual level of data to national registers. The survey sample included 95,150 individuals. Copenhagen Municipality was divided into ten areas according to official administrative districts (Fig. 1), and these were treated as individual municipalities in the sampling process, resulting in a total of 38 municipalities. A random sample of 2450 individuals aged 16 years or older was drawn from each municipality. Due to differences in population size, the sample size in Frederiksberg Municipality was 4500 individuals. Each individual received a mailed invitation and a paper questionnaire

Fig. 1. The Capital Region of Denmark in 2010 divided into four municipality SES groups and urbanicity. Light purple - Municipality SES group 1 (Most affluent); purple - Municipality SES group 2; light orange - Municipality SES group 3; orange - Municipality SES group 4 (Less affluent). Municipalities within the solid line comprise the Copenhagen inner-city area defined as urban. This area comprises ten official administrative districts and Frederiksberg Municipality, which are highlighted in the box to the right. Municipalities between the dotted and solid line comprise six municipalities of Greater Copenhagen defined as suburban areas. Remaining municipalities are defined as rural. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.).

103

Health & Place 48 (2017) 102–110

K.A. Bernsdorf et al.

identified in metres. Based on national travel trends, cubic spline modelling and Akaike's information criterion (AIC), fast food proximity was defined by three distance intervals: < 1 km (short distances), 1– 4 km and > 4 km (long distances) (Technical University of Denmark, 2013b).

(a web-based version was also available). The questionnaire contained questions on health behaviour. The response rate was 52.3% (N = 49,806). Among those who did not respond to the questionnaire, a higher proportion were men, were unemployed, had a short education, had a low gross income and were of another ethnicity than Danish (Hammer-Helmich et al., 2011). Of those who participated, a minority did not inform of their home address (N = 14). Of the remaining respondents, some chose not to provide information on frequency of fast food intake (N = 404), educational level and ethnicity (N = 1083). Thus, the final study population included 48,305 individuals.

2.4. Individual-level variables Individual-level characteristics included gender, age, educational level and ethnicity. Information on age and gender was obtained from CRS (Pedersen et al., 2006), while data on educational level (highest completed education) and ethnicity was drawn from the Danish Population's Education Register (Jensen and Rasmussen, 2011) and Statistics Denmark (Norredam et al., 2011), respectively. This information was linked with questionnaire data on fast food intake using CPR numbers. For descriptive purposes, age was categorized into seven groups: 16–24 years, 25–34 years, 35–44 years, 45–54 years, 55–64 years, 65–79 years and 80 + years. Educational level was categorized into four groups: primary or secondary school; vocational education; academy or bachelor degree; and Master or PhD degree. Educational level was used as an indicator for individual socioeconomic status (SES) since there is considerable evidence demonstrating that an individual's educational status is an important predictor of dietary patterns (Groth et al., 2014). Furthermore, individual SES may also influence the association between fast food access and consumption (Boone-Heinonen et al., 2011b). Ethnicity was categorized into three groups: Danish, Western or non-Western background. This was based on information about own and/or parents’ country of birth. Supplementary analyses included information on car ownership for the respondent or spouse (yes/no), which was drawn from the Digital Motor Register at Statistics Denmark (Statistics Denmark, 2017).

2.2. Outcome – fast food intake Frequency of fast food intake was assessed from the survey questionnaire. Participants were asked the following question: “How often do you eat fast food (pizza, burgers, sausages, shawarma etc.)?”, choosing between six possible responses (rarely or never, 1–3 times/ month, 1–2 times/week, 3–4 times/week, 5–7 times/week, and more than once/day). Based on the distribution of answers and with the intention to reflect frequent intake, outcome was dichotomized into fast food intake one or more times/week (yes/no). 2.3. Exposure - fast food outlet accessibility To decide whether an FO was an FFO or not, FFOs were defined according to the following criteria: food retailers primarily serving pizza, burger, French fries, sausages, barbeque food or shawarma/ kebab AND at least two of the following characteristics: take-away food, customers pay before eating, limited or no table service, limited furnishing. For identification and geographical location of relevant FFOs, information was extracted from the national food safety and hygiene regulation register (Smiley Register), which is administered by the Ministry of Environment and Food (Ministry of Environment and Food of in Denmark, 2017). For identification of relevant FFOs, we used the Danish Industrial Classification system DB03 with relevant branch codes (553020 and 553010). The first four digits correspond to the EU's industry classification NACE rev. 1.1, whereas the last two digits are Danish sub-groupings corresponding to the UN's industrial classification ISIC (Statistics Denmark, 2005). These categories included FFOs but also full-service restaurants, ice cream bars etc. Therefore, branch codes were combined with common restaurant names. Thus, relevant FFOs were defined as restaurants within the abovementioned branch codes AND with a restaurant name that included foods defined as fast food (i.e. pizza, burger, sausages, barbeque (grill), kebab, shawarma and falafel) OR being a part of the large fast food chains McDonalds, Burger King, Sunset or Kentucky Fried Chicken. From the register, universal transverse mercator (UTM) coordinates were extracted for each FFO. Names were checked by a researcher and coordinates were verified through ground-truthing of a random sample (May-June 2010). The sensitivity and positive predictive value of the Smiley Register was 82% and 92%, respectively (Toft et al., 2011). Individual accessibility of fast food restaurants was described by both a density and proximity measure using geocoding and network analyses in Network Analyst and ArcGIS 9.3. Network buffers of 1 km around each respondent's home were used to calculate the number of FFOs within this area, i.e. the density (Thornton et al., 2011). The intent was to capture all FFOs accessible primarily by walking around the local neighbourhood. This was based on data from the Danish National Travel Survey, showing that 73% of all trips on foot are less than 1 km and with an average trip length between home and shopping of 0.8 km (Technical University of Denmark, 2013a). Home addresses were geocoded based on the exact address obtained from CRS. Together, this information defined the density variable, which was categorized into the following six categories: 0, 1, 2, 3–5, 6–10 and 11 + FFOs. Proximity to the nearest FFO from the respondent's home was

