The study used data from the 2006 Brazos Valley Food Environment Project (BVFEP) and the 2006 Brazos Valley Community Health Assessment (BVHA), which were approved by the Institutional Review Board at Texas A&M University for six rural counties in the Central Texas Brazos Valley region (see Figure 1). These counties include five urban clusters (population 3,181-11,952), and are considered rural based on population density (persons/mi²) [37]. The rural region covers a land area of 4,466 mi² and includes more than 119,654 residents [38,39].

The BVFEP used ground-truthed methods in a two-stage approach to determine the access to fast food for residents living in the six rural counties [26,40]. 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 in order to identify all retail food venues. All traditional (supercenters, supermarkets, and grocery stores), convenience (convenience stores and food marts), and non-traditional (dollar stores, mass merchandisers, and pharmacies) food stores, full-service restaurants, and fast-food restaurants were enumerated, This process involved direct observation and on-site determination of geographic coordinates using a Bluetooth Wide Area Augmentation System (WAAS)-enabled portable GPS receiver and the World Geodetic System 1984 datum [27,40]. Table 1 provides a definition used to classify specific types of traditional and non-traditional food stores. Traditional fast-food restaurants were defined as limited-service restaurants where customers pay before eating. These outlets sell quick service foods that are ready for immediate consumption on premises, off premises, or foods that are delivered to the customer's location. Traditional fast-food restaurants did not include snack and nonalcoholic beverage bars [41,42]. In the second stage, an observational survey instrument was developed, tested, and administered in all fast-food restaurants and food stores by trained observers to determine the availability and variety of fast-food entrées and side dishes [26]. The enumerated census of food stores included 23 grocery stores/supermarkets and 143 convenience stores, which were considered to be non-traditional fast-food outlets [26]. Twelve (52.2%) grocery stores/supermarkets and 109 (76.2%) convenience stores marketed fast-food entrées.

BVHA participants were recruited by a professional survey research firm using random digit dialed telephone screening and provided consent to participate in the survey. Details on the recruitment methodology have been previously published [43-45]. Briefly, prospective participants were initially contacted by telephone; a survey booklet, cover letter, small monetary incentive, and postage paid envelope were mailed to those adults who agreed to participate. Complete data on demographic characteristics and fast-food consumption were available for 1,409 adults who resided in one of the six rural counties (216 rural respondents were excluded because of missing data in the variables of interest). All participants were geocoded to their residence using Centrus Geocoder for ESRI ArcGIS™ with Tele Atlas street reference data. Street address includes streets, County Road, Private Road, or Farm-to-Market Road. Seventeen rural survey participants were unable to be geocoded and were not included in this sample.

Personal characteristics included age, gender, race/ethnicity, household income, household size (number of people living in household), employment status, and body mass index (BMI). Household income was categorized as poverty (≤100% Federal Poverty Level [FPL]), low income (101% to 199% FPL), and above low income (≥200% FPL). Body mass index (BMI) was calculated from self-reported height and weight, and was categorized as normal (<25 kg/m²), overweight (25-29.9 kg/m²), and obese (≥30 kg/m²). Employment status was dichotomized as employed full-time for wages vs. all others. The frequency of fast-food meal consumption was assessed with the following survey question: "How many times a week do you eat fast-food meals?" Six response categories ranged from "0" to "more than 6" times a week. This variable was modeled as a continuous variable given no dietary recommendation for fast-food consumption [46]. This question is similar to one that has been used in prior studies [10,13,17,18,46-51].

The network distance measure was computed for the distance along the road network; network distance was calculated with ESRI's Network Analysis extension in ArcInfo 9.2 from each BVHA participant's residence to the geographic position in front of each traditional fast-food restaurant and non-traditional fast-food outlet (convenience stores, supermarkets, and grocery stores that marketed fast-food items) [40]. Two criteria of potential spatial access to fast-food were calculated from the residence of each BVHA respondent: 1) proximity and 2) coverage. Proximity (minimum network distance) was used to measure distance to the nearest traditional fast-food restaurant, non-traditional fast-food outlet, or fast-food opportunity (traditional fast-food restaurant or non-traditional fast-food outlet). Separate network distances in miles were calculated from the residence of each participant to the nearest traditional fast-food restaurant or non-traditional fast-food outlet within the six-county study area [52]. Network data were provided by the 2003 Tele Atlas Dynamap Transportation version 5.2. Coverage (number of different shopping opportunities) was used to identify the number of different fast-food locations within a specified network distance of the respondent's residence [52]. While proximity is typically used to measure distance to the nearest food venue, coverage adds the dimension of variety and competition within a specific distance, regardless of administratively-defined areas [52]. Using the shortest network distance from each residence, Network Analyst computed the total number of traditional fast-food restaurants, non-traditional fast-food outlets, and all fast-food opportunities within 1, 3, and 5 mile network buffers to be consistent with previous work [52].

