Farm Fresh Foods for Healthy Kids (F3HK): An innovative community supported agriculture intervention to prevent childhood obesity in low-income families and strengthen local agricultural economies

In-depth interviews will be conducted across the four states with approximately 12 CSA farmers who operate a CO-CSA program and approximately 12 CSA farmers who are interested in CO-CSAs but who do not currently have a program in place. Farmers will be identified though internet searches and researchers’ networks and screened for eligibility over the phone. Topic guides will include questions about farm and CSA operations, marketing strategies, customer profiles, most popular produce offerings, views on accepting SNAP/EBT, and experience using or opinions on subsidizing CSA shares for lower income customers. Farmers who operate a CO-CSA program will also be asked about lessons learned. All participating farmers will be asked to complete a checklist of seasonal produce grown on their farms and indicate approximate time of harvest. Farmers will be compensated $50 for their time.

In-person, in-depth interviews will be conducted with approximately 10 adults in low-income households in each state. Sampling will be focused on the communities in which the randomized trial (component 4) will be implemented, and will seek to include a balance of caregivers of children aged 2–7 years and those of children aged 8–12 years. Eligibility will include (1) being the primary caregiver of a child in the household aged 2–12 years; and (2) meeting an income threshold determined by being eligible for the Expanded Food and Nutrition Education Program (EFNEP) or being the primary caregiver of a child who receives free or reduced-price school lunch. Participants will be recruited using a variety of strategies, including flyers, emails, and in-person recruitment at community locations (e.g. social service departments). A screening form will be administered to interested adults. Those with children aged 8–12 years will be asked if they are willing to let their child participate in a simultaneous, but separate interview.

At the time of the interview, low-income adults will complete a brief demographic survey to document their age, sex, race/ethnicity, and the number of adults and children living in the household. The interview guide will cover eating and cooking habits, perceptions of healthy foods, community supports for healthy eating, children’s preferences for types of fruits and vegetables, child snacking, tools needed to prepare produce, children’s involvement in food preparation, thoughts on local farms and farmers, knowledge of seasonal produce, CO-CSA program thoughts and preferences, and what would make it easier or more challenging to participate in a CO-CSA program.

Immediately following the interview, an adaptation of a stated preference non-market evaluation econometric technique, commonly known as a “choice experiment” [69], will be utilized to determine individual preferences for CSA shares in hypothetical market situations. This adapted method previously has been used to examine willingness to shop at a farmers’ market [70]. The goal is to understand a participant’s stated preference for purchasing a CSA given variations in specific price, amount, variety, frequency, and distance values. Each factor will be examined individually and in combination, as food shopping conditions are often multi-factorial. Participants also will be asked about ideal values for each factor, as well as the particular produce items and amount of each of those items they would want in their CSA share. Lastly, preferences for purchasing a CSA versus shopping at a supermarket will be determined given hypothetical cost and travel time differences.

In each state, 5 children will take part in semi-structured interviews to explore fruit and vegetable familiarity and preferences, involvement in meal preparation, snacking habits, family and peer influences, nutrition knowledge and skills obtained through school or community, experiences with gardening and farms, and knowledge of seasonal produce. Child interviews will include photo elicitation [71] and a modified draw and write activity [72], where children will be asked to draw a meal and snack that they usually make for themselves and describe what they are drawing. Adults will be compensated $25 for their time and participating children will be compensated $5.

In-person, in-depth interviews will be conducted with 20 full-paying CSA members, who purchase a share from a farmer who is interviewed as part of the formative evaluation (component 1) or a farmer who will partner in the randomized trial (component 4). In each state, we will seek to include CSA members of at least two different farms in the sample, with a balance of individuals without children and those with children aged 2–12 years. The interview guide will cover views on CSA features, CSA preferences, factors that influence participation, child involvement in CSAs, food shopping preferences, opinions on food assistance programming for limited-income families, willingness to help offset the price of CSA shares, and thoughts on various CSA cost-offset strategies. Participating full-pay CSA members also will be asked to complete the same demographic survey and choice experiment outlined above. They will be compensated $20 for their time.

