The study used data from the 2006-2007 Brazos Valley Food Environment Project (BVFEP), which was approved by the Institutional Review Board at Texas A&M University, and the decennial 2000 U.S. Census Summary File 3 (SF-3) for six rural counties in the Central Texas Brazos Valley region (see Figure 2). The BVFEP has a history of working with community partners in the seven-county Brazos Valley region of Texas (one urban and six rural counties); all six rural counties were included in this study. These counties, which consist of 101 Census block groups (CBGs) and include five urban clusters (population >2,500), are considered rural based on population density [80,81]. The rural region covered a land area of 4,466 m² and included a population of 119,654 people [82]. According to the 2000 U.S. Census at the area-level of CBG, the median proportion of minority resident was 24.6% (range 3.6%-89.8%); unemployment was 2.4% (range 0%-8.8%); median of 37.9% of households (range 0%-72.8%) reported an income under 200% federal poverty level (FPL); a median of 14.2% (range 0%-39.8%) completed less than nine years of education; 49.4% (range 0%-71.8%) of housing owner-occupied; median income of $32,269 (range $8235-$51,776), and 37% (range 0%-74.8%) travel at least 30 minutes to work [83]. More than 41% of the 101 CBG are considered low-income areas; that is, areas in which at least 40% of residents has income at or below 200% FPL [17]. Regular public transportation services, such as fixed route, commuter, or taxi services, were not available in the study area [84,85].

The CBG, which is the smallest unit of Census geography for which the detailed "long-form" social and economic data from the Census are tabulated, was selected to define a neighborhood [36,86]. Neighborhood social and material deprivation. We applied the Neighborhood Socioeconomic Deprivation Index to each of the 101 CBG in the rural study area [36]. This measure of compound social and material deprivation was calculated from the 2000 U.S. Census Summary File 3 (SF-3) and included unemployment (persons age 16 years and older in the labor force who were unemployed and actively seeking work), poverty (persons with incomes below the federal poverty level), low education attainment (persons age 25 years and older, with less than a 10^th-grade education), household crowding (occupied households with more than one person per room), public assistance (households receiving public assistance), vehicle availability (occupied housing with no vehicle available), and telephone service (occupied housing with no telephone service). Based on the distribution of scores for the index, a three-category variable for overall neighborhood socioeconomic deprivation was constructed and assigned to each CBG: low deprivation (highest overall socioeconomics and lowest quartile of deprivation scores), middle deprivation (middle two quartiles), and high deprivation (lowest overall socioeconomics and highest quartile of deprivation scores) [36]. Vehicle ownership. Data from SF-3 were used to determine CBG-level vehicle ownership (occupied housing with vehicle available). Tertiles were used to construct a three-category variable for vehicle ownership (range of 62% to 100%); <90.5% of households with a vehicle was considered low ownership, 90.5% to 95.4% medium, and >95.4% high vehicle ownership.

In the 2006 Brazos Valley Health Assessment, random digit dialing methodology was used to recruit adults from the six rural counties in this study for a mailed survey [Wendel, Alaniz, Burdine, Sharkey, Felix, Windwehen: Regional Efforts Aimed at Health Equity: A Case Study in the Brazos Valley, submitted]. More than 1,400 participants responded to a question that asked in what type of store they buy most of their groceries. Almost 80% identified a supermarket or warehouse store; 15.9% a small grocery store; and 4.8% identified a convenience store or other. The BVFEP used ground-truthed methods in a two-stage approach to determine the access to and availability of fruits and vegetables to residents of the 101 CBG. In the first stage, trained observers systematically drove all highways (Interstate, U.S., and State), farm-to-market roads, and city or town streets/roads within the study area. All traditional (supercenters, supermarkets, and grocery stores), convenience (convenience stores and food marts), and non-traditional (dollar stores, mass merchandisers, and pharmacies) food stores were enumerated through direct observation and on-site determination of geographic coordinates using a Bluetooth Wide Area Augmentation System (WAAS)-enabled portable Global Positioning System (GPS) receiver and the World Geodetic System 1984 datum [34,36]. In the second stage, an observational survey instrument was developed, tested, and administered in all food stores by trained observers to determine the availability and variety of fruits and vegetables [58]. Definitions used to classify specific types of food stores are shown in Table 1.

