Measures of Perceived Neighborhood Food Environments and Dietary Habits: A Systematic Review of Methods and Associations

Access to healthy food is a necessity for all people. However, there is still a lack of reviews on the assessment of respondent-based measures of neighborhood food environments (perceived food environments). The aim of this systematic review was to evaluate the measurement tools for perceived food environments by five dimensions of food access and to obtain the overview of their associations with dietary habits among people aged 18 years and older in middle- and high-income countries. Observational studies using perceived food environment measures were identified through a systematic review based on two databases for original studies published from 2010 to 2020. A total of 19 final studies were extracted from totally 2926 studies. Pertaining to the five dimensions of food access, 12 studies dealt with accessibility, 13 with availability, 6 with affordability, 10 with acceptability, 2 with accommodation, and 8 with a combination of two or more dimensions. Perceived healthy food environments were positively associated with healthy dietary habits in 17 studies, but 8 of them indicated statistically insignificant associations. In conclusion, this review found accessibility and availability to be major dimensions of perceived food environments. The relationship between healthy food environments and healthy diets is presumably positive and weak.


Introduction
The United Nations sustainable development goals include Zero Hunger, a goal targeted at ending hunger, achieving food security, and improving nutrition [1]. Food environments are characterized by the availability, affordability, convenience, promotion, quality, and sustainability of foods and beverages in wild, cultivated, and built spaces [2]. Healthy food environments are essential to ensure food security, such that all citizens can sustainably access healthy food [3]. Empirical evidence of the health impact of neighborhood food environments has accumulated especially in deprived areas in high-income countries since the early 1990s [4]. A review [5] has reported inequalities in food access in the United States, and there has been a paucity of studies in other developed countries.
Proper measurement of food environments is required to investigate the relationship between food environments and dietary habits. Objective measures of food environments, Nutrients 2022, 14, 1788 2 of 33 such as geographic information systems, are direct observations and are a common methodology for assessing food environments [6]. However, these objective measures may not necessarily capture individual behaviors or the actual situation of food access [7]. For example, studies conducted in the United States have demonstrated that consumers traveled beyond their nearest supermarket to obtain cheaper [8] and healthier food [9], indicating that physical distance may not be the only factor involved in choosing primary food stores. Perceived (respondent-based) measures, such as individual perceptions and experiences may support the limitations of objective measures for assessing food environments.
Nevertheless, perceived measures face some challenges. First, there are no standardized measures of perceived food environments. One of the processes for developing a standardized measurement is to classify them based on the different aspects of food environments. Penchansky and Thomas [10] proposed the utilization of the five dimensions of "food access" (accessibility, availability, affordability, acceptability, and accommodation). Glanz et al. [11] suggested that community and consumer environments impact individual behaviors. Accordingly, accessibility, availability, and accommodation can be included in the community environment, and affordability and acceptability can be grouped in the consumer environment. Second, there is still a lack of evidence on the relationship between perceived food environments and dietary habits. Only one review [12] classified the perceived and objective measures, and indicated that perceived measures of availability within the neighborhood food environments were consistently associated with healthy diets among studies published through 2011. However, a review [12] targeted both children and adults, wherein it was reported that food environments for children may be influenced by the household main shopper. A review targeting adults who are likely to be the main shoppers is required.
The aims of this study were to systematically review existing tools used to measure perceived food environments according to the five dimensions of "food access", and to assess the association of perceived food environment measures with dietary habits.

