A Systematic Review on Socioeconomic Differences in the Association between the Food Environment and Dietary Behaviors

Little is known about socioeconomic differences in the association between the food environment and dietary behavior. We systematically reviewed four databases for original studies conducted in adolescents and adults. Food environments were defined as all objective and perceived aspects of the physical and economic food environment outside the home. The 43 included studies were diverse in the measures used to define the food environment, socioeconomic position (SEP) and dietary behavior, as well as in their results. Based on studies investigating the economic (n = 6) and school food environment (n = 4), somewhat consistent evidence suggests that low SEP individuals are more responsive to changes in food prices and benefit more from healthy options in the school food environment. Evidence for different effects of availability of foods and objectively measured access, proximity and quality of food stores on dietary behavior across SEP groups was inconsistent. In conclusion, there was no clear evidence for socioeconomic differences in the association between food environments and dietary behavior, although a limited number of studies focusing on economic and school food environments generally observed stronger associations in low SEP populations. (Prospero registration: CRD42017073587)


Introduction
Socioeconomic inequalities in dietary behavior are persistent and widespread [1] and are contributing to inequalities in diet-related chronic diseases [2]. Several explanatory mechanisms for these inequalities have been proposed. Individuals with lower socioeconomic position (SEP) according to educational attainment, income levels or occupation status may lack the material and psychosocial resources that generally accompany a higher SEP. Indeed, material resources such as higher food budgets and access to health-promoting goods and services [3,4] and psychosocial resources such as nutrition knowledge, cooking skills and positive attitudes towards healthy eating [5][6][7][8] are known to contribute to healthier dietary behavior.
Having fewer material and psychosocial resources may limit individuals' capacity to resist unhealthy temptations in the food environment [9] or to take advantage of healthy options in the food environment. For example, higher educated individuals may be better able to deal with an unhealthy food environment because of their individual-level resources such as higher food budgets, better planning skills or more nutritional knowledge compared to those with lower education levels. If food environments are characterized by the availability and promotion of high-energy and ultra-processed foods-as is common in most Western countries [10]-food choices of those having fewer material and psychosocial resources are more likely to be unhealthy.
While there is some evidence that food environments are unhealthier in more deprived areas [11]-also referred to as the 'deprivation amplification' or the double burden of deprivation [12,13]-little is known about the differential effects of the food environment on dietary behavior in higher and lower SEP groups. Such socioeconomic inequalities in the effects of the food environment on dietary behavior could in fact provide an explanation for the weak or inconsistent associations described in the numerous systematic literature reviews summarizing the influence of the food environment on dietary behavior so far [14][15][16][17][18][19][20][21][22][23][24]. If individuals with a high or low SEP respond differently to their food environments, this may confound the association between aspects of the food environment and dietary behavior in studies that do not specifically consider the role of SEP. There is indeed some evidence that the food environment impacts dietary behavior differentially across socioeconomic strata. Three UK studies indicated that having a higher SEP is protective against exposure to unhealthy food environments [25][26][27]. However, evidence for this hypothesis has not been systematically reviewed. A better understanding of how food environments impact high and low SEP groups differentially would contribute to public health strategies targeting dietary inequalities.
Therefore, the aim of this study was to systematically review the evidence for socioeconomic differences in the association between the food environment and dietary behavior. in adolescents and adults. We included studies that stratified their population on the basis of SEP and studied the food environment-diet association in these strata. In addition, we included studies that investigated a single SEP group to assess if associations between the food environment and dietary behavior are generally stronger or more consistent in either high or low SEP populations.

Materials and Methods
We conducted a systematic literature review according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines [28]. The protocol for this literature search was registered in the Prospero database, registration number CRD42017073587 (can be found via https://www.crd.york.ac.uk/prospero/).