The availability of fruits and vegetables was separately determined from the presence and variety of fresh and processed fruits and vegetables [30,58]. Processed fruits and vegetables included healthier canned, frozen, and juice [74]. Healthier forms of processed fruits included fruits canned in natural juice, fruits canned in light syrup, frozen fruits without added sugar, and 100% fruit juice. Healthier forms of processed vegetables included vegetables canned or frozen without oil or a sauce and 100% vegetable juice. Variety was operationalized as the number of different food items within a fruit or vegetable category (e.g., number of different fresh fruits or number of different types of canned fruits in natural juice).

The population-weighted centroid, which is a more accurate measure than the geographic centroid [36], for each of the 101 CBG was calculated using the ArcGIS Desktop tool Mean Center (Version 9.2, Environmental Systems Research Institute). This tool constructs the CBG mean center based on the mean-weighted x and y values of the block population centroids [36]. Two criteria of potential spatial access were calculated from each CBG [36,87]: 1) proximity, and 2) coverage [37]. Proximity was chosen since it is typically used to measure distance to the nearest food store. Coverage adds the dimension of variety and competition within a specific distance and is not limited to the food stores within an administratively-defined area, such as CBG or Census tract. Proximity: ESRI's Network Analyst extension in ArcInfo 9.2 was used to calculate the shortest network distance along the road network between two sets of paired point data: neighborhood (population-weighted CBG centroid) and the nearest corresponding food store within the six-county study area. The 2003 Tele Atlas Dynamap Transportation version 5.2 provided network data. Separate distances were calculated from each CBG to the nearest supercenter, supermarket, grocery store, convenience store, mass merchandiser, dollar store, and pharmacy in miles. Coverage: Network Analyst computed the total number of each type of food store within one, three, five, and 10 miles, using the shortest network distance from the population-weighted center of each CBG. Since the study area is not a large, highly dense urban area as much of the limited literature describes (e.g., Chicago, Detroit, Montreal, Los Angeles) [32,37,38], coverage distances were selected that represented a long walk (1 mile) and reachable by car (within 3, 5, and 10 miles). Proximity measured the shortest distance needed to travel to a specific type of food store, while coverage indicates the number of opportunities. More opportunities equates to greater accessibility [88]. Proximity and coverage measures were calculated for all traditional, convenience, and non-traditional food stores with a good selection of fresh fruits or vegetables, and food stores with a good selection of fresh or processed fruits or vegetables.

Release 11 of Stata Statistical Software was used for all statistical analyses; p < 0.05 was considered statistically significant. Descriptive statistics were estimated for availability of fresh and processed fruits and vegetables by type of food store. Nonparametric tests for trend were estimated across categories of increasing neighborhood deprivation [89]. Distances from the population-weighted centroid of each CBG to the nearest supermarket, traditional food store (supercenter, supermarket, or grocery store), convenience store, and non-traditional (dollar store, mass merchandiser, and pharmacy) food store were calculated. Distance to the nearest supermarket was compared with the distance to the nearest food store with a good variety of fresh fruits, fresh vegetables, fresh and processed fruits, or fresh and processed vegetables by testing for equalities in mean, median, and distribution of distance measures, using Wilcoxon Matched-Pairs signed-rank test. This is a nonparametric test to determine differences between groups of paired data. The null hypothesis is that there is no difference between both distributions. Finally, three multivariate regression models were fitted to determine the relationship of neighborhood deprivation or vehicle ownership to potential spatial access to a good variety of fruits and vegetables, controlling for population density: 1) proximity, 2) 3-mile coverage, and 3) 5-mile coverage. In multivariate regression, several dependent variables are jointly regressed on the same independent variables. The multivariate model approach was chosen instead of four separate multiple regression models (one for each outcome variable) for two reasons: 1) the four outcome variables are correlated with each other and the multivariate regression accounts for this correlation when testing hypotheses about the predictor variables; and 2) the final collection of models is easier to interpret if the same predictor variables are identified.