Search Strategy
The review protocol was registered in the public domain (PROSPERO registration number: CRD42020201881) in accordance with the PRISMA guidelines [13]. The present systematic search was conducted to identify studies with observational designs published online in English, between 1 January 2010 and 6 August 2020, using the PubMed and Web of Science databases (Figure 1). Systematic keyword searches were developed and agreed upon by all authors to identify studies that investigated the relationship between perceived food environments and dietary habits among community-dwelling people aged 18 years and older in middle-and high-income countries, with at least 200 people [14] (Supplementary Materials Tables S1 and S2). The two databases were explored by one reviewer on 6 August 2020, to unify the run time, and duplicates were excluded. Basic data cleaning was performed before the first screening. The articles were screened to identify those that failed to be eliminated in the keyword search: off-topic studies and those that investigated food environments in low-and lower-middle-income countries [15]. We defined studies as off-topic when relevant text words, such as "food environments", "dietary habits", and "food access" were not included in the title or the abstract. In the first screening, we excluded studies that stated the following in the title and abstract: (1) duplication, (2) the criteria of the number of population, (3) studies without an observational design, (4) studies that did not use perceived measurements for assessing neighborhood food environments, and (5) studies that did not investigate dietary habits as an outcome. The second screening was conducted by investigating the full text using the same exclusion criteria as the first screening. The basic data cleaning and first-and second-screenings of studies were independently conducted, blinded, and then jointly reviewed by three reviewers using the free web application of Rayyan [16]. The three reviewers assessed the risk of bias individually, although not blinded, and decisions were confirmed by four other reviewers. ond screening was conducted by investigating the full text using the same exclusion criteria as the first screening. The basic data cleaning and first-and second-screenings of studies were independently conducted, blinded, and then jointly reviewed by three reviewers using the free web application of Rayyan [16]. The three reviewers assessed the risk of bias individually, although not blinded, and decisions were confirmed by four other reviewers.

Analyses
We summarized the studies by the number of study participants, location (i.e., urban and rural), country, project data source, study design, and target population. Furthermore, we described the measurement tools and types of analyses (i.e., continuous and dichotomous) of perceived food environments and dietary habits, respectively. Measures of perceived food environments were classified according to the five dimensions of food access (accessibility, availability, affordability, acceptability, and accommodation) [10]. The definition of the five dimensions are as follows [10,12]: accessibility, the location of the food supply source and the ease of getting to that location, accounting for travel time and distance; availability, the adequacy of the supply of healthy food; affordability, food prices and people's perceptions of worth relative to the cost, which is often measured by store audits of specific foods, or regional price indices; acceptability, people's attitudes about attributes

Analyses
We summarized the studies by the number of study participants, location (i.e., urban and rural), country, project data source, study design, and target population. Furthermore, we described the measurement tools and types of analyses (i.e., continuous and dichotomous) of perceived food environments and dietary habits, respectively. Measures of perceived food environments were classified according to the five dimensions of food access (accessibility, availability, affordability, acceptability, and accommodation) [10]. The definition of the five dimensions are as follows [10,12]: accessibility, the location of the food supply source and the ease of getting to that location, accounting for travel time and distance; availability, the adequacy of the supply of healthy food; affordability, food prices and people's perceptions of worth relative to the cost, which is often measured by store audits of specific foods, or regional price indices; acceptability, people's attitudes about attributes of their local food environment, and whether the given supply of products meets their personal standards; and accommodation, how well local food sources accept and adapt to the needs of local residents (e.g., store hours and types of payment accepted). Dietary outcomes were classified by measures of fruit and/or vegetable intake, other healthy food intake, unhealthy food intake (i.e., fast food), and a diet-quality index that the selected studies employed. We identified healthy and unhealthy foods according to the definitions of the selected studies. The risk of bias was assessed across seven domains, each of which consisted of three to nine indicators based on two scales: the Risk of Bias for Nutrition Observational Studies Tool [17,18], which is applicable to observational studies in public health nutrition, and the Newcastle Ottawa Scale [19] which indicates potential biases of food environment studies (Supplementary Materials Table S3). The seven domains were (1) confounding, (2) selection of participants, (3) classification of exposures, (4) departures from intended exposures, (5) missing data, (6) measurement of outcomes, and (7) selection of reported results.
Finally, we summarized the associations of perceived food environments with dietary habits in the final model of the analysis. The association was defined as "positive" when healthy perceived food environments were significantly associated with a higher intake of healthy food or a lower intake of unhealthy food; and was defined as "negative" when healthy perceived food environments were significantly associated with a lower intake of healthy food or a higher intake of unhealthy food. To consider not only the statistical significance but also the trend of the association [20], we assessed the trend even if the association was statistically insignificant. We counted the studies on the use of the five dimensions of food access among the selected studies. We counted each dimension when two or more dimensions were reported in one study. In addition, we counted studies that indicated positive or negative associations in the indicator of the dimension. When two or more indicators were indicated in one dimension, we counted them as one statistically significant association. For example, we counted one significant association from six types of indicators in one dimension of affordability in one study [21]. Statistical significance was set at a two-sided p-value of 0.05.
We decided not to conduct a meta-analysis because of the heterogeneity in exposure and outcome measurements across the studies.