All statistical analyses were conducted with Release 11 of Stata Statistical Software (College Station, TX); p < 0.05 was considered statistically significant. Descriptive statistics were calculated for personal characteristics, spatial access to fast food (proximity and coverage), and frequency of fast-food meal consumption. Proximity and coverage measures of traditional fast-food restaurants was compared with non-traditional fast-food outlets, and all fast-food opportunities (fast-food restaurants, convenience stores, supermarkets, and grocery stores) by testing for equalities in mean, median, and distribution of measures using the Wilcoxon signed-ranks test [53]. Bivariate correlations between frequency of fast-food meal consumption and personal characteristics were estimated using Pearson's product-moment correlation. Three multiple variable regression models were fitted to determine the association of fast-food proximity (to nearest traditional fast-food restaurant, non-traditional fast-food outlet, and any fast-food opportunity) with frequency of fast-food meal consumption. In one model, spatial access was entered as the distance to the nearest traditional fast-food restaurant; in the second model, distance to the nearest non-traditional fast-food outlet; and in the third model, spatial access was defined as distance to the nearest fast-food opportunity, regardless of type. In all models, backward elimination was used for variables not significant (p < 0.10), which included race/ethnicity, household size, and employment. Since preliminary analysis showed a significant association of gender, an interaction variable was constructed for gender and spatial access (proximity and coverage). Nine multiple variable regression models were fitted to determine the association of fast-food coverage with frequency of fast-food meal consumption. In three models, spatial access was entered as the number of traditional fast-food restaurants within 1, 3, and 5 miles; in three models, spatial access was entered as the number of non-traditional fast-food outlets within 1, 3, and 5 miles; and in three models, spatial access was defined as the number of all fast-food opportunities within 1, 3, and 5 miles.

Unlike studies that asked specifically about fast-food restaurant use from major national/international chains, our rural respondents were asked about weekly, general fast-food meal consumption [10,13,14,17-19,34,47,49-51,56]. By not restricting the responses to major chains, respondents were able to consider the more prevalent local and regional fast-food restaurants as well as other fast-food venues found in rural areas. On the average 1.3 fast-food meals were consumed by the 1,409 rural adults in our sample. However, compared with other studies, the rural men and women in this sample consumed fewer fast-food meals than reported in predominately urban areas [17,46].

Compared with a previous study of rural areas in three states that measured access by self-reported walking time from residence to the nearest fast-food restaurant [19], this study used ground-truthed methods to identify, geocode, and calculate the network distance to the nearest traditional and non-traditional fast-food venues and the number of fast-food venues within a specified distance of the residence. Other studies have used perceived access [7,57], secondary lists [11,34,35], or a combination of direct observation and telephone books [36]. Prior work has shown that the use of secondary lists distorts representations of the food environment, and that ground-truthing is the most accurate measure [27,40], especially in rural areas [27,40]. Considering that the average distance to the nearest traditional fast-food restaurant was 5.8 miles and 3.2 miles to the nearest non-traditional fast-food outlet, walking time was an inappropriate measure in our rural area. Spatial access to fast food, using both proximity (distance to the nearest location) and coverage (number of fast-food locations) measures, was significantly better to non-traditional fast-food outlets than to traditional fast-food restaurants alone. On average, the distance to the nearest fast-food opportunity was nearly half the distance to the nearest fast-food restaurant. There was twice the number of all fast-food opportunities compared to traditional fast-food restaurants within all three buffer distances. This suggests that limiting measures of spatial access to traditional fast-food restaurants alone or to major chains may significantly underestimate "true" exposure to fast food [24]. Our work overcomes limitations of prior studies that examined fast-food consumption, but failed to consider spatial access as a determinant [10,46,47].

Residential distance was associated with fast-food consumption. In regression models, after adjustment for personal characteristics, increased distance to the nearest traditional fast-food restaurant, non-traditional fast-food outlet, or all fast-food opportunities, was associated with decreased weekly frequency of fast-food meals. This is in contrast to the work of Jeffery and colleagues in an urban area where they found that proximity to the home was not related to frequency of fast-food consumption [11]. In addition, coverage for traditional fast-food restaurants, non-traditional fast-food outlets, and all fast-food opportunities was associated with fast-food consumption, regardless of the coverage area. However, the magnitude of the association of coverage, or the number of the different venues, for traditional fast-food restaurants, non-traditional fast-food outlets, or all fast-food opportunities was almost three times greater with one-mile of the residence, compared with three- and five-mile buffer areas. This suggests that both proximity to the nearest fast-food venue and the number of different choices were associated with fast-food consumption. Further, proximity and coverage for non-traditional fast-food outlets had a similar independent association with fast-food consumption as traditional fast-food restaurants. It is possible that rural residents may be similarly attracted to traditional and non-traditional fast-food venues.