2.5. Area-level variables Differences in fast food intake may be accounted for by specific characteristics of the area such as the SES of an area and urbanicity. Area SES was defined from the SES of the 38 municipalities using a composite measure across each municipality. This measure considers the distribution of residents with the highest attained educational level, the distribution of the residential employment status and the mean gross income of the municipality (Appendix A). Information was derived from central registers, i.e. the Danish Population's Education Register, the Employment Classification Module and the Income Statistics Register (Baadsgaard and Quitzau, 2011; Jensen and Rasmussen, 2011; Petersson et al., 2011). This resulted in a variable of four municipality SES groups of decreasing affluence, with municipality SES group 1 including the most affluent municipalities (Fig. 1). Urbanicity was based on information on place of residence drawn from the CPR numbers and categorized into urban, suburban and rural. Urban areas included the ten official administrative districts of Copenhagen as well as Frederiksberg Municipality, which is surrounded by these districts (Fig. 1). Thus, urban areas comprise the Copenhagen inner-city metropolitan area. The remaining municipalities were defined as suburban areas if they were located in the remaining area within a 1 km metropolitan polygon, and as rural areas if they were located in the remaining area outside a 1 km metropolitan polygon (Ministry of Business and Growth Denmark, 2015). Consequently, suburban areas included the six municipalities of Greater Copenhagen; Taarnby, Gentofte, Gladsaxe, Herlev, Roedovre and Hvidovre, while rural areas comprised the remaining 21 municipalities. 2.6. Statistical analyses A total of 48,305 individuals were included in the statistical analyses performed using survey procedures in SAS statistical software (version 9.4, SAS Institute Inc., Cary, NC, USA). We applied multilevel 104

Health & Place 48 (2017) 102–110

K.A. Bernsdorf et al.

logistic regression analyses (GLIMMIX procedure) to investigate the association between fast food access (density and proximity) and frequent fast food intake (≥ once/week). All analyses were weighted to account for the difference in non-response by socio-demographic characteristics and cause-specific morbidity and mortality (Christensen et al., 2015, 2014b). Thus, results are representative for the entire population in the Capital Region in 2010. The multilevel model strategy accounts for the fact that individuals residing within the same area/ unit tend to be more alike in their health behaviour given that they are exposed to the same characteristics, e.g. they use the same facilities and face the same social structure (Cummins et al., 2007; Merlo et al., 2005; Voigtländer et al., 2013). Thus, the hierarchical structure of the two-level model comprised information on residents (level 1) nested within municipalities (level 2) (Merlo et al., 2005). Although parishes are the smallest official administrative units defining a geographical area in Denmark, the present study used municipalities to comprise the second level of the multilevel model. This was due to the computation of weights from Statistics Denmark, which were based on the stratified sampling design of the municipalities. Thus, letting parishes comprise the second level of the model would not allow us to weigh the statistical analyses for sampling design and non-response. Consequently, the area-level was defined by administrative municipality boundaries and estimated as random effects, while individual-level and area-level characteristics were estimated as fixed effects (Merlo et al., 2005). Weights were computed by Statistics Denmark based on information about gender, age, municipality, educational level, income, civil status and hospitalization (Christensen et al., 2012). Six models were analysed (Table 1). Model 1 included no explanatory variables, while model 2 included individual accessibility of FFOs (i.e. density or proximity). In model 3, adjustments were made for individual characteristics, i.e. gender, age and ethnicity. Model 4 included variables from model 3 and educational level. A multilevel interaction term was also added to model 4 to explore whether fast food access had a different impact according to individual educational level. Thus, for model 4 we also tested for interaction between FFO access and educational level. Model 5 included variables from model 4 and urbanicity of the municipality. For model 5, we also tested for crosslevel interaction (e.g. environment-individual associations) to examine whether associations were modified by area characteristics. Thus, cross-level associations were tested between FFO access and urbanicity. Model 6 included variables from model 5 and municipality SES. Finally, in relation to model 6, we also tested for cross-level interaction between FFO access and municipality SES. For associations between FFO density and fast food intake, estimates are presented as odds ratios (ORs) with 95% confidence intervals (CIs), indicating the odds of frequent fast food intake relative to a reference group. Most variables included in the analyses were categorical. Based on AIC, age was included in the analyses as a linear

Table 2 Characteristics of the study population in the Capital Region of Denmark 2010. Fast food intake

Gender* Women Men Age* 16–24 years 25–34 years 35–44 years 45–54 years 55–64 years 65–79 years 80+ years Ethnicity* Danish Western background Other Educational level* Primary or secondary school Vocational education Academy or bachelor degree Master or PhD degree Urbanicity* Urban Suburban Rural Municipality socioeconomic status group* 1 (Most affluent) 2 3 4 (Less affluent) a *

≥ Once per week n = 6203

< Once per week n = 42,102

%a

Number

%a

Number

29.5 70.5

2176 4027

55.7 44.3

24,925 17,177

29.4 27.0 22.7 12.6 4.9 2.7 0.7

1690 1441 1411 951 410 250 50

9.9 13.8 18.5 18.0 17.8 17.0 5.0

3390 4485 7066 7918 5568 8707 1671

81.5 4.7 13.8

5413 186 604

88.0 5.2 6.8

39,080 1402 1620

46.9

2682

33.0

12,380

24.1 18.1

1573 1200

30.5 23.5

13,605 1552

10.9

748

13.0

5565

38.7 56.9 4.4

2397 3508 298

27.5 65.4 7.1

10,890 28,106 3106

18.4 30.6 22.9 28.1

1164 1777 1497 1765

24.7 31.7 21.6 22.0

10,744 12,055 9664 9639

Weighted for non-response and survey design. Chi2 test p < 0.0001.

spline representing a piecewise linear interpolation with two nodes at 50 years and 66 years, respectively. Based on AIC, FFO proximity was included in the analyses as a linear spline representing a piecewise linear interpolation with two nodes at 1 km and 4 km, respectively. Thus, for associations with FFO proximity, predicted values express how much a resident's odds of eating fast food at least once a week would be for each 100 m increase in distance to the nearest FFO up to a distance < 1 km. For distances from 1 to 4 km and > 4 km, respectively, estimates express the odds for each 1 km increase in distance. 3. Results

Table 1 Multilevel regression analysis models of fast food intake ≥ once/week in 38 municipalities in the Capital Region of Denmark. For each model, variables are additionally included to the previous model.

