A possible dose–response association between distance to farmers’ markets and roadside produce stands, frequency of shopping, fruit and vegetable consumption, and body mass index among customers in the Southern United States

This cross-sectional study took place in a 17 county region in northeastern North Carolina, as part of the evaluation of the Centers for Disease Control and Prevention (CDC)-funded Albemarle Regional Health Services (ARHS) Partnerships to Improve Community Health (PICH) grant. The 17-county PICH region consists of mostly non-metro, rural counties with higher rates of poverty and obesity than the rest of the state [20, 21]. This region is primarily agricultural and sparsely populated. For example, one of the PICH counties, Gates County, has a 2015 population estimate of 11,431 persons, and a 2010 population per square mile of 35.8. Bertie County has a population of 20,199 persons, and a population density of 30.4. A few of the more highly populated PICH counties have populations and population densities of 39,829 and 179.2 persons per square mile (Pasquotank County), and 54,150 and 111.9 persons per square mile (Edgecombe County) [21]. These counties have high rates of adult obesity (ranging from 29.2% for Currituck to 37.0% for Edgecombe), and often have very few retail outlets offering healthy food and beverage options, highlighting the importance of direct farm-to-consumer outlets in the PICH regions.

Trained surveyors recruited a convenience sample of farmers’ market customers between July 2015 and September 2015 at farmers’ markets (n = 7) and roadside produce stands (n = 8) located in 14 of the 17 county region of northeastern North Carolina. (For the purposes of this paper, markets and stands are referred to as “farmers’ markets”.) Between 5 and 46 customers were surveyed at each market. Eligibility criteria were being English-speaking and over 18 years of age. Potential participants were approached at the entrance of the farmers’ market and asked if they would be interested in participating in a survey about farmers’ markets. If the participant agreed (verbal consent), (s)he was given a 5-page questionnaire, and if requested, the questions were read aloud. This study was reviewed and approved by the East Carolina University Institutional Review Board (15–000427). Study participants were given a reusable grocery tote upon questionnaire completion.

Participants were queried about their frequency of farmers’ market shopping by asking “How often in the past 12 months did you buy fruits or vegetables locally grown from a farmer’s market, CSA (community supported agriculture), roadside stand, or pick-your-own produce farm?” Response options included 2 or more times per week, one time per week, 2–3 times per month, once a month, a few times per year, and never. Due to distribution of responses, there were four categories established: A few times per year or less (combination of a few times per year and never); One to several times per month (combination of once a month and 2–3 times per month); Once a week; and 2 or more times per week.

Participants could select from a list of motivators of and barriers to shopping at farmers’ markets, used in a prior study, [8] with motivators including support local farmers, fresher produce, better prices, and variety of the products, and barriers including no Supplemental Nutrition Assistance Program (SNAP), Electronic Benefit Transfer (EBT), no credit/debit accepted, market days and hours aren’t convenient, and “I get what I need from other places.” Participants could also select “other” and write in responses. The perceived relative expense of produce at farmers’ markets versus supermarkets was assessed by asking “compared to other places you purchase fruits and vegetables, is the farmers’ market more or less expensive?” with response options including more expensive, less expensive, the same price, and it depends.

The questionnaire also assessed whether fruit and vegetable purchase and consumption had increased given shopping at farmers’ markets using the following question: “As a result of your shopping at this Farmers’ Market, have you been eating more fruits and/or vegetables than before you started to shop here?” Response options were fewer fruits/vegetables, no change, more fruits and vegetables, and “this is my first time at this market.” The questionnaire also assessed the proportion of fruits/vegetables purchased at farmers’ markets relative to other goods. The number of servings of fruits and vegetables eaten per day was assessed using the following items: “On a typical day, how many servings of fruits do you eat? (A serving of fruit is like a medium sized apple or a half cup of fresh fruit. – this does not include fruit juice)” and “On a typical day, how many servings of vegetables do you eat, not including French fries? (A serving of vegetables is like one cup of green salad or half a cup of cooked vegetables.)” Response options ranged from 1 to 6 or more servings per day, and this was summed as self-reported fruit and vegetable consumption. These questions were similar to the “Food Within Reach” assessment and the items on the Child Health Assessment and Monitoring Program (CHAMP) [22, 23]. Body mass index was calculated from self-reported height and weight, as weight in kilograms divided by height in meters squared. BMI was corrected for systematic reporting error for height and weight using a method previously described [24].