In each state, 5 community health educators will be recruited via emails and direct contact to take part in the formative research. For each participant, there will be three sources of data: (1) an online questionnaire including multiple choice and open-ended questions; (2) an individual, in-depth phone interview; and (3) a phone-based focus group discussion. The online questionnaire will include the same demographic questions asked of participating low-income adults and full-paying CSA members, as well as questions about professional experience, work related to local foods promotion, fruits and vegetables that are familiar and appealing to low-income families in their communities, and nutrition education programs that their organizations offer to low-income families. The interview guide will include questions about the types of nutrition education they provide, where they deliver the education, participants served, and how they suggest CSA produce could be highlighted in the curriculum that will be developed as part of component 3 (curriculum development). Follow-up phone focus group discussions, each including educators from at least two states, will be held to extend findings from the questionnaire and interviews and gather educators’ perspectives on how to engage low-income populations in nutrition education, recommendations for identifying community organizations to partner with for this type of program, and suggestions for sustainability.

All interviews and focus groups will be conducted by trained study personnel, audio-recorded, transcribed verbatim, and imported into qualitative data management software for analysis. Initially, one researcher will read and code each transcript by question, and conduct rapid analysis to inform the intervention. Subsequently, two researchers will read each transcript, create and apply an initial codebook based on the questioning structure to a subset of the data. Following this, researchers will meet to discuss emerging themes, check inter-coder reliability, resolve discrepancies, and create a consensus codebook. Data then will be recoded using the revised codebook, and key themes will be reviewed and discussed.

In addition to the audio-recording, the choice experiment exercise completed by low-income adults and full-paying CSA members will be detailed with hand-written notes, which will be used to inform the development of the coding schema. Each participant will be assigned a “willingness to shop or buy” score based on number of times they said “yes” to each purchasing situation. Frequencies and percentages of participants willing to utilize CSAs for each price/accessibility situation will be generated. Nominal variables reflecting produce preferences, given frequency, size, and price, also will be tabulated. Results will be examined separately by race/ethnicity, geographic region, and household size. Researchers will meet to resolve discrepancies between coding and tabulation.

Study personnel and educators will recruit participants via flyers, newspapers, websites, and social media, as well as direct recruitment at schools, churches, libraries, community service organizations, local events, and “word of mouth.”

Potential participants will be screened for eligibility using an electronic tablet-based screening tool. The eligibility criteria for participation is: (1) age 18 years or older; (2) English-speaking; (3) parent or legal guardian of at least one child between 2 and 12 years of age in the household (herein referred to as a caregiver); (4) meet income requirements of ≤185% of FPL (self-reported); (5) have not participated in a CSA in the past 3 years; (6) be willing to use their SNAP benefits or their own money to purchase the 50% CO-CSA share; (7) have an active email address and access to a computer or tablet for online data collection; and (8) be willing to attend CSA-tailored education classes as part of the intervention.

To enroll into the study, participants must: (1) attend a baseline assessment visit and provide written consent; (2) complete the first online questionnaire; and (3) if assigned to intervention, pay the farm a deposit equal to two weekly CO-CSA payments (refundable at the end of the season if all payments are made). The decision to require a deposit draws on the experiences of a program that provided no-cost CSA shares to low-income consumers in North Carolina, which indicate participants may be more engaged with a CSA arrangement if they make a financial commitment [55].

In each participating household, one caregiver and one child aged 2–12 years (herein referred to as the focal child) will enroll in the study. Caregivers will provide written informed consent to take part and written permission for the focal child to participate. Children 7–12 years will provide oral assent. Consent and assent will be obtained by trained study staff. Any ancillary studies that develop out of this project will obtain participant consent separately. Informed consent materials are available as Additional file 1.

The shares offered through the CO-CSA program will be identical to standard summer CSA shares offered to full-paying CSA members. If the farm offers more than one size share (e.g. small share, standard share, large share), study participants will select their preferred option. Shares can be paid for at the start of the season or on a weekly basis at pick-up using cash, EBT, and another form of payment accepted by the farm. Participants will be sent reminder messages for the first 4 weeks of the CSA season.

Educators will deliver the curriculum developed in component 3 on consecutive weeks in three-lesson clusters during early, mid-, and late CSA season, with timing determined by performance centers based on the local harvest calendar. The classes will be scheduled for days, times, and locations that are convenient for most participants. Participants will be encouraged to attend all classes and will be reminded of the date and time of each session by the local educator. Handouts will be provided so that participants can share what they learn with non-attending family members and a password-protected project website will be developed where participants who are unable to attend classes will be able to access materials that align with lesson content, including lesson handouts and recipes that feature common CSA vegetables.