3.1. Characteristics of population A total of 6203 participants (16.2%) reported consuming fast food at least once per week (Table 2). Frequent intake of fast food was highest among men, younger age groups, individuals of Danish origin, individuals with shorter education, and among those living in suburban municipalities. Frequent intake of fast food was lowest among the most affluent municipalities.

Models Fast food intake Individual-level variables Model 1 + Fast food outlet access Model 2 + Age, gender, ethnicity Model 3 + Educational level Area-level variables Model 4 + Urbanicity Model 5 + Municipality socioeconomic status

Model 1 Model 2 Model 3 Model 4a

3.2. Accessibility of fast food outlets

Model 5b Model 6c

Low FFO density and long distances to the nearest FFO were more prevalent in municipalities of higher SES (group 1) (Table 3). No strong gradient was found in accessibility across municipality SES. A high FFO density was more prevalent in urban areas, and the majority of residents (90%) within rural areas did not have access to an FFO

a

We also tested for interaction between fast food outlet access*educational level. We also tested for interaction between fast food outlet access and urbanicity. c We also tested for interaction between fast food outlet access*municipality socioeconomic status. b

105

Health & Place 48 (2017) 102–110

K.A. Bernsdorf et al.

3.4. Associations between FFO proximity and fast food intake

Table 3 Accessibility of fast food outlets (FFOs) across municipality socioeconomic status (SES) and urbanicity in the Capital Region of Denmark 2010. Results are shown as frequencies (%) for FFO density and as mean kilometre (km) for FFO proximity. Municipality SES group

FFO density 0 1 2 3–5 6–10 11+ Total (%) FFO proximity km (mean)

Table 5 shows results from the multilevel analyses of FFO proximity. Distance to the nearest FFO was significantly associated with frequent fast food intake (p < 0.0001). A resident's odds of eating fast food at least once a week decreased for each 100 m increase within short distances to the nearest FFO (< 1 km). Similarly, a decrease in odds was also found within distances of 1–4 km, although this decrease was for each 1 km in distance to the nearest FFO. In contrast, for residents with a long distance to the nearest FFO (> 4 km) an increase in odds of frequent fast food intake was seen for each 1 km increase in distance. These findings remained the same when both individual and area characteristics were included in the analyses. Educational level did not modify the association between FFO proximity and fast food intake (model 4, p = 0.08). However, both urbanicity and municipality SES significantly modified the association (model 5 and 6, p < 0.001). Cross-level interactions with urbanicity are seen in Fig. 3. Cross-level interactions with municipality SES showed that residents within municipality SES group 1 with a long distance to the nearest FFO (> 4 km) had a significant increase in odds of frequent fast food intake for each 1 km increase in distance (Results not shown). In supplementary analyses, we found that car ownership significantly modified the association between FFO proximity and fast food intake (model 4, p < 0.0001). Thus, the association with long distances to the nearest FFO (> 4 km) was only significant among residents who owned a car (OR (95% CI) = 1.08 (1.05–1.11)).

Urbanicity

1

2

3

4

Urban

Suburban

Rural

40.63 21.33 10.97 20.95 5.50 0.61 100

23.30 8.74 8.31 13.41 13.24 33.00 100

14.04 12.76 13.85 23.18 21.47 14.70 100

16.15 11.11 9.65 16.16 16.96 29.97 100

0.70 1.36 2.42 7.68 22.75 65.08 100

27.32 19.07 15.07 24.53 11.55 2.46 100

90.45 7.99 1.50 0.05 0 0 100

1.04

1.03

0.56

0.97

0.26

0.90

4.02

within 1 km of their home. Distance to the nearest FFO increased with decreasing urbanicity. Thus, FFO access was limited particularly in rural and suburban areas, while greater access was found in urban areas.

3.3. Associations between FFO density and fast food intake Table 4 shows results of the multilevel analyses of FFO density. FFO density was significantly associated with frequent fast food intake (p < 0.0001), showing a trend towards having an increased odds ratio of frequent fast food intake with increasing numbers of FFOs within 1 km from the resident's home. This finding remained when individual and area characteristics were included in the analyses. Thus, in the fully adjusted model, odds of frequent fast food consumption were 1.37 times higher for individuals with 11 + FFOs within 1 km compared to individuals with no FFOs within 1 km. Neither educational level nor urbanicity modified the association between FFO density and fast food intake (model 4, p = 0.45 and model 5, p = 0.63, respectively). However, municipality SES significantly modified the association (model 6, p = 0.001). Within municipality SES group 2–4, cross-level interactions showed that high FFO density was significantly associated with frequent fast food intake (Fig. 2).

4. Discussion We found an association between FFO access and fast food intake in the Capital Region of Denmark. Regardless of both individual and area characteristics, the likelihood of frequent fast food intake increased with increasing FFO density. Similarly, a resident's odds of frequent fast food intake decreased significantly with increasing distance to the nearest FFO (for distances up to 4 km). For distances greater than 4 km, the opposite association was found, although this applied only for car owners. The associations were independent of individual SES, but were modified by characteristics of the residential area; municipality SES modified the association between FFO density and fast food intake, while both urbanicity and municipality SES modified association with FFO proximity. These cross-level interactions suggest that within certain population sub-groups, the built food environment could influence dietary behaviour. Thus, Danish preventive efforts and public health policy aimed at decreasing unhealthy food habits such as fast food intake may benefit from focusing on limiting the number of FFOs within less affluent municipalities. However, this interaction could also be due to retailers targeting these areas due to cheaper land prices, or it may be a result of a demand-led process within these areas. Regardless, this is an important issue to highlight to policy decision makers as land use restrictions on incoming FFOs in less affluent municipalities may help influence the association. To our knowledge, this study is the first to examine the association between fast food access and intake in an adult population within a Danish context. Previous research examining the association between fast food access and intake has reported conflicting findings (BooneHeinonen et al., 2011b; Dunn et al., 2012; Moore et al., 2009; Oexle et al., 2015; Richardson et al., 2011; Thornton et al., 2009; Turrell and Giskes, 2008). For example, in line with our study, Boone-Heinonen et al. (2011) and Moore et al. (2009) found a positive association between increased access to fast food and fast food consumption among young and middle-aged adults. However, in the study by Boone-Heinonen et al. (2011) the association was only seen among low-income men. In contrast to these findings, Oexle et al. (2015) found no significant association between GIS-based presence of FFOs and weekly fast food consumption among adults living in the region of