Study Overview
Among the 2926 studies identified by the two databases, 2519 were excluded during the basic data cleaning (Figure 1). Of the remaining 407 studies, we excluded 343 that met the exclusion criteria in the first screening. After a full article review during the second screening, 41 of the 64 studies were excluded. Of the 23 studies remaining, we excluded an additional four studies at the risk of bias assessment. Of these four, one study [22] was excluded because it targeted a specific population that received healthcare services, and another [23] did not use a perceived measurement tool for food environments; two sets of studies used the same measurement tools from the same research project: Lucan et al. [24] and Lucan and Mitra [25]; Bivoltsis et al. [26] and Trapp et al. [27]. We selected Lucan and Mitra [25] and Bivoltsis et al. [26] targeting a wider study areas and larger population.
Among the indicators in the 19 studies, we omitted the indicator of home food environment in the studies of Alber et al. [28], Kegler et al. [29], and Springvloet et al. [30], because the current review did not focus on household food environments. However, we did not omit the indicator of the accessibility of unhealthy food at the workplace, as investigated by Carbonneau et al. [31], because the indicators of neighborhood and workplace were integrated into one score.

The Assessment of the Risk of Bias
There was no serious or critical risk of bias in the 19 studies, and most of them had a moderate or low risk of bias against the seven classified domains (Supplementary Materials  Table S4). All studies were determined to have a moderate risk of bias in confounding, selection of participants, and departure from intended exposures. The moderate risk of bias against the classification of exposures (i.e., perceived food environments) was identified in three studies; one study [32] did not mention the validity or reliability of the perceived measurement tools, and two studies [30,33] used the indicators of perceived food environments that were employed in previous reports but did not investigate validity and reliability. Four studies that were judged as having a low risk of bias of missing data performed multiple imputation [31,34], engaged in a listwise deletion of data by not observing a missing data pattern [35], and did not exclude missing data amounting to 1.5% of the total [25]. Five studies [7,28,29,36,37] did not describe missing data. Seven studies [7,25,35,[38][39][40][41] with a low risk of bias in measurement outcomes (i.e., dietary habits) used data from interviews conducted by trained interviewers and not through self-description of participants.

Characteristics of the Study Design
Thirteen studies were conducted in the United States, and five studies [21,26,30,31,37] were conducted in Western countries and Australia in the Oceania region (Table 1). Only one study [33] has been conducted in Japan in East Asian countries. The study areas of 10 studies [7,21,26,28,30,31,35,37,38,42] were urban areas, while six studies [25,33,[39][40][41] were conducted in both urban and rural areas. With respect to the study design, 18 studies were cross-sectional, and only one study [26] used a longitudinal study design from baseline 2003-2005 to 2004-2006 to investigate changes in food environments and dietary habits. Two studies included minority populations, such as French-speaking adults [31] and African American (with White) adults [29]. Three studies [7,35,42] targeted adults with low income. Furthermore, Lo et al. [34] targeted middle-aged and older women, and Sharkey et al. [32], and Yamaguchi et al. [33] targeted older adults.