To help participating families prepare their CSA produce, free kitchen tools will be provided. Participants will be able to select two to four of the following items: food processor, crockpot, stockpot, large cutting board, chef’s knife, salad spinner, reusable grocery bag. Participants will choose kitchen tools just prior to their first year of CO-CSA participation. These kitchen tools will be mailed directly to participants’ homes. Additionally, a vegetable peeler, vegetable scrub brush, and paring knife will be distributed at the healthy eating classes for participants to keep.

We will assess changes from baseline in the quantity and variety of fruits and vegetables consumed, intake of energy-dense and processed foods, and dermal carotenoid levels for both children and adults. We also will measure estimated energy intake and dietary quality for children, food-related changes in the household environment (including household food security), caregivers’ knowledge and skills related to healthy eating, and children’s BMI percentile, physical activity, and sedentary behavior (Table 4).

The study team at Cornell will oversee all online questionnaire-based data collection, including dietary recalls. Study staff at each performance center will travel to the community sites to collect anthropometric and dermal measures. The schedule for data collection is shown in Table 5. Data collection instruments can be requested from the Principal Investigator (RAS).

Anthropometric and dermal measures will be collected at baseline and in the fall of Years 2–4.

Anthropometric measures. Anthropometric measures will be taken without shoes and in light clothing according to standard procedures [81]. Weight will be measured to the nearest 0.2 lbs. using the Seca Clara 803 scale. Height will be measured to the nearest 0.125 in. with the Seca 213 portable stadiometer. Weight and height measurements will be collected in duplicate unless the two measurements are more than 0.4 lbs. or 0.125 in. apart. In that case, a third measurement will be taken.

The National Cancer Institute’s Automated Self-Administered 24-h Recall system (ASA24–2016™) will be used to collect sets of three dietary recalls for the focal child [73]. This tool is based on the US Department of Agriculture’s interviewer-administered Automated Multiple-Pass Method (AMPM) and collects data on foods, drinks, and dietary supplements consumed the previous day. For children aged 2–5 years, the adult caregiver will complete the recall for the child. For children aged 6 years and older, the child will be asked to help the caregiver complete the recall.

Study participants will complete online questionnaires at baseline, in fall of Years 2–4, and spring of Years 3–4.

Brief dietary measures. Dietary intake of fruits and vegetables will be measured using the National Cancer Institute’s All-Day Fruit and Vegetable Screener (FVS) [82]. The FVS is a validated, 19-item instrument, which asks about the frequency and quantity of consumption of fruit and vegetables during the past month. We will adapt the FVS to an online format and add visual aids to improve portion size estimation [83]. Intake of sugar sweetened beverages and processed foods will be collected using the second version of the beverage and snack questionnaire (BSQ2) [84, 85]. The BSQ2 asks about frequency of consumption of beverages, savory snacks, and sweets in the last 7 days. To simplify the tool and reduce duplication with the FVS, we will remove questions about consumption of vegetables, fruits, or 100% fruit juice and will not ask participants to specify if foods were consumed in school or out of school. We also will ask whether eating habits in the past month were typical, the proportion of meals prepared and/or eaten together, and who completed the FVS and BSQ2 for the child.

Fruit and vegetable knowledge, attitudes, and beliefs. Caregivers’ knowledge of recommended fruit and vegetable consumption will be assessed by two ordinal variables, one with response choices in cups and the other relative to plate images. Caregivers’ attitudes toward food preparation will be captured by the Negative Cooking Attitude scale which sums ordinal responses to four negative statements about cooking [86]. The Self-Efficacy for Eating/Cooking Fruits and Vegetables scale [86] will be streamlined to focus only on items relating to fruits and vegetables, and record ordinal responses to four items. General nutrition beliefs will be assessed using the General Nutrition Knowledge Belief Score which sums ordinal responses to 11 healthy eating statements [87]. These beliefs include items directly relevant to F3HK, including the importance to participants of eating diets that include plenty of fruits and vegetables, use salt or sodium only in moderation, are low in saturated fat, and use sugars only in moderation.