Table 4 The association between fast food outlet (FFO) density and frequent fast food intake (≥ once/week) in the Capital Region of Denmark 2010. Odds ratio for frequent fast food intake (95% CI)a FFO density Model 1 Model 2

Model 3

Model 4

Model 5

Model 6

1 – 1.24 (1.12– 1.37) 1.23 (1.09– 1.39) 1.24 (1.10– 1.39) 1.21 (1.07– 1.37) 1.18 (1.05– 1.34)

2 – 1.33 (1.13– 1.57) 1.30 (1.15– 1.48) 1.31 (1.60– 1.48) 1.27 (1.12– 1.45) 1.24 (1.09– 1.41)

3–5 – 1.28 (1.15– 1.42) 1.37 (1.25– 1.51) 1.37 (1.25– 1.50) 1.32 (1.18– 1.49) 1.30 (1.16– 1.45)

6–10 – 1.43 (1.29– 1.58) 1.49 (1.33– 1.66) 1.49 (1.36– 1.67) 1.39 (1.22– 1.59) 1.38 (1.22– 1.55)

11 + – 1.58 (1.37– 1.81) 1.50 (1.34– 1.66) 1.54 (1.37– 1.71) 1.37 (1.20– 1.57) 1.37 (1.20– 1.57)

a Estimates are shown as odds ratios (ORs) with 95% confidence intervals (CIs), indicating the odds of frequent fast food intake relative to the reference group (Fast food density = 0.).

106

Health & Place 48 (2017) 102–110

K.A. Bernsdorf et al.

Fig. 2. Association between fast food outlet density and frequent fast food intake across municipality socioeconomic status (SES) in the Capital Region of Denmark 2010. Results are shown as odds ratios (ORs) for frequent fast food intake adjusted for age, gender, ethnicity, educational level, urbanicity and municipality SES (model 6). Reference: fast food density = 0 within the same municipality SES group.**ORs with significance level of p < 0.05.

of GIS technology. The appropriate size of the area to be used for calculating accessibility measures remains subject to debate (Charreire et al., 2010; Wilkins et al., 2017). We emphasize the strength of estimating the density of FFOs from individual-level measures (i.e. around the place of residence) rather than from the centroid of large administrative areas (Thornton et al., 2011). Conversely, Lyseen et al. (2015) state that areas defined from a single fixed location fail to capture people's complete exposure. This is reasoned by the increasing connectivity within the built environment which makes the human mobility pattern complex. Therefore, Lyseen and colleagues emphasize the need for methods, concepts and measures of individual activity and exposure, e.g. GPS tracking (Lyseen et al., 2015). This consideration is supported by our supplementary analyses including information on car ownership. However, there are advantages and drawbacks of each method of defining accessibility. For example, GPS tracking is subject to technical limitations such as lack of connection to an adequate number of satellites. Furthermore, GPS tracking creates enormous amounts of data, which hampers data processing. Within the field of accessibility and health, there are other limitations than the choice of accessibility measure. These involve the heterogeneity in study designs, populations and outcome measures (Black et al., 2014a, 2014b; Charreire et al., 2010; Cobb et al., 2015; Gamba et al., 2014; McKinnon et al., 2009). For example, dietary habits such as fast food intake are determined and measured with great variability. Furthermore, in epidemiological studies such as the present study, predictors of dietary habits are obliged to be crude and reliant on

Table 5 The association between fast food outlet (FFO) proximity and frequent fast food intake (≥ once/week) in the Capital Region of Denmark 2010. Odds ratio for frequent fast food intake (95% CI) FFO proximity Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 a b

< 1 kma – 0.96 (0.95–0.97) 0.97 (0–96 to 0.98) 0.97 (0.96–0.98) 0.97 (0.96–0.98) 0.97 (0.96–0.98)

1–4 kmb – 0.94 (0.87–1.01) 0.89 (0.83–0.96) 0.89 (0.83–0.96) 0.91 (0.85–0.99) 0.92 (0.85–0.99)

> 4 kmb – 1.03 (1.01–1.06) 1.07 (1.04–1.09) 1.06 (1.04–1.09) 1.06 (1.04–1.08) 1.05 (1.03–1.08)

Reference: distance to the nearest FFO minus 100 m (for distances < 1 km). Reference: distance to the nearest FFO minus 1000 m.

Central South Carolina. Richardson et al. (2011) support this finding in a rather large population of young American adults. As mentioned, the conflicting findings may be due to methodological diversity of dimensions within the GIS research. A systematic review by Cobb et al. (2015) found that out of 45 studies using FFOs as an exposure, accessibility was measured in 31 different ways. Differences apply both to proximity and density measures as well as the geographical boundary from which these are determined. Thus, in order to facilitate an effective translation of research into practice, the present study seeks to be transparent regarding choices within the use

Fig. 3. Association between fast food outlet (FFO) proximity and frequent fast food intake across municipality urbanicity in the Capital Region of Denmark 2010. Reference: distance to the nearest FFO minus 100 m (for distances < 1 km) and distance to the nearest FFO minus 1000 m (for distances > 1 km). ** Odds ratios (ORs) with significance level of p < 0.05. Results are shown as ORs of eating fast food at least once a week for each 100 m increase in distance to the nearest FFO up to a distance < 1 km, and for each 1 km increase in distance to the nearest FFO for distances > 1 km within the same urbanicity. In urban municipalities, FFOs did not exist within distances > 4 km. ORs are adjusted for age, gender, ethnicity, educational level and urbanicity (model 5).