Overview of the Measurement Tools of Perceived Food Environments
The frequency of usage of dimensions of food access were 12 studies in accessibility, 13 studies in availability, six studies in affordability, 10 studies in acceptability, and 2 studies in accommodation (Table 2). Studies have integrated two [29,35,36,39], three [32,34,37], and five [31] dimensions to form one score. Chapman et al. [21] used one dimension of affordability with three indicators, and five studies [7,26,33,41,42] used one dimension with a single-item indicator; three studies [7,26,33] used accessibility, one study used availability [41], and the other used acceptability [42].
A total of 17 studies used measurements of perceived food environments that were previously validated or pilot tested. Two studies [30,33] used measurements that were previously used but were not validated, and one study [32] did not validate the measurement. Indices of perceived food environments in five studies [25,28,35,38,40] exhibited moderate validity using objective measurements as a standard. Eight studies indicated that perceived measurements demonstrated a moderate level of internal consistency, as analyzed by test-retest reliability [7,28,29,34,38,39,41] and inter-item reliability [37].
Affordability investigated the perception of worth relative to the cost of all food items [32] or the specific foods, such as fruits and vegetables [21,28,30,32,38], and healthy foods [31].
The measurement of the Nutrition Environment Measures Survey-Perceived [49] Perceived store consumer nutrition environment    Response: 5-point agree/disagree Likert scale (the response of (e) was reverse-coded, and the higher score reflects healthy food environments) Perceptions of neighborhood barriers [62] Perceived neighborhood nutrition barriers: 5 items (a) Too many fast-food restaurants (b) Not enough food stores with affordable fruits and vegetables (c) Not enough restaurants with healthy food choices (d) Not enough farmer's markets or fruit stands (e) No place to buy a quick, healthy breakfast to go Response: 5-point Likert scale, not a problem to a big problem, was summed as "perceived neighborhood nutrition barriers" score (ranged 5 to 2; the lower score reflects healthy food environments).  The produce in my neighborhood is more expensive than that in other neighborhoods * (j)

Continuous
The low-fat products in my neighborhood are more expensive than those in other areas * (k) The fresh produce in my neighborhood is of high quality (l) The low-fat products in my neighborhood are of high quality Response: 4-point agree/disagree Likert scale (the higher summed score reflects healthy food environments). * An item was reverse-scored. The access-related score was integrated accessibility, availability, and acceptability. The score of affordability was used itself as the Food affordability.  Dichotomous F&V: fruits and vegetables. a Applicable types of perceived food environments were selected from the five types provided below. If there were two or more types, the applicable types were described (i.e., a−l) in the measurement column. Accessibility: The location of the food supply source and the ease of getting to that location, counting for travel time and distance. Availability: The adequacy of the supply of healthy food; examples in the food environment might include the presence of certain types of restaurants near people's homes, or the number of places to buy produce. Affordability: Food prices and people's perceptions of worth relative to cost, which is often measured by store audits of specific foods, or regional price indices. Acceptability: People's attitudes about attributes of their local food environment and whether the given supply of products meets their personal standards. Accommodation: How well local food sources accept and adapt to the needs of local residents (e.g., store hours and types of payment accepted).

Continuous
Indicators for assessing the influence of media on food and nutrition on one's diet [31] and neighborhood social cohesion [29] were investigated by accommodation. Five studies investigated unhealthy food environments: availability of fast-food restaurants [41], accessibility of cafés or restaurants [26], and accessibility of fast-food restaurants in the score [31,36,37].

The Outcomes of the Dietary Habits
The intake of fruits and vegetables or only vegetables [30] was the most common method of measuring dietary habits. Two studies [25,41] investigated the frequency (times/week) of fast-food intake, and one study [29] investigated fat intake ( Table 3). The intake of fruits and vegetables (servings, times, and grams per day) was calculated using the validated food frequency questionnaires [7,29,30,34,38,39,42] and the measurement tools that were previously used [25,33,40]. Score indices of diet quality were employed in four studies [26,31,36,37]. Bivoltsis et al. [26] used an unhealthy dietary score.