Caregiver’s ability to select, store, and prepare CSA produce. The Cooking Techniques and Meal Preparation Self-Efficacy scale [86] assesses participant confidence in knife skills, preparation techniques, and basic cooking skills. We will expand the scale to include additional items that align with the aims and activities of the CSA-tailored education. Items on washing, storing, freezing, and drying produce will be added, as well as the preparation of leafy greens and cooking with herbs, onions, and garlic, which will be emphasized in the curriculum.

Fruit and vegetable availability and accessibility within the home. Home fruit and vegetable availability will be assessed using questions from a 3-item scale, which asks caregivers “How often are the following true? 1) We have fruits and vegetables in my home, 2) In my home, vegetables are served at meals, 3) In my home, fruit is served for dessert” [88]. Home fruit and vegetable accessibility will be assessed using questions from a 4-item scale, which asks caregivers, “How often are the following true? 1) In my home, there is fruit available to have as a snack, 2) In my home, there is vegetable available to have as a snack, 3) In my home, there are cut-up vegetables in the fridge for my child to eat, and 4) In my home, there is fresh fruit on the counter, table, or somewhere else where my child can easily get to it” [88]. For both scales, four ordinal response options will be offered (hardly ever, rarely, sometimes, often). Caregivers also will be asked to report how easy or difficult it is to financially afford and physically access acceptable quality fruits and vegetables from five ordinal response options.

Household food security. Household food security will be classified by the 6-item short form of the USDA Food Security Survey Module (FSSM), with a 30-day reference period [89]. Items ask respondents about the conditions and experiences of food insufficiency due to resource limitations, and the behavioral responses of household members to these conditions. The form will be converted to an online format, but questions and response options will not be modified. Participants that affirm two or more items are considered to live in a food insecure household.

Nutrition assistance and education participation. Participation in school meal programs will be measured using seven questions from the National Health and Nutrition Examination Survey (NHANES) Diet Behavior and Nutrition Questionnaire [90]. Minor modifications will be made to simplify some questions and response options, and improve their suitability to an online format. We also will ask a follow-up question to determine whether the focal child’s school offers universal free breakfast. Three questions from the NHANES Food Security Module [91] will be used to assess participation in SNAP and the Special Supplemental Food Program for Women, Infants, and Children (WIC). These questions will be modified to specify a one-month reference period, rather than a one-year reference period. Study participants also will be asked if they participated in nutrition education classes other than those for the study since the last assessment, and, if yes, how many classes they attended.

Physical activity and sedentary behavior. To assess child physical activity, caregivers will be asked the number of days in past week the focal child was active for at least 60 min [92], and two questions about the amount of time the child “typically” spent playing outdoors in the last month – one related to weekdays and the other related to weekend days [93]. To assess sedentary behavior, caregivers will be asked to report the number of hours on an average school day that the focal child watches television, plays video games or computer games, or uses a computer for something that is not school work [92]. Questions drawn from the Middle School Youth Risk Behavior Survey [92] will be adapted for caregiver-report and to an online format.

General health status. Caregivers will be asked to report their general health and the general health of the focal child from five ordinal response options using questions from the NHANES converted to an online format [94]. They also will be asked to report how much time they (the adult) have spent outside per day between 10 am and 4 pm.

Household composition and demographics. At baseline, all adult participants will compete questions drawn from the American Community Survey [95] and BRFSS [96] about household composition (age, gender, and relationship of all household members) and demographic characteristics (race/ethnicity, education, employment, and household income). Questions and response options will be modified to an online format and simplified, where needed. Adult participants also will report whether anyone in the household smokes. Caregivers will be asked to provide the name, age, and gender of the focal child and the number of days in a typical week that he or she lives with the respondent.

On follow-up questionnaires, participants will be asked to update household composition, education, employment, income, smoking status, and days that child lives with respondent.