Table 2 shows the distribution of neighborhood socioeconomic characteristics and high levels of neighborhood need in the study area, which are presented as mean and standard deviation for the overall study area and by area-level (CBG) deprivation. There were 26 CBG (25.7%) in the five urban clusters (population range 3,181 to 11,952); 61.5% (n = 16) were considered high deprivation neighborhoods; and 81.8% of supermarkets (n = 9) were located in five urban clusters [36]. The percent of all socioeconomic characteristics increased significantly with increasing levels of deprivation. Table 2 also shows the distribution of proximity and coverage measures to supermarkets, traditional, convenience, and non-traditional food stores. Access to the nearest food store and the number of the food stores improves with increasing deprivation. Table 3 shows characteristics of CBG that are at least 10 miles one-way from the nearest supermarket or traditional food store. Residents in 47.5% of the 101 neighborhoods (CBG) and 39.3% of rural residents had to travel at least 10 miles; 26.7% of CBG and 20.7% of the population were at least 15 miles from a supermarket; and 12.9% of CBG and 10.5% of population were at least 20 miles from a supermarket. More than 40% of low-income CBG were at least 10 miles from nearest supermarket.

Observational surveys were completed in 185 (99.5%) food stores; one convenience store refused to participate. Table 4 shows the availability and variety of fresh and processed fruits and vegetables by food store type. As a group, convenience stores provided less availability of fruits and vegetables than traditional or non-traditional food stores. Almost 100% of non-traditional food stores (i.e., dollar stores, mass merchandisers, and pharmacies) offered processed fruits and vegetables. A greater percentage of convenience stores offered processed vegetables rather than processed fruits. The data show that variety of fresh or processed fruits and vegetables was better in supermarkets compared with grocery stores, and in dollar stores compared with convenience stores. Summary scores for availability and variety of fruits and vegetables, which combine fresh and processed fruits or vegetables show that dollar stores offered a greater variety of fruits than either convenience stores or mass merchandisers; variety of vegetables was greater at supermarkets compared with grocery stores. Table 5 shows the level of availability and variety of fruits and vegetables, based on a summary score of the number of different types of fruits or vegetables present. All traditional food stores marketed a good variety fruits and vegetables. More than 43% of dollar stores and 10% of convenience stores offered a good variety of fruits and almost 94% of dollar stores and 28% of convenience stores displayed a good variety of vegetables.

As shown in Table 6, access was best for high deprivation neighborhoods - in proximity and in the number of shopping opportunities. Overall, residents had to travel a shorter distance for a good variety of fresh fruits, fresh vegetables, fresh and processed fruits, or fresh and processed vegetables than to the nearest supermarket. The difference in distance to the nearest supermarket compared with the nearest food store with a good selection of fruits or vegetables remained significant for low and medium economically deprived neighborhoods; however, the differences were not statistically significant for high deprivation neighborhoods. Access, regardless of level of neighborhood deprivation, was better for a good variety of fresh and processed fruits or vegetables than for a good variety of fresh alone. Chloropleth maps (Figures 3 and 4) illustrate the spatial distribution of potential access to good varieties of fresh fruits and vegetables and neighborhoods with medium and high socioeconomic deprivation. The darkest color area indicates CBG that are greater than 10 miles one-way to the nearest food store with a good selection of fresh fruits (16.7% of medium and 11.1% of high deprivation areas) or vegetables (18.8% of medium and 14.8% of high deprivation areas). Table 7 and 8 show proximity and coverage of supermarkets, traditional food stores, and food stores with a good variety of fruits or vegetables by area-level vehicle ownership. Areas with lowest vehicle ownership had relatively better access to food stores and fruits or vegetables. Figures 5 and 6 indicate those areas with low (>9.5% of occupied households without a vehicle) or medium (4.6-9.5% without a vehicle) vehicle access. More than 20% of low vehicle access areas lacked access to fresh fruits and 23.5% to fresh vegetables within 10 miles. Chloropleth maps (see Figures 7, 8, 9, and 10) show access improved when fresh and processed fruits or vegetables were combined into overall fruits or vegetables. Still, there were areas where travel greater than 10 miles was necessary to access the nearest food store with a good variety of fresh or processed fruits or vegetables.

Multivariate linear regression models were used to examine the relationship between neighborhood socioeconomic deprivation or vehicle ownership and access to a good variety of fruits or vegetables, controlling for population density. Table 9 shows that residents in high deprivation or low vehicle ownership areas, compared with low deprivation or high vehicle ownership areas, had to travel a significantly shorter distance to the nearest food store for fruits or vegetables. Population density was significant; the greater the population density, the better the access to the nearest fruits or vegetable opportunity. A similar relationship for 3-mile coverage is shown in Table 10 and for 5-mile coverage in Table 11. High deprivation or low vehicle ownership areas were associated with a greater number of shopping opportunities for fruits or vegetables. As with the prior analysis, increasing population density was similarly associated with greater coverage.