We also collected customers’ residential addresses for Geographic Information Systems Analyses. Customer demographics included assessment of age (in years), sex, ethnicity, race, income, and educational attainment.

Farmers’ market audits were conducted to quantify amenities offered at each farmers’ market (e.g., payment methods, farmers’ market signage, fruits and vegetables available). The audits were completed by trained auditors (who were also administering the customer intercept questionnaires) on the day the intercept surveys were conducted. The audit included items from the validated Farmers’ Market Audit Tool (FMAT) [25] and the North Carolina Fruit and Vegetable Outlet Inventory (NCFVOI) Tool [26]. Audit data were entered into a Qualtrics survey by the auditor. To quantify the amenities offered at each market, a farmers’ market amenities index was created based upon the farmers’ market audit data. The index consisted of whether or not SNAP/EBT was accepted (0 = no, 1 = yes), the forms of payment accepted (1 point for each form of payment accepted, of cash, check, credit/debit, SNAP, WIC), farmers’ market sign (yes/no), sign promoting SNAP/EBT, a welcome booth, and availability of 17 types of fruits and vegetables (coded as the number of vendors selling that item). The amenities index was a sum of all characteristics, and ranged from 5 to 38. We also created a fruit and vegetable availability sub-score, and used this in analyses. The score was a sum of the number of vendors selling each of 17 fruit and vegetable items, and ranged from 4 to 28.

To learn more about the shopping patterns of farmers’ market shoppers, a GIS database was created. For the GIS analyses, all respondents who did not provide their address or did not live in the 17-county PICH region were omitted from the GIS analyses. Of the 263 who were surveyed, 31 lived outside a PICH county, and 33 did not give an address that was complete enough for geocoding, leaving 199 respondents for GIS analyses. All addresses were geocoded to the highest level of accuracy possible, either to the city centroid, street centroid, or to the rooftop level. Addresses of all farmers’ markets in the 17 county area and all surveyed customers were batch geocoded using the Google Maps geocoding Application Programming Interface (API) through the BatchGeo website. Address data were verified using Google maps and satellite images. Distances from participant’s home address to the closest farmers’ market, as well as the distance from their home to the farmers’ market where they completed the survey (if different from the closest farmers’ market), were calculated using ArcGIS Spatial Analyst. Distances were calculated over an integrated statewide street network to reduce edge effects and to account for customers’ ability to traverse county boundaries. Three GIS variables were calculated: (1) The distance (in miles) to the closest farmers’ market from the participant’s residential address; (2) the distance (in miles) to the farmers’ market where the participant was surveyed; and (3) the difference between the two distances, which would be zero if the participant was surveyed at the farmers’ market closest to his or her residential address.

Customer and market characteristics were analyzed using descriptive statistics, including means and standard deviation for continuous variables, and frequencies for categorical variables. Bivariate associations included correlation (two continuous variables), t-tests and Analysis of variance (for a categorical and a continuous variable), and for two categorical variables, a chi-square analysis of independence. To examine potential differences between farmers’ markets and roadside produce stands, we used t-tests and Fisher’s exact tests to examine the differences between farmers’ markets and produce stands in terms of overall amenities score, fruit and vegetable availability, sign availability, and SNAP/EBT availability.

To examine associations hypothesized in our conceptual model in Fig. 1, we used multinomial logistic regression analyses (adjusted for age, race, sex, and education) to examine associations between frequency of farmers’ market shopping (dependent variable) and distance to farmers’ market and amenities index for markets (both used as independent variables in separate analyses). The multinomial logistic regression used “a few times per year or less” as the reference group. Poisson and linear regression were used to examine the association between farmers’ market shopping frequency (independent variable) and separate dependent variables of (1) fruit and vegetable consumption and (2) BMI, respectively, adjusting for age, race, sex, and education. Poisson regression analyses were also used to examine associations between fruit and vegetable consumption and the farmers’ market amenities index. We examined the need for multi-level models, with farmers’ markets as the second level, but the random effects were not statistically significant, suggesting there was no need for multi-level models. We examined potential mediation of the relationship between shopping frequency and BMI by fruit and vegetable consumption using Baron and Kenny criteria [27]. This was an exploratory study, and we did not conduct an a priori power analysis. All analyses were conducted in SAS version 9.4 (SAS Institutes, Cary, North Carolina).

The datasets generated and analyzed for this project are not publicly available due to participant confidentiality, but de-identified datasets may be available from the corresponding author on reasonable request.