107

Health & Place 48 (2017) 102–110

K.A. Bernsdorf et al.

unhealthy dietary patterns. Furthermore, within similar types of FFO there can be variability in relation to products sold, price and marketing strategies. Moreover, fast food access might not be restricted to FFOs, and thus presence of FFOs does not capture all characteristics of the fast food environment (Cobb et al., 2015; Gamba et al., 2014). Consequently, the chosen FFO definition may be a conservative estimate of fast food access. In addition, including several types of FOs could yield a more accurate measurement of the food environment, its relationship to individual dietary habits and, ultimately, the risk of obesity and chronic diseases. Glanz and colleagues have identified two aspects of the food environment. These include the community nutritional environment defined by proxy measures, as in the present study (e.g. the number, type, location and accessibility of FOs), and the consumer nutritional environment defined by directly measured food availability, i.e. what consumers encounter in and around FOs (e.g. prices, promotions and nutritional quality) (Glanz et al., 2005). Linking these two types of proxy measures could provide a better theoretical understanding of where and how people access food.

self-reported measures due to cost and time. This inevitably causes heterogeneity across populations. 4.1. Car ownership The Danish National Travel Survey shows that 21% of the Danish population use a car for shopping purposes. The average distance between home and shopping is 10.6 km by car (Technical University of Denmark, 2013a). In agreement with the national travel trends, odds of frequent fast food intake for residents with long distances (> 4 km) was only significant among residents with a car, suggesting that when the nearest FFO is distant, the resident will simply drive there. This underlines that accessibility of FFOs is determined not only by their distribution across space but also by mobility factors such as car ownership or public transport networks. 4.2. Strengths Strengths of the present study include the large representative sample of residents in the Capital Region of Denmark with different SES characteristics and living within both urban and rural areas. Complete information on areas and individual-level characteristics was provided by good quality registers (Baadsgaard and Quitzau, 2011; Jensen and Rasmussen, 2011; Pedersen et al., 2006; Petersson et al., 2011). Information on fast food intake and the spatial location of FFOs was obtained at the same time, which provides a current picture of both exposure and outcome. Though outcome was based on subjective data, the question on fast food intake was based on a validated 48-item food frequency questionnaire (Toft et al., 2015). However, true associations may be weaker than those reported here, as we did not ask about fast food intake near home specifically. A majority of studies within the accessibility research use secondary FO data, but few verify and augment the presence and location. In the present study, information used to define the exposure was based on secondary source data which has previously been validated against ground-truthing by Toft et al. (2011) and found acceptable for the purpose of identifying and locating FFOs. Thus, accessibility measures were objective and defined from verified register-based measures of the built environments. Together this enhances the precision and credibility of the findings.

5. Conclusion This study presents results relevant for environmental influences on fast food intake in a Danish context, which differs in socioeconomic structures and urban form to the US and Australia. It is the first study based on validated FFO data to show that fast food accessibility measures are associated with weekly fast food intake in the Capital Region of Denmark. The association depended on both urbanicity and municipality SES. These results suggest that Danish health promotion strategies need to consider not only the contribution of the built environment to unhealthy eating but also the characteristics of groups and the urbanicity of the area in which they live. Future broad-based intervention and policy research is needed to support these findings. Ethics The research project was approved by the Danish Data Protection Agency according to the Danish Act on Processing of Personal Data. Approval from the Danish Health Research Ethics Committee System was not required according to Danish law, as the research project was purely based on data obtained from questionnaires and national registers. Written informed consent for publication based on the questionnaire data was provided by the participants when returning the questionnaires.

4.3. Limitations The present data is based on a cross-sectional design from which it is not possible to determine whether the associations observed are causal. Similarly, we are not able to determine the direction of the association, as it is quite difficult to account for self-selection (BooneHeinonen et al., 2011a). Thus, selective population movements may have imitated the impact of municipality SES and urbanicity we found on the association between fast food access and intake. A potential statistical limitation lies within the choice of using municipalities as level 2 units of the multilevel model. As mentioned, this was reasoned by the computation of weights from Statistics Denmark which were based on a stratified sampling design of the municipalities. A municipality within the Capital Region of Denmark comprises a rather large geographical area in which the individual demography and SES can vary widely. Consequently, parishes may have reflected the context more accurately. Nonetheless, preliminary supplementary analyses that did not account for the stratified sampling design and nonresponse showed similar associations between fast food access and intake as presented here. The same applied for supplementary unweighted analyses of a multilevel model comprising information on three levels: residents (level 1) nested within municipalities (level 2) nested within parishes (level 3). Thus, it seems that the chosen multilevel model was acceptable. Finally, a limitation lies within the FFO definition. As FFOs increasingly offer healthier fast food alternatives, the term “fast food” may no longer serve as a suitable proxy for

Declaration of conflicting interests The authors declare that there are no conflicts of interest. Acknowledgements The authors thank the team behind the survey at the Research Centre for Prevention and Health. Furthermore, we thank all the participants who took part in the survey. Funding This project was funded by the Capital Region of Denmark. Appendix A. - Municipality socioeconomic status groups description of a composite measure The socioeconomic status (SES) of the 38 municipalities was described through a composite measure combining the distribution of the educational level, employment status and mean gross income of the residents. Information was derived from central registers, i.e. the Danish Population's Education Register, the Employment Classification Module 108

Health & Place 48 (2017) 102–110

K.A. Bernsdorf et al.