The interaction of gender and spatial access was associated with fast-food consumption, controlling for the independent association of gender and spatial access. The association of distance and coverage on fast-food consumption was greater for women than for men. For comparable increase in distance to the nearest traditional fast-food restaurant, non-traditional fast-food outlet, or any fast-food opportunity, the association for women was for a decrease in frequency of fast-food meals, compared with men. Similar significant associations were observed for coverage; the greater the coverage, the less frequent consumption of fast-food meals for women. The interaction revealed that proximity and coverage do not have the same direction of association as the independent effect of proximity and coverage. There may be gender-based inequities in access to transportation, and given the average distances to both traditional fast-food restaurants and non-traditional fast-food outlets, women without available transportation may visit fast-food destinations less often. The coefficients for interaction with traditional fast-food restaurants and non-traditional fast-food outlets were similar in magnitude, indicating an association on frequency of fast-food meal consumption among women. This finding may be associated with men and women having different travel activity patterns that are influenced by family and work demands, which also influence interactions within the environment [58-62]. The relationships between better access to fast food based on home residence and less frequent fast-food consumption for women may be related to less-frequent use of fast-food restaurants than men [54], and daily travel patterns. One possible explanation is that women may be less influenced by the food outlets nearest their home because of family, social, volunteer/professional responsibilities that require travel beyond a home-centered activity space; women's travel patterns may be more varied from accommodating to changing work schedules or family demands compared to men's, whose travel patterns between work and home are relatively more stable. Another possibility may be that women, who were less likely to be employed for wages than men, consumed less fast food because they spent more time at home or other locations where fast food was not readily available. Alternatively, women may have distinct motivations for seeking out fast food, which may explain these findings [6]. Understanding men and women's interactions with the food environment in context with other behaviors and dimensions of the environment, such as the location and distribution of nonfood destinations (e.g., work, school, childcare facilities, banking and other service centers, and retail outlets), may elucidate the distinct relationship between access to fast food and consumption in men compared with women.

It is especially important to note that, in addition to spatial access to fast food, personal characteristics were significantly associated with frequent fast-food consumption. Overall, poverty status was associated with fewer fast-food meals. This is opposite of a previous finding that lower income was associated with more frequent fast-food use [56], and similar to one where higher income was associated with more frequent fast-food consumption [54].

Obesity was consistently associated with increased fast-food consumption. Prior studies have shown mixed results for the relationship between obesity and fast-food consumption [10,11,36]. Interestingly, three additional personal characteristics were not independently associated with fast-food consumption in our rural sample; namely, race/ethnicity, household size, and employment status. Correlations between these three characteristics and household income may explain the lack of statistical significance in the regression models.

There are several methodological strengths to this study. First, data for the location and availability of fast food in this large rural area were collected using ground-truthed methods, which have been shown to be more accurate in small-town and rural areas than secondary or publicly available lists [27,40,63]. Second, data were not restricted to fast-food restaurants or the major fast-food restaurant chains, but included all traditional fast-food restaurants and non-traditional fast-food outlets, such as convenience stores, supermarkets, and grocery stores that market fast-food meals. This provides a comprehensive picture of the fast-food environment. Third, our measures of access included two dimensions; namely, proximity (distance to the nearest location) and coverage (number of locations within a specified area). Finally, data on fast-food consumption was collected from a large sample of rural men and women who participated in a community health status assessment.

Although data permitted the examination of potential spatial access to fast food from the residence, travel patterns, car ownership, origin of food purchase trips, or distance to employment were not available in the data. It is not known whether other personal (e.g., fast-food preferences, available transportation) or environmental aspects, including the location and distribution of nonfood destinations (e.g., work, schools, banking and other service centers, retail outlets), may be more relevant than residential access and eating behavior measures for understanding fast-food consumption in rural women. Future work with rural families will allow documentation of other missing information, such as specific fast foods and amount consumed, location of fast-food purchase, and criteria for selecting fast-food outlet. As with other reports of food consumption, whether fast food or not, the assumption is that data represent "usual" intake, which may not be the case. A further limitation is the lack of definition of fast food in the survey and lack of validity that may result from a difference in perceived meaning of the term "fast food" among respondents. Finally, our rural study area may not be representative of other rural areas.