Literature Search
Original, peer-reviewed studies that examined associations between the food environment and dietary behavior in different socioeconomic strata or in a single SEP group were included. Food environments were defined as all objective and perceived aspects of the economic and physical food environment outside the home. Dietary behavior was defined as all measures of dietary behavior of foods and food groups, dietary patterns and food purchasing behavior. Socioeconomic groups were defined as individual, household or area-level measures of education, income, occupation or receiving benefits. Only studies with a study population of adolescents or adults (aged twelve years or over) were considered, as the food choices of children younger than twelve years are less likely to be directly influenced by the food environment (but rather via their parents' food choices). Furthermore, only studies with an observational study design (including baseline data of experimental studies) were included since we were interested in the differential response to long-term environment rather than the differential response to a (short term) change in the food environment as is the case in experiments. A detailed overview of the inclusion criteria is available in Table 1. A systematic literature search was performed in 4 electronic databases (Medline, Embase, Psycinfo and Web of Science) for studies published up to June 2018 in the English or Dutch language. The search strings can be found in Supplementary File S1. An additional manual search was performed to identify relevant articles based on the reference list of included studies. Papers identified by the search strategy were uploaded in Rayyan for screening. Rayyan is a free web and mobile app that facilitates multi-author screening of abstracts and titles [29]. To refine the inand exclusion criteria, the first 100 retrieved articles were screened on the basis of title and abstract. Inclusion rates were compared and if necessary, adjustments were made to the criteria. Thereafter, titles and abstracts were equally divided among five of the authors for screening of relevance according to the review inclusion criteria.
Full text versions of all records deemed eligible on the basis of title and abstract were searched through the four electronic data bases or alternatively searched via Google Scholar or requested by e-mail from the corresponding authors. The retrieved full texts were reviewed for inclusion.

Data Extraction
The following information was extracted from the included studies: • Study characteristics (author, year of publication, sample size, response rate, country, study design, objective); • Population characteristics (e.g., age group); • Type of dietary behavior (e.g., healthy eating index, adherence to dietary guidelines, fruit and vegetable (F & V) intake); • Aspect of food environment studied (e.g., distance to nearest supermarket); • Indicator of SEP (e.g., education, income, social class); • Results and conclusion.
Extracted data was summarized in tables based on type of food environment measure (e.g., perceived food environment, school food environment).

Assessment of Methodological Quality
All included studies were independently assessed for methodological quality using the 14-item NIH quality assessment tool for observational cohort and cross-sectional studies [30]. With regard to the item 'accuracy, objectivity, validity and reliability of the outcome measures', studies were rated positively when using a dietary assessment tool that was validated in the population under study or when using objective information on dietary purchases. Studies were rated neutrally when using a previously validated dietary assessment tool or using a combination of self-reported and objective dietary purchase outcomes. Studies were rated negatively when using a non-validated dietary assessment tool, or when a previously validated tool was adapted without further validation. Generally, articles were rated 'Good' when they had ≥6 times 'Yes', 'Fair' when they had 3-5 times 'Yes', and 'Poor' when they had 0-2 times 'Yes', but per instruction of the quality assessment tool, an assessment of the overall quality of the article was also included in the rating.

Results
After removal of duplicates, 18,838 articles were screened on the basis of title and abstract. A total of 18,132 records were excluded after reading title and/or abstract, leaving 706 articles for full-text screening. A further 668 articles were excluded on the basis of inclusion/exclusion criteria. Based on the reference lists of the thirty-eight studies included for data extraction, five additional papers were identified. Most exclusions were done because authors did not present food environment-diet associations by SEP, but only associations between the food environment and dietary behavior adjusted for SEP as covariate. A total of forty-three papers were included in the review, of which twenty-three studied the association between aspects of the food environment and dietary behavior across different SEP strata (Table 2) and twenty studied this association in a single SEP group ( Table 3). The study selection flowchart is presented in Figure 1. Generally, articles were rated 'Good' when they had ≥6 times 'Yes', 'Fair' when they had 3-5 times 'Yes', and 'Poor' when they had 0-2 times 'Yes', but per instruction of the quality assessment tool, an assessment of the overall quality of the article was also included in the rating.