Fresh fruits and vegetables were available in all traditional food stores (supercenters, supermarkets, and grocery stores) and in none of the non-traditional food stores (dollar stores, mass merchandisers, and pharmacy); however, 6% (n = 9) of 140 convenience stores marketed both fresh fruits and vegetables and 22% marketed either fresh fruits or vegetables. The inclusion of healthier canned fruits expanded the picture of availability to include all non-traditional food stores and 48% of convenience stores. All non-traditional food stores and 90% of convenience stores marketed canned vegetables. The variety of fruits or vegetables was greater at supermarkets compared with grocery stores. Among non-traditional food stores, the largest variety was found at dollar stores, which was greater than that found at convenience stores. These results highlight the limitations of prior methods for examining fruit and vegetable availability which may misrepresent actual availability in a number of ways [17]. For example, prior work focused on supermarkets as sole source for fruits and vegetables and omitted non-traditional food stores, such as dollar stores or mass merchandisers, which are dramatically growing in numbers and increasing the opportunities for food and beverage shopping [65,90]. Over the past 10 years, dollar stores have increased the variety of lower-price shopping and food options to consumers [91-93]. According to the Nielsen Company, dollar stores are also attracting high and middle income shoppers, in addition to their primary customer - low-income shoppers [90]. In addition to the dynamics of the retail food environment, the impact of the economic downturn and increased vehicle costs may alter food shopping patterns. Prior examinations of fruit and vegetable access failed to include the presence of canned or frozen forms even though dietary recommendations specify fresh, canned, frozen, or 100% juice [70]. Furthermore, the variety (i.e., the number of different types of fruits or vegetables) is often not considered as part of availability, which results in a lack of differentiation between a location having one type of fruit compared with multiple types of fruits.

Spatial access to a good variety of fresh or processed fruits or vegetables, using proximity (distance to the nearest food store) and coverage (number of shopping opportunities), was better for rural neighborhoods than access to the nearest supermarket. On average, rural neighborhoods (CBG) were 9.9 miles to the nearest supermarket, 7.0 miles to the nearest traditional food store (supercenter, supermarket, or grocery store), 6.7 miles and 7.4 miles to the nearest food store with a good variety of fresh fruits and vegetables, respectively, and 4.7 miles and 4.5 miles to a good variety of fresh and processed fruits or vegetables. As demonstrated in this study, the distance varied greatly depending whether access was to the nearest supermarket, the nearest food store regardless of type with a good variety fresh fruits or vegetables, or the nearest food store with a good variety of fresh and processed fruits or vegetables. Access to an available supply of fruits or vegetables provides a more realistic picture of rural access than distance alone to a supermarket or traditional food store, or store density for limited types of fresh fruits and vegetable [36,57]. Interestingly, the differences in distance remained significant for neighborhoods classified as having low or medium socioeconomic deprivation, but not for high deprivation, where proximity and coverage was generally best. High deprivation neighborhoods were located at a distance that would require access and resources for a car for transportation and where the largest percentage of occupied housing does not have an available vehicle. For neighborhoods with lower vehicle ownership, the median distance to a food store or selection of fruits or vegetables was beyond walking distance.

Food deserts have been described as areas, particularly lower income neighborhoods, with limited access to affordable and nutritious food [17]. In small-town and rural areas, food deserts have been defined as areas more than 10 miles from a supermarket [55,94]. The 10 mile threshold is considered "somewhat arbitrary, considering that without a car, any distance of more than a mile or so could be considered unacceptably far" [17]. As shown in the choloropleth maps, there were several medium and high deprivation neighborhoods that could be described as food deserts, where residents lacked access to supermarkets or to fresh or processed fruits or vegetables [95]. Maps also illustrated vehicle ownership areas where food stores or supplies of fruits or vegetables were not accessible.