Christensen, A.I., Ekholm, O., Glumer, C., Andreasen, A.H., Hvidberg, M.F., Kristensen, P.L., Larsen, F.B., Ortiz, B., Juel, K., 2012. The Danish National Health Survey 2010. Study design and respondent characteristics. Scand. J. Public Health 40, 391–397. http://dx.doi.org/10.1177/1403494812451412. Christensen, A.I., Ekholm, O., Gray, L., Glümer, C., Juel, K., 2015. What is wrong with non‐respondents? Alcohol‐, drug‐and smoking related mortality and morbidity in a 12‐year follow up study of respondents and non‐respondents in the Danish Health and Morbidity Survey. Addiction. Cobb, L.K., Appel, L.J., Franco, M., Jones‐Smith, J.C., Nur, A., Anderson, C.A.M., 2015. The relationship of the local food environment with obesity: a systematic review of methods, study quality, and results. Obesity 23, 1331–1344. Cummins, S., Curtis, S., Diez-Roux, A., 2007. Understanding and representing “place”in health research: a relational approach. Soc. Sci. Dunn, R.A., Sharkey, J.R., Horel, S., 2012. The effect of fast-food availability on fast-food consumption and obesity among rural residents: an analysis by race/ethnicity. Econ. Hum. Biol. 10, 1–13. Fraser, L.K., Edwards, K.L., Cade, J., Clarke, G.P., 2010. The geography of fast food outlets: a review. Int. J. Environ. Res. Public Health 7, 2290–2308. Gamba, R.J., Schuchter, J., Rutt, C., Seto, E.Y.W., 2014. Measuring the food environment and its effects on obesity in the United States: a systematic review of methods and results. J. Community Health, 1–12. Glanz, K., Handy, S.L., Henderson, K.E., Slater, S.J., Davis, E.L., Powell, L.M., 2016. Built environment assessment: multidisciplinary perspectives. SSM - Popul. Heal. 2, 24–31. http://dx.doi.org/10.1016/j.ssmph.2016.02.002. Glanz, K., Sallis, J.F., Saelens, B.E., Frank, L.D., 2005. Healthy nutrition environments: concepts and measures. Am. J. Heal. Promot. 19, 330–333. Groth, M.V., Sorensen, M.R., Matthiessen, J., Fagt, S., Landvad, N., Knudsen, V.K., 2014. Disparities in dietary habits and physical activity in Denmark and trends from 1995 to 2008. Scand. J. Public Health 42, 611–620. http://dx.doi.org/10.1177/ 1403494814547668. Hammer-Helmich, L., Buhelt, L.P., Andreasen, A.H., Magtengaard, K.R., HildingNørkjær, H., Glümer, C., 2011. The Danish Capital region Health Profile 2010 (Sundhedsprofil for region og kommuner 2010). Research Center for Prevention and Helath;The Capital Region of Denmark. He, M., Tucker, P., Irwin, J.D., Gilliland, J., Larsen, K., Hess, P., 2012. Obesogenic neighbourhoods: the impact of neighbourhood restaurants and convenience stores on adolescents' food consumption behaviours. Public Health Nutr. 15, 2331–2339. Jensen, V.M., Rasmussen, A.W., 2011. Danish education registers. Scand. J. Public Health 39, 91–94. http://dx.doi.org/10.1177/1403494810394715. Lachat, C., Nago, E., Verstraeten, R., Roberfroid, D., Van Camp, J., Kolsteren, P., 2012. Eating out of home and its association with dietary intake: a systematic review of the evidence. Obes. Rev. 13, 329–346. Lau, C.J., Lykke, M., Andreasen, A.H., Bekker-Jeppesen, M., Buhelt, L.P., Robinson, M.K., Glümer, C., 2015. The Danish Capital region Health Profile - Chronic disease 2013 (Sundhedsprofil 2013 - Kronisk sygdom). Forskningscenter for Forebyggelse og Sundhed, Region Hovedstaden, København. Laxy, M., Malecki, K.C., Givens, M.L., Walsh, M.C., Nieto, F.J., 2015. The association between neighborhood economic hardship, the retail food environment, fast food intake, and obesity: findings from the Survey of the Health of Wisconsin. BMC Public Health 15, 237. Lind, P.L., Jensen, P.V., Glümer, C., Toft, U., 2016. The association between accessibility of local convenience stores and unhealthy diet. Eur. J. Public Health, ckv242. Longacre, M.R., Drake, K.M., MacKenzie, T.A., Gibson, L., Owens, P., Titus, L.J., Beach, M.L., Dalton, M.A., 2012. Fast-food environments and family fast-food intake in nonmetropolitan areas. Am. J. Prev. Med. 42, 579–587. Lyseen, A.K., Hansen, H.S., Harder, H., Jensen, A.S., Mikkelsen, B.E., 2015. Defining neighbourhoods as a measure of exposure to the food environment. Int. J. Environ. Res. Public Health 12, 8504–8525. Macintyre, S., Ellaway, A., Cummins, S., 2002. Place effects on health: how can we conceptualise, operationalise and measure them? Soc. Sci. Med. 55, 125–139. McKinnon, R.A., Reedy, J., Morrissette, M.A., Lytle, L.A., Yaroch, A.L., 2009. Measures of the food environment: a compilation of the literature, 1990–2007. Am. J. Prev. Med. 36, S124–S133. Mendis, S., Banerjee, A., 2010. Cardiovascular disease: equity and social determinants. In: Blas, E., Kurup, A.S. (Eds.), Equity, Social Determinants and Public Health Programmes. World Health Organization. Merlo, J., Chaix, B., Yang, M., Lynch, J., Rastam, L., 2005. A brief conceptual tutorial of multilevel analysis in social epidemiology: linking the statistical concept of clustering to the idea of contextual phenomenon. J. Epidemiol. Community Health 59, 443–449, (doi:59/6/443)(pii). Ministry of Business and Growth Denmark, 2015. Proclamation of act on planning (Bekendtgørelse af lov om planlægning) [WWW Document]. URL 〈https://www. retsinformation.dk/forms/r0710.aspx?Id=176182#id7168b482-14bd-4765-a47bdc023e66d03b〉 (accessed 23 November 2015). Ministry of Environment and Food in Denmark, 2017. About the Danish Smiley Scheme [WWW Document]. URL 〈http://www.findsmiley.dk/english/Pages/About.aspx〉 (accessed 1 January 2017). Moore, L.V., Diez Roux, A.V., Nettleton, J.A., Jacobs, D.R., Franco, M., 2009. Fast-food consumption, diet quality, and neighborhood exposure to fast food: the multi-ethnic study of atherosclerosis. Am. J. Epidemiol. 170, 29–36. http://dx.doi.org/10.1093/ aje/kwp090. Mozaffarian, D., Afshin, A., Benowitz, N.L., Bittner, V., Daniels, S.R., Franch, H.A., Jacobs, D.R., Jr, Kraus, W.E., Kris-Etherton, P.M., Krummel, D.A., Popkin, B.M., Whitsel, L.P., Zakai, N.A., American Heart Association Council on Epidemiology and Prevention Physical Activity and Metabolism, Council on Clinical Cardiology, Council on Cardiovascular Disease in the Young, Council on the Kidney in Cardiovasc, C. on