Results
After removal of duplicates, 18,838 articles were screened on the basis of title and abstract. A total of 18,132 records were excluded after reading title and/or abstract, leaving 706 articles for fulltext screening. A further 668 articles were excluded on the basis of inclusion/exclusion criteria. Based on the reference lists of the thirty-eight studies included for data extraction, five additional papers were identified. Most exclusions were done because authors did not present food environment-diet associations by SEP, but only associations between the food environment and dietary behavior adjusted for SEP as covariate. A total of forty-three papers were included in the review, of which twenty-three studied the association between aspects of the food environment and dietary behavior across different SEP strata (Table 2) and twenty studied this association in a single SEP group ( Table  3). The study selection flowchart is presented in Figure 1.
Overall, studies that considered the association between economic aspects of the food environment and dietary behavior found differential associations on the basis of SEP. In five studies, objectively measured higher food prices of unhealthy foods were associated with either lower consumption of unhealthy foods or higher consumption of healthier foods, and higher prices of F & V were associated with lower consumption of F & V [31,33,44,46,61]. Four of these studies found that low SEP groups were more responsive to food prices [30,32,45,60]. One study did not find differential effects by SEP when linking fast food prices to fast food consumption but did observe that higher F & V prices were only associated with higher F & V consumption in a low SEP group [44]. The authors speculated that other, unmeasured, competing factors may have led to this unexpected finding [44]. Finally, one study found that a higher SEP group was more responsive to price promotions, most notably price promotions on healthier foods [45].
Studies that examined objectively measured access, proximity and quality of the food environment often did not find significant associations with dietary behavior, nor interactions by SEP [47,49,70,71]. Most studies focused on access and proximity of food retailers in the neighborhood. Three studies found associations between these aspects of the food environment and dietary behavior, but without any indication of moderation by SEP [50,52,71]. Four studies reported that associations between access, proximity and quality of the food environment were more strongly associated with dietary behavior in the socioeconomically disadvantaged subgroup compared to the higher socioeconomic groups [25,34,51,69], of which three studies focused on the neighborhood food environment and one on the in-store food environment. That is, these studies showed that less healthful in-store supermarket environments, poorer food environments, a higher proportion of convenience stores and shopping at supercenters or convenience stores were associated with a lower diet quality or unhealthier dietary behavior among those with low SEP, but this association was weaker, non-significant or in the opposite direction among those with high SEP. The authors suggested that fewer individual or neighborhood-level material and psychosocial resources make individuals with low SEP more vulnerable to availability and marketing of unhealthier foods [25,34,51,69]. Finally, one study observed that dietary inequalities between low and high income individuals were only present in neighborhoods with a low density of supermarkets and fresh produce markets [48] and one study reported that educational inequalities in fast food consumption were stronger in areas with higher fast food outlet exposure than in areas with lower fast food exposure [26]. Among low income adolescents: higher FF prices were associated with a higher number of days nonmeat protein consumption. Increased supermarket availability was associated with higher frequency vegetable intake. FF restaurant availability was not significantly associated with any of the food consumption patterns. Greater fast food outlet exposure was associated with greater fast food consumption. The difference in fast food consumption between those with lowest and highest education level was strongest in those most exposed to fast food outlets. Having supermarkets and supercenters nearby was associated with shopping in supermarkets and supercenters, respectively, but only in SNAP households. Only in non-SNAP households, having grocery stores nearby was associated with shopping there. Shopping at supercenters or convenience stores was associated with higher odds of purchasing SSB. Shopping at supercenters was associated with higher odds of purchasing water/low calorie beverages in both SNAP and non-SNAP households. Shopping at grocery stores was only associated with higher odds of purchasing SSB in SNAP households. Shopping at supermarkets was only associated with higher odds of purchasing water/low calorie beverages in SNAP households.

Individual-level income
A higher proportion of convenience stores relative to total food stores/restaurants was associated with lower diet quality scores and this association was stronger among low income participants. For specific food groups; only whole grain consumption was negatively associated with the % neighborhood convenience stores relative to total food stores/restaurants, and this association was also stronger among low income participants.       Shopping more frequently at convenience stores was associated with greater consumption of added sugars; buying food more often at neighborhood stores predicted significantly greater intake of SSBs and discretionary fats (e.g., butter); and buying food more often at supercenters was significantly associated with greater intake of discretionary fats. Conversely, shopping more often at specialty grocery stores and F & V stores was significantly associated with greater F & V consumption. All four studies examining socioeconomic differences in the association between the school food environment and dietary behavior showed interaction by SEP, although not all in the same direction. Two studies showed that low SEP adolescents benefitted more from healthy options in the school food environment than high SEP adolescents [63,64], one study showed that high SEP adolescents benefitted more from healthy options in the school food environment than low SEP adolescents [62] and one study showed that a fast food outlet or grocery store close to school was associated with irregular eating habits (described as an undesirable behavior) in low SEP adolescents only [65].
Finally, one study considered the perceived food environment and found that perceived availability of FF outlets, restaurants and convenience stores close to home was associated with unhealthy intakes, with larger effect sizes in adolescents from less affluent families than in adolescents from more affluent families [66].