Overview of the Associations
Nine studies indicated significant positive associations of perceived food environments with healthy dietary habits within the dimensions of accessibility [7,33], affordability [21], acceptability [28,42], and mixed scores of accessibility, availability, acceptability, and/or affordability [32,34,36,37,39] (Tables 4 and 5). Eight studies reported positive, but not statistically significant, associations [25,26,29,31,34,[37][38][39]. Four studies showed significant negative associations of availability [28,30], affordability [30], and the mixed score of availability and acceptability [35] with healthy diets. Five studies [25,26,28,41] investigated the association of perceived food environments with unhealthy diets. Bivoltsis et al. [26] indicated a significant positive association of improved accessibility of healthy food environments with a high intake of unhealthy food. Bivoltsis et al. [26] also found that changing low accessibility to unhealthy food environments was significantly and positively associated with high intake of unhealthy food. A study conducted by Lucan et al. [25] showed that poor accessibility of fruits and vegetables and supermarkets and poor acceptability of grocery quality were significantly and positively associated with higher fast-food intake.
With  [29,35,39,40] investigated the pathways and mediations of perceived food environments in relation to dietary habits. Three studies [29,39,40] did not control for any possible confounders in the path model to prevent over-specification of the results [90]. Springvloet et al. [30] analyzed perceived food environments based on availability and affordability as mediators of the association between education level and vegetable intake using a linear regression model. Bivoltsis et al. [26] investigated the association of the change (improved and worsened) in perceived food environments with changes in the dietary habits of people after one to two years of changing residence in a longitudinal study.  F&V intake F&V intake Fruit and vegetable intakes were separately measured by self-reported two-item screener [86,87].
(1) The number of servings of fruit (1/2 cup of fruit or 3/4 cups fruit juice) usually consumed each day (2) The number of servings of vegetables (1/2 cup cooked or 1 cup raw) consumed daily.
Total fruit and vegetable intakes were calculated by combining (1) and (2).

Continuous
Springvloet et al., 2014 [30] Vegetable intake Vegetable intake Food frequency questionnaire [88,89] Four items using a reference period of one month (g/day) (1) How many days per week they usually consume cooked and raw vegetables or salads (ranging from 0 to 7 days per week)? (2) How many tablespoons of cooked and raw vegetables or salads they usually ate on these days (ranging from one to six or more)?

Continuous
Yamaguchi et al., 2019 [33] F&V intake Meat and fish intake F&V intake and Meat and fish intake Average intake of vegetables/fruits and meat/fish over a one-month (times/day) [65] was calculated by the response of 'every day and over twice/day, every day and once/day, 4-6 times/week, 2-3 times/week, once-a-week, less than once-a-week, or almost never'.

Continuous
F&V: fruits and vegetables. a Diet quality was assessed by scores based on an indicator.        : no significance (the direction of the association): no information. Statistically significant associations: * p < 0.05, ** p < 0.01, and *** p < 0.001. a A "positive" association existed when that healthy perceived food environments were significantly associated with a higher intake of healthy food or a lower intake of unhealthy food, and "negative" association, when healthy perceived food environments were significantly associated with a lower intake of healthy food or a higher intake of unhealthy food. The "positive" association indicated instances when unhealthy perceived food environments were significantly associated with a lower intake of healthy food or a higher intake of unhealthy food, and "negative" indicated instances when unhealthy perceived food environments were significantly associated with a higher intake of healthy food or a lower intake of unhealthy food.

Discussion
This is the first systematic review to assess the measures of the perceived food environments and their associations with dietary behaviors in middle-and high-income countries in 19 studies. Accessibility and availability were the most commonly measured dimensions of food access. A positive relationship between healthy perceived food environments and healthy dietary habits was observed among 17 studies, with nine of studies having a statistically significant relationship.