Participant, educator, and performance site coordinator-level process evaluation. A process evaluation will assess intervention dose delivered, dose received, and fidelity [97]. “Dose delivered” will be determined based on CSA share pick-up and lesson attendance. Participating farms will provide data on CSA share pick-up. Participants will complete an identifiable, paper-based questionnaire after each of the nine lessons to report their utilization of CSA share contents and their perceived utility of lesson components (“dose received”). Utilization also will be determined through a question on the fall online questionnaire which asks the portion of the share usually eaten. For households that report that they did not usually eat all of their share, follow-up questions related to preservation, spoilage, and food sharing will be asked. “Fidelity” will be assessed via post-lesson educator questionnaires and site coordinator quality assurance (QA) visits. Educators will report their fidelity to the written curriculum and their perceptions of the lesson’s acceptability, feasibility, and adaptability. Performance center coordinators will conduct two QA visits to each community during Years 2 and 3 and will audit that week’s CSA share pick-up and lesson facilitation. Audits will be used to collect information on the functionality of pick-up sites and CSA share contents. Class observations will be used to supplement educator’s reporting of lesson fidelity and to assess educators’ exhibition and use of qualities and techniques deemed important for adult education: preparedness, adaptability, knowledge, facilitation of group cohesion, respect, and promotion of a safe environment, immediacy of content and engagement. Following the CSA season, participant focus group discussions and educator interviews will be conducted to explore and elaborate upon findings from the above-described questionnaires.

Most data will be entered online directly by participants. All information collected will be kept private and confidential, and stored on a secure computer to which only members of the research team will have access. Participant’s names and all personal identifiers will be removed from data before analysis. All personal identifiers will be encrypted and password protected.

Intent-to-treat analyses will allocate all participants to initial random assignment group for measurement at Year 2 baseline, Year 2 fall assessment, and Year 3 spring assessment. Subsequently, Group 2 will be considered to have received the intervention for measurement at Year 3 fall assessment, Year 4 spring assessment, and Year 4 fall assessment. Likewise, Groups 3 and 4 will maintain initial random assignment for measurement at Year 3 spring assessment, Year 3 fall assessment, and Year 4 spring assessment, but then Group 4 will be considered to have received the intervention for Year 4 fall assessment.

Change variables for linear outcomes (Table 4) will be created by subtracting the baseline from the measurement at the end of the CSA season, as well as during subsequent measurement points. Multilevel linear models with fixed and random effects will be used because observations are repeated measures for participants nested within farm communities. Models also will include intervention group status, baseline measures of outcome variables, and relevant covariates. Additional models will control for CO-CSA and education dose received. Multiple imputation will be used to estimate partial missing data.

Logistic regression models will be used to assess whether differences in intervention implementation across sites (via “fidelity” measures) and differences in experiences between participants (via “dose delivered” and “dose received” measures) influence participants’ outcomes. Qualitative assessments of participants’ and educators’ experiences will be used to understand factors associated with dose and fidelity at each site. Data on non-participation will be analyzed using the Five Dimensions of Access framework [99]. Results will inform the use and refinement of the curriculum and CO-CSA model for future community-based nutrition interventions and will be helpful for CSA farmers interested in incorporating the CO-CSA model into their business operations.

Using the data collected from performance center coordinators, a cost analysis will first be conducted to identify and measure the direct, tangible costs of the resources used in administration and implementation of the F3HK program. This analysis will be done from both the program perspective and the broader societal perspective. Analysis from the program perspective will focus on costs directly incurred during preparation for and implementation of the various components of the F3HK program, including the monetary value of in-kind contributions. Analysis from the societal perspective will estimate all costs incurred and health effects obtained as a result of the program, including estimation of costs and effects experienced beyond the participant and program perspectives [100].

A preliminary cost-effectiveness analysis (CEA) then will be completed, which will build on the cost analyses conducted from the broader societal perspective and will include opportunity costs experienced by participants (i.e. time at CSA site or education class that could have been used in other valued ways). We will use the human capital method and value participants’ time based on wage rates [101]. We will estimate the value of participants’ time using income data collected from participant questionnaires and national average wage rates from the Bureau of Labor Statistics [102]. An incremental cost-effectiveness ratio (ICER) will be calculated by (1) subtracting the cost of the F3HK program (as determined by the cost analysis) from the cost of the alternative (no program, $0); (2) subtracting the change in fruit and vegetable intake from baseline to late-season in the control group from the change in fruit and vegetable intake during that same interval in the intervention group, as measured by the FVS; and (3) dividing the program cost by the difference in the change in fruit and vegetable intake between the control and intervention groups. This ratio will thus provide an estimate of the program cost per unit change in fruit and vegetable intake. Epidemiologic data will be used to estimate the impact of changes in fruit and vegetable intake on life expectancy and quality, or Quality Adjusted Life Years (QALYs) [101], which will allow for a comparison of cost-effectiveness between F3HK and other programs aimed at improving the public’s health.