Associations of the Food Environment and Dietary Behaviours in a Single SEP Group
All but one of the twenty studies that reported on the association between the food environment and dietary behavior in a single SEP group (Table 3) focused on a socioeconomically disadvantaged group in terms of receiving benefits, living in a deprived area, having low income, being low educated or having food insecurity status. The exception was the study by Leischner et al. which focused on university college students, thereby focusing on higher educated young adults [41]. Sixteen out of the twenty studies were conducted in the USA. Most of these twenty studies considered more than one aspect of the food environment: fourteen papers considered availability and quality of stores in the neighborhood [32,35,[38][39][40][41][42][43]53,56,57,59,67,68]; ten papers studied access, distance or time taken to travel to stores [32,36,37,40,43,[57][58][59][60]67,68]; and seven papers studied economic aspects of the food environment such as objective food cost and/or perceived affordability [32,43,[53][54][55][56]67].
In the studies conducted among a socioeconomically disadvantaged group that considered availability and quality of stores in the neighborhood [32,35,[38][39][40][41][42][43]53,56,57,59,67,68], five studies observed that perceived [39,40,56] and objective [36,41,42,57] availability of stores selling healthier products was associated with healthier dietary behavior and two studies observed that availability or use of stores selling unhealthier products was associated with unhealthier dietary behavior [42,60]. Six studies found no association between availability in food stores and dietary behavior [32,35,38,43,59,60]. One study showed that perceived food store access was not associated with F & V intake, while having both a supercenter and convenience store nearby was [35]. Another study showed that F & V and SSB consumption was higher in specific food shopping locations [53] but provided no explanation for this finding.
Of the six papers that studied the role of economic aspects of the food environment for dietary behavior [32,43,[53][54][55][56], two found no significant associations with objective food prices or perceived costs [32,56], and three found a negative association, such that higher objective food prices, higher perceived food costs and lower self-reported affordability were associated with lower diet quality or lower intake of healthy foods [43,54,55]. One study did not find an association between objective prices of F & V and F & V consumption but did find that higher SSB prices were associated with higher consumption of SSBs [53], which is an unexpected direction of the association. The authors suggested that this finding may be due to insufficient variation in SSB prices or misreporting of SSB consumption [53].

Quality Assessment
Of the forty-three included studies, twenty-six received a 'good' rating, fifteen received a 'fair' rating and two received a 'poor' rating (Table 4). Most studies scored poorly on the sample size justification and most studies did not use a validated tool to measure dietary behavior or used a previously validated tool but did not validate it in their study population. The two studies that received a 'poor' rating additionally did not describe their population clearly. 2009 N.B. '+' stands for a positive evaluation; '-' stands for a negative evaluation; '+/-' stands for a neutral evaluation; ' ' means the information was not provided/found in the article; NA = not applicable.