Characteristics of the Study Design
The reviewed studies mostly investigated food environments in the United States and other Western countries. Global changes in the food system associated with global economic growth have increased availability of unhealthy food [91] and consequently transformed dietary habits. Therefore, more evidence from different regions of non-Western countries, such as Asian countries, is required. No significant difference in the association of subjective food environments and dietary intake between urban and rural areas was observed in this review. However, studies in rural areas [29,32,34,36] considered the physical distance and/or number of healthy food stores in neighborhoods. This is because rural-urban inequality, such as infrastructure challenges and low population density, was in existence [92]. Therefore, specific strategies for rural communities are required.
In accordance with the present results, significant associations were observed in studies that targeted specific populations. For example, studies that targeted socially vulnerable people, such as those with low incomes [7,35,42] and older adults [32,33], indicated a significant association between perceived food environments and dietary habits. One review [93] proposed the stigma and food inequity conceptual framework which is composed of the structural (e.g., neighborhood infrastructure and targeted marketing) and individual (e.g., awareness and endorsement of negative beliefs, thoughts, and beliefs) levels. These stigmas are associated with food inequities due to access to resources, home food environments, and psychosocial and behavioral processes, which ultimately undermine healthy dietary intake and contribute to food insecurity [93]. To understand the food environments among vulnerable people, it is necessary to consider the contexts of poverty, race, nationality, gender, age, malnutrition, and their intersection.

The Assessment of the Risk of Bias
The moderate level of bias in confounding, selection of participants, intended exposures, and selection of results may be reasonable in the present review because the articles were observational studies that had limitations in the relevant confounder adjustment, eligible participant selection, and precise exposure setting compared to a well-designed randomized trial. We did not investigate the statistical power as heterogeneity in the exposure measurement and outcomes made comparing the effect size difficult, although we selected studies that targeted at least 200 people. A review observed that evidence depended on not only the statistical power but also the research methodology [94]. Therefore, this review assessed the risk of bias (i.e., study quality) comprehensively.
Regarding the bias of missing data, statistical approaches are expected to be considered for missing data in accordance with the missing patterns [95]. Four studies [25,31,34,35] considered proper imputation approaches in accordance with data missing completely at random, missing at random, and not missing at random [95]. A description of the statistical approaches for missing data is important to assess measurement bias. With respect to bias in dietary measurements, there were seven studies [7,25,35,[38][39][40][41] with a low risk of bias as interviews were conducted by trained staff, which would reduce measurement error.
In the statistical model, all studies in this review considered confounders, such as age, sex, ethnicity, income, and/or other social determinants of health. However, only Bivoltsis et al. [26] considered the duration of residence in an area after relocation. The year of residence can possibly affect the geographic knowledge of the location of neighborhood stores and also impact the cultural acceptability of foods for people moving to a new neighborhood. Therefore, the duration of residence should be considered in food environment research.

Overview of Measurement Tools of Perceived Food Environments
Five studies [25,28,35,38,40] examined the validity of perceived food environments using objective measures as a standard. However, it is unclear whether objective (i.e., geographic) measurements accurately reflect the location of neighborhood primary food stores [8,9]. In addition, objective measurements are yet to be standardized using a consistent measure [12]. Nevertheless, using both objective and perceived measures is necessary to capture the complexity of food environments using different measurement tools.
According to the present review, accessibility of food stores within a walkable distance or convenient time, availability of a variety of fresh fruits and vegetables in the neighborhood could be some of the basic indicators to measure perceived food environments. In addition, affordability of prices of fruits and vegetables and acceptability of the quality of fruits and vegetables are necessary to consider the gap between individual perceptions and neighborhood retail. The indicators using accessibility, availability, affordability, and acceptability proposed by previous studies [43][44][45][46][47][48] could be optimized for structuring a food access measure, given that these indicators were employed in the present eight studies. These dimensions are useful and helpful from the viewpoint of public health to understand the measurement of perceived food environments the studies used.
However, the definition of dimensions have to be clearer since one dimension could overlap with another dimension. For example, certain studies were found to name only one dimension even when other dimensions were involved [29,31,35,37]. Most studies in this review did not clearly specify the dimensions. Especially in the definition of accommodation, convenience of store hours and types of payments are likely to be classified as accessibility, availability, and acceptability. The difficulty in the classification may limit the utility of these dimensions.