and the Income Statistics Register (Jensen and Rasmussen, 2011; Petersson et al., 2011; Baadsgaard and Quitzau, 2011). This resulted in four municipality SES groups (Fig. 1). The composite measure was developed in 2007. The highest level of achieved education was identified for each participant and the prevalence of individuals with a short education (e.g. primary or secondary school) was calculated for each municipality, ranging from 14.3% to 39.6%. These percentages were used to group the municipalities into quartiles, giving 1 point to the lowest quartile (lowest prevalence) and 4 points to the highest quartile (highest prevalence). The same procedure was used to group the prevalence of unemployed individuals, ranging from 7.8% to 25.5%. The mean gross income in each municipality was calculated (from DKK 171,500 to DKK 364,900) and the municipalities were grouped into quartiles, giving 4 points to the lowest quartile (lowest income) and 1 point to the highest quartile (highest income). Points were summed into a total score taking values between 3 and 12. Based on a ranking of the total score, four municipality social groups were formed. In order to form groups of equal size (nine or ten municipalities), information on mean income was used to differentiate between municipalities with an equal total score. The municipalities in SES group 1 had 3–4 points and were all in the lowest quartile for unemployment and the highest quartile for income. The municipalities in SES group 2 had 5–8 points and most were in the second quartile for unemployment and in the second highest quartile for income. The municipalities in SES group 3 had 8– 10 points and were mostly in the third quartile for low education and unemployment. The municipalities in SES group 4 had 10–12 points; all were in the highest quartile for unemployment and most were in the highest quartile for low education and the lowest quartile for income. Hence, municipality SES group 1 includes the municipalities with the lowest prevalence of individuals with a low educational level, the lowest prevalence of unemployment and the highest mean income, i.e. the most affluent municipalities. Municipality SES group 4 includes the municipalities with the highest prevalence of individuals with a low educational level, the highest prevalence of unemployment and the lowest mean income, i.e. the less affluent municipalities. References Beaulac, J., Kristjansson, E., Cummins, S., 2009. A systematic review of food deserts, 1966–2007. Prev. Chronic Dis. 6, A105. Black, C., Moon, G., Baird, J., 2014a. Dietary inequalities: what is the evidence for the effect of the neighbourhood food environment? Health Place 27, 229–242. Black, C., Ntani, G., Inskip, H., Cooper, C., Cummins, S., Moon, G., Baird, J., 2014b. Measuring the healthfulness of food retail stores: variations by store type and neighbourhood deprivation. Int. J. Behav. Nutr. Phys. Act. 11, 69. http://dx.doi.org/ 10.1186/1479-5868-11-69. Boone-Heinonen, J., Gordon-Larsen, P., Guilkey, D.K., Jacobs, D.R., Popkin, B.M., 2011a. Environment and physical activity dynamics: the role of residential selfselection. Psychol. Sport Exerc. 12, 54–60. Boone-Heinonen, J., Gordon-Larsen, P., Kiefe, C.I., Shikany, J.M., Lewis, C.E., Popkin, B.M., 2011b. Fast food restaurants and food stores: longitudinal associations with diet in young to middle-aged adults: the CARDIA study. Arch. Intern. Med. 171, 1162–1170. Bowman, S.A., Vinyard, B.T., 2004. Fast food consumption of US adults: impact on energy and nutrient intakes and overweight status. J. Am. Coll. Nutr. 23, 163–168. Burgoine, T., Forouhi, N.G., Griffin, S.J., Brage, S., Wareham, N.J., Monsivais, P., 2016. Does neighborhood fast-food outlet exposure amplify inequalities in diet and obesity? A cross-sectional study. Am. J. Clin. Nutr., (doi:ajcn128132 )(pii). Baadsgaard, M., Quitzau, J., 2011. Danish registers on personal income and transfer payments. Scand. J. Public Health 39, 103–105. http://dx.doi.org/10.1177/ 1403494811405098. Caspi, C.E., Sorensen, G., Subramanian, S.V., Kawachi, I., 2012. The local food environment and diet: a systematic review. Health Place 18, 1172–1187. http:// dx.doi.org/10.1016/j.healthplace.2012.05.006. Charreire, H., Casey, R., Salze, P., Simon, C., Chaix, B., Banos, A., Badariotti, D., Weber, C., Oppert, J.-M., 2010. Measuring the food environment using geographical information systems: a methodological review. Public Health Nutr. 13, 1773–1785. Christensen, A.I., Davidsen, M., Ekholm, O., Pedersen, P.V., Juel, K., 2014a. The health of Danes - The National Health Profile 2013 (Danskernes Sundhed - Den Nationale Sundhedsprofil 2013). Sundhedsstyrelsen, Axel Heides Gade 1, 2300 København S. Christensen, A.I., Ekholm, O., Glumer, C., Juel, K., 2014b. Effect of survey mode on response patterns: comparison of face-to-face and self-administered modes in health surveys. Eur. J. Public Health 24, 327–332. http://dx.doi.org/10.1093/eurpub/ckt067.

109

Health & Place 48 (2017) 102–110

K.A. Bernsdorf et al.