Finally, the findings from multivariate regression models confirmed that the most deprived or lowest vehicle ownership rural neighborhoods were not the most food-disadvantaged neighborhoods. Compared with low deprivation or high vehicle ownership neighborhoods, high deprivation or low vehicle ownership neighborhoods had better spatial access to a good variety of fruits and vegetables, both in the distance to the nearest source for fruits or vegetables and in the number of shopping opportunities, after controlling for the influence of neighborhood population density. Population density was included in the models, based on prior work in rural areas that found that food stores were located in closer proximity to neighborhoods or communities with greater population density [36,96].

It has been suggested that poorer quality neighborhoods amplify individual disadvantages through poorer access to healthy foods [21]. Key findings from this study add to the discussion and understanding of the influence of neighborhood inequalities on physical access to healthy foods. Access to food stores (primarily large supermarkets) has been studied in the U.S., U.K., Europe, Canada, Australia, and New Zealand, and with mixed results. In some studies, poor or minority areas provided little or no access to supermarkets [25,38,39,48,53,97-102]; other studies found little or no difference between deprived and affluent areas in access to supermarkets [37,59,71,78,103,104], or better access from deprived neighborhoods [36,60,61,104,105]. Importantly, we followed the recommendation of Macdonald and colleagues [61] that the nutritional value of the foods available in food stores should be considered and not just the proximity or density of retail food stores. In fact, the U.S. Department of Agriculture's food guidance system in its dietary recommendations for fruits and vegetables identify canned, frozen, and 100% juice in addition to fresh as a way to help people achieve the recommended variety and amount of fruits and vegetables [70]. Supporting this is the research that confirmed nutrient benefits of canned and frozen fruits and vegetables [69,71-77]. Our results show that, in this large rural area, the more disadvantaged neighborhoods had relatively better potential access to fresh fruits and vegetables and to combined fresh and processed fruits and vegetables than less disadvantaged neighborhoods. This suggests different processes operating in rural areas compared with large urban areas in U.S., many of which are socially and racially segregated.

Physical access to food stores has been shown to be a major problem for people in deprived communities; those without cars, older residents, people on low incomes, and residents of rural areas [36,47,48,53-57]. There may be challenges, even for those with relatively close potential access to healthy foods, due to the availability of a vehicle, lack of public transportation, limited financial resources (type, amount, timing, and competing demands), problems with the home environment (food storage, meal preparation area, and refrigeration), and the constraints of household size, and employment including location and work schedule. For example, limited household refrigeration may require a consumer to make frequent, costly trips for perishable food items; or purchase more expensive or less healthy food items from a retail store closer to home [14,30,31].

There are several major methodological strengths to this study. First, this study relied on the identification of all traditional, convenience, and non-traditional food stores through ground truthing, a methodology that we have previously shown to be more accurate in small-town and rural areas than secondary or publicly acquired lists [34,36]. Second, we included multiple store types, such as convenience stores, dollar stores, and mass merchandisers, which reflect a more realistic picture of potential retail food opportunities. Finally, availability and variety of fruits and vegetables were determined through a comprehensive on-site observational survey that included fresh, canned, frozen, and 100% juice forms of fruits and vegetables. This is in response to the criticism of others who posit that it is not enough to determine location of food stores in relation to neighborhoods without also considering the quality or healthiness of the food that is available [17,21,61].

Data allow us to examine potential spatial access, but do not capture purchase behavioral characteristics; that is, where and how frequently rural residents choose to shop for fruits and vegetables. An underlying assumption with most studies is that shopping trips originate from the residence; however, the starting point for food shopping may vary and depend on time and location of work or other activities in multiple stops that include food shopping [17,106]. Although we did not identify famer's markets or fruit/vegetable stands during our ground truthing, there was no measure of community or individual gardens. Future work with rural families will allow us to understand the role of non-retail sources for fruits or vegetables, such as food sharing, home gardens, and canning or freezing. The cost of collecting ground-based, comprehensive in-store data makes it difficult to replicate this study in numerous settings. However, we believe the value from the breadth and depth of data justifies the resources. Another limitation is the use of area-level data from the 2000 U.S. Census. We acknowledge that there have been changes since 2000; most notably, increased proportion of minority residents, greater unemployment, and increased proportion of low-income residents. We plan to update our work when data from the 2010 U.S. Census is released. Still, this is the best available source of CBG-level data. Finally, we are limited in our ability to generalize beyond our rural region. However, the use of all store formats and forms of fruits and vegetables has relevance to an international audience. Future plans call for a similar examination in small-town and rural areas outside the U.S.