Overview of the Association of Perceived Food Environments with Dietary Habits
According to the present review, healthy perceived food environments are positively associated with healthy dietary habits but the association is weak. One study indicated that the individual-level factors accounted for the largest variation in fruit and vegetable intake as compared to that at the area level [96]. Nevertheless, health behaviors interact with physical and social environments, including food environments [97]. Therefore, interventions for both individual dietary behaviors and food environments may be important.
From the present results of the inverse relationship between healthy perceived food environments and unhealthy dietary habits, it is possible that people, especially those with low incomes [35], do not necessarily make healthy choices when both healthy and unhealthy foods are accessible, available, acceptable, and affordable. Indeed, the present review observed a significantly higher fast-food intake in healthy perceived food environments despite good accessibility of supermarkets/greengrocers [26]. Lucan and Mitra [25] indicated that poor accessibility of supermarkets, poor availability of produce, and poor acceptability of grocery quality were significantly associated with high intake of fast food. The present results imply that even when food environments are subjectively perceived as healthy, they may still have several choices of unhealthy food, given high availability of both healthy and unhealthy foods sold in stores according to consumer demands. A systematic review [98] investigated serving size labeling which is important to ensure an accurate understanding of nutritional content and food choices and consumption, among 14 articles (12 articles were from North America) and indicated that consumers in several studies had a poor understanding of serving size labeling. Another possibility would be suggested for why some people do not necessarily make healthy choices in good food environments. People who live in areas that are not familiar with healthy food may experience certain barriers in making healthy choices. One study reported that people who were introduced to the Mediterranean diet, which is considered a healthy diet, found a difficulty in purchasing food items due to an increase in food costs and found work, stress, and time pressures undermined adherence to the diet [99]. Therefore, nutrition education, as well as the improvement of food marketing, are required at the policy level to combat unhealthy food consumption so that people can effectively make healthy choices in complex food environments [100].
To determine food environments, a comprehensive assessment using both objective and subjective measures at structural and individual levels is required [93]. To terminate food inequity, it is necessary that policymakers collaborate with communities and private companies to devise a healthy city plan that includes measures, such as zoning the location of grocery/convenience stores. Simultaneously, further studies on the perceived food environments and literacy of healthy diets are required to monitor and assess the policy intervention.

Limitations
The study selection was conducted by three reviewers independently in the first and second screenings of the study records, keeping each decision blinded. However, the present review had some limitations that warrant mention. First, the present classification of perceived food environments according to the five dimensions of food access was inconclusive. The classification of the dimensions of food access on perceived food environments decided by the reviewers may differ from those of other reviewers. Second, there was a possibility of publication bias in the present review [101]. However, publication bias could be minimized by conducting a systematic review [13]; as a result, we extracted representative articles demonstrating evidence-based measurements and outcomes. Third, selection bias was not completely excluded, although it was minimized by blinding. Fourth, the causality of the relationship between perceived food environments and dietary habits is still unclear because all studies in this review were cross-sectional, except for one study [26]. Additionally, this review did not conduct the meta-analysis. A meta-analysis using longitudinal studies is needed to capture the causal relationship between perceived food environments and dietary habits, as perceived food environments change over time due to the turnover of food stores, constructed roads, and individual situations.

Conclusions
The order of frequent use of the perceived food environments was availability, accessibility, acceptability, affordability, and accommodation. Positive association of perceived food environments with dietary habits was observed, although this association may be weak. The characteristics of the relationship between perceived food environments and dietary habits are complex due to socioeconomic and latent background characteristics at the individual and community levels. Therefore, it is necessary to measure multiple aspects, such as the combination of food access dimensions of perceived food environments and to consider the effect of both healthy and unhealthy food in food environments and dietary habits.
Supplementary Materials: The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/nu14091788/s1, Table S1: Keyword search conducted on Pub Med; Table S2: Keyword search conducted on the Web of Science; Table S3: Assessment sheet of the risk of bias; and Table S4: Results of the risk of bias assessment.