Discussion
The aim of this study was to systematically review the literature on socioeconomic differences in the association between the food environment and dietary behavior of adolescents and adults. We included studies that stratified their population on the basis of SEP as well as studies that considered the association between the food environment and dietary behavior in a single SEP group (e.g., only low-income groups). The included studies were diverse in their measures of the food environment and dietary behavior, indicators of SEP, and their findings.
We hypothesized that the food environment would have a stronger effect on dietary behavior in those with lower SEP, and that associations between the food environment and diet would be more consistent if only one socioeconomic group was considered. We found some evidence to support the first hypothesis: In the studies that focused on economic (n = 6) and school food (n = 4) environments, associations with dietary behavior tended to be stronger in the socioeconomically disadvantaged subgroups. However, this was not the case for studies focusing on objectively measured access, proximity and quality of the food environment (n = 12). Only one study focused on perceived food environments, therefore little can be concluded about the strength of evidence for socioeconomic differences in these types of studies. We did not find strong evidence for the second hypothesis since associations in specific socioeconomic groups (mostly in low SEP groups) were inconsistent, with about half of the studies finding non-significant associations. Studies among adolescents (n = 7) and non-cross-sectional studies (n = 4) generated most consistent results.
The more consistent evidence for the interaction by SEP for economic and school food environments may be due to the fact that these aspects of the food environment are more delimited and that 'exposure' to these aspects of the food environment is easier to define compared to aspects of availability and accessibility in the overall food environment. The significant amount of time ('exposure') adolescents spend at school may explain why this type of environment has a relatively consistent influence on dietary behavior. It may be speculated that adolescents with a high SEP have a healthier home food environment, while low SEP with unhealthier home food environments may therefore benefit more from a healthy school food environment [72]. The results for economic aspects of the food environment echo the findings from studies demonstrating a stronger response to tax and subsidy policies from those with lower SEP [73,74]. Future studies could examine the pathways through which these socioeconomic differences arise; we speculated that both material and psychosocial resources may play a role, but literature on these pathways is scarce [75].
In the studies considering a single SEP group, predominantly focused on socioeconomically disadvantaged populations, evidence for an association of the availability and quality of stores, access, distance or time taken to travel to stores, and (perceived) food costs with dietary behavior was inconsistent. About half of the studies found significant associations in the expected direction, a few found significant associations in an unexpected direction, and the remainder found no significant associations. This is comparable to the findings of systematic literature reviews on the association between the food environment and dietary behavior across socioeconomically diverse populations [14][15][16][17][18][19][20][21][22][23][24], providing little evidence that associations are more consistent when a more socioeconomically homogeneous population is considered. Many of the studies that focused on socioeconomically disadvantaged populations defined their population on the basis of community-level deprivation or income. This may leave room for socioeconomic variability within these communities, particularly if those with higher SEP were more likely to participate in the study. As such, the studies focusing on one specific SEP group may not truly have resulted in studies conducted in a socioeconomically homogeneous group. Additionally, on the basis of this literature review, little can be concluded about the role of the food environment for dietary behavior in a high SEP population, as we only identified one study that focused on such a population.
On the basis of previous literature reviews [14][15][16][17][18][19][20][21][22][23][24] we speculated that observed null associations in a socioeconomically diverse sample may be due to opposing associations in higher and lower SEP groups, but many studies did not find significant differences between SEP groups. It is likely that the inconsistencies observed in this literature review have similar causes as the inconsistencies observed in general literature reviews on associations between the availability and accessibility of the food environment and diet. Namely: That similar measures of the food environment are difficult to compare between different contexts; that food environments are often simplified to metrics of single types of food retailers (i.e., proximity to supermarkets, or availability of F & V in convenience stores), while the food environment encompasses a broad range of interacting factors (e.g., an interplay of proximity, availability, marketing, labelling, etc.); and that researchers make many assumptions about the places and ways in which food environments influence dietary behavior [20,22]. This may be reflected in our finding that SEP differences were most consistent for studies focusing on economic and school food environments, which represent much more narrow aspects of the food environment than access, availability and quality of food retailers. In general, adherence to reporting guidelines on food environment studies such as the Geo-FERN reporting checklist [76] would facilitate the comparison of such studies in systematic reviews.

Strengths and Limitations
This is the first systematic literature review that examined socioeconomic differences in the association between the food environment and dietary behavior. Strengths of this study were the broad definition of food environment variables in order to capture all relevant literature; the use of four search engines; the performance of a rigorous quality assessment of the included studies; and the fact that screening, data extraction and quality assessment was performed by at least two researchers each. However, although systematic literature reviews occupy a top position in the hierarchy of evidence, they, including this one, suffer from a number of limitations. Although we piloted the screening process, the involvement of multiple authors in the screening process and the high number of potentially relevant articles in general may have led to the erroneous exclusion of relevant articles. Furthermore, the heterogeneous nature of the included studies prevented us from performing a meta-analysis of the findings, and this hampers the assessment of publication bias: Authors may not have reported non-significant interaction terms with SEP, which may have led to an overestimation of the SEP-differences in this review. The classification of studies into categories of food environment measures may also be noted as a limitation: As studies in single SEP groups examine different aspects of the food environment than studies stratified by SEP we were unable to use the same classification for both types of studies. Finally, whilst there was no limitation for language during the search strategy, our review consists entirely of articles published in English. This could be due to the fact that other relevant articles may not have been indexed in the electronic databases used for this review.

Conclusions
Evidence for socioeconomic differences in association between the food environment and dietary behavior was inconsistent, although a limited amount of studies focusing on economic and school food environments generally observed stronger associations in low SEP populations than in high SEP populations. Studies on the association between food environment and dietary behavior in a single SEP group were no more consistent than studies in a mixed population observed in previous literature reviews. As such, it is unlikely that the inconsistencies in the association between the food environment and diet that have been observed thus far are attributable to a differential response to food environments from high and low SEP groups.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-6643/11/9/2215/s1, File S1: Search strings used for the systematic review on 'A systematic review on socioeconomic differences in the association between the food environment and dietary behaviors'.

Conflicts of Interest:
The authors declare no conflict of interest.