2017. Beyond supermarkets: food outlet location selection in four U.S. cities over time. Am. J. Prev. Med. 52, 300–310. http://dx.doi.org/10.1016/ j.amepre.2016.08.042. Statistics Denmark, 2017. Documentation of statistics for Car Register and Publications 2017 [WWW Document]. URL 〈http://www.dst.dk/en/Statistik/dokumentation/ documentationofstatistics/car-register-and-publications〉 (accessed 3 July 2017). Statistics Denmark, 2005. Industrial Classification (NACE/DB03) [WWW Document]. URL 〈http://www.dst.dk/Site/Dst/Udgivelser/GetPubFile.aspx? Id=18503&sid=classuk〉 (accessed 3 January 2017). Sushil, Z., Vandevijvere, S., Exeter, D.J., Swinburn, B., 2017. Food swamps by area socioeconomic deprivation in New Zealand: a national study. Int. J. Public Health. http://dx.doi.org/10.1007/s00038-017-0983-4. Technical University of Denmark, 2013a. Travel survey - Fact sheet on transport by car in Denmark 2013 (Faktaark om cykeltrafik i Danmark 2013). Technical University of Denmark, 2013b. Travel survey - Fact sheet on transportation by foot in Denmark (Faktaark om transport til fods i Danmark 2013). Thornton, L.E., Bentley, R.J., Kavanagh, A.M., 2009. Fast food purchasing and access to fast food restaurants: a multilevel analysis of VicLANES. Int. J. Behav. Nutr. Phys. Act. 6, 28. http://dx.doi.org/10.1186/1479-5868-6-28. Thornton, L.E., Lamb, K.E., Ball, K., 2016. Fast food restaurant locations according to socioeconomic disadvantage, urban–regional locality, and schools within Victoria, Australia. SSM - Popul. Heal. 2, 1–9. http://dx.doi.org/10.1016/ j.ssmph.2015.12.001. Thornton, L.E., Pearce, J.R., Kavanagh, A.M., 2011. Using Geographic Information Systems (GIS) to assess the role of the built environment in influencing obesity: a glossary. Int. J. Behav. Nutr. Phys. Act. 8, 10.1186. Toft, U., Erbs-Maibing, P., Glumer, C., 2011. Identifying fast-food restaurants using a central register as a measure of the food environment. Scand. J. Public Health 39, 864–869. http://dx.doi.org/10.1177/1403494811423431. Toft, U., Kristoffersen, L.H., Ladelund, S., Lau, C., Ovesen, L., Borch-Johnsen, K., Jørgensen, T., 2015. Validation of the 48-item Food Frequency Questionnaire used in the Inter99 Study. Forskningscenter for Forebyggelse og Sundhed. Turrell, G., Giskes, K., 2008. Socioeconomic disadvantage and the purchase of takeaway food: a multilevel analysis. Appetite 51, 69–81. Voigtländer, S., Razum, O., Berger, U., 2013. Methods to measure neighbourhoods and analyse their impact on health: an overview, in: Neighbourhood Structure and Health Promotion. Springer, pp. 93–115. Wilkins, E.L., Morris, M.A., Radley, D., Griffiths, C., 2017. Using Geographic Information Systems to measure retail food environments: discussion of methodological considerations and a proposed reporting checklist (Geo-FERN). Health Place 44, 110–117. http://dx.doi.org/10.1016/j.healthplace.2017.01.008.

N, 2012. Population approaches to improve diet, physical activity, and smoking habits: a scientific statement from the American Heart Association. Circulation 126, 1514–1563, (doi:CIR.0b013e318260a20b )(pii). Norredam, M., Kastrup, M., Helweg-Larsen, K., 2011. Register-based studies on migration, ethnicity, and health. Scand. J. Public Health 39, 201–205. http:// dx.doi.org/10.1177/1403494810396561. Oexle, N., Barnes, T.L., Blake, C.E., Bell, B.A., Liese, A.D., 2015. Neighborhood fast food availability and fast food consumption. Appetite. Orfanos, P., Naska, A., Trichopoulou, A., Grioni, S., Boer, J.M.A., van Bakel, M.M.E., Ericson, U., Rohrmann, S., Boeing, H., Rodriguez, L., 2009. Eating out of home: energy, macro-and micronutrient intakes in 10 European countries. The European Prospective Investigation into Cancer and Nutrition. Eur. J. Clin. Nutr. 63, S239–S262. Pearce, J., Blakely, T., Witten, K., Bartie, P., 2007. Neighborhood deprivation and access to fast-food retailing. Am. J. Prev. Med. 32, 375–382. http://dx.doi.org/10.1016/ j.amepre.2007.01.009. Pedersen, C.B., Gøtzsche, H., Møller, J.Ø., Mortensen, P.B., 2006. The Danish civil registration system. Dan. Med. Bull. 53, 441–449. Pereira, M.A., Kartashov, A.I., Ebbeling, C.B., Van Horn, L., Slattery, M.L., Jacobs, D.R., Ludwig, D.S., 2005. Fast-food habits, weight gain, and insulin resistance (the CARDIA study): 15-year prospective analysis. Lancet 365, 36–42. Petersson, F., Baadsgaard, M., Thygesen, L.C., 2011. Danish registers on personal labour market affiliation. Scand. J. Public Health 39, 95–98. http://dx.doi.org/10.1177/ 1403494811408483. Powell, L.M., Chaloupka, F.J., Bao, Y., 2007. The availability of fast-food and full-service restaurants in the United States. Am. J. Prev. Med. 33, S240–S245. http:// dx.doi.org/10.1016/j.amepre.2007.07.005. Powell, L.M., Nguyen, B.T., Han, E., 2012. Energy intake from restaurants: demographics and socioeconomics, 2003–2008. Am. J. Prev. Med. 43, 498–504. Richardson, A.S., Boone-Heinonen, J., Popkin, B.M., Gordon-Larsen, P., 2011. Neighborhood fast food restaurants and fast food consumption: a national study. BMC Public Health 11, 543. http://dx.doi.org/10.1186/1471-2458-11-543. Richardson, A.S., Meyer, K.A., Howard, A.G., Boone-Heinonen, J., Popkin, B.M., Evenson, K.R., Shikany, J.M., Lewis, C.E., Gordon-Larsen, P., 2015. 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 36, 74–87. Robinson, M.K., Lykke, M.J., Hansen, H.H., Andreasen, A.H., Jeppesen, M., Buhelt, L.P., Lau, C.J., Glümer, C., Robinson, M.K., 2014. The Danish Capital region Health Profile 2013 (Sundhedsprofil for region og kommuner 2013). Forskningscenter for Forebyggelse og Sundhed. Rummo, P.E., Guilkey, D.K., Ng, S.W., Popkin, B.M., Evenson, K.R., Gordon-Larsen, P.,

110