Study design and recruitment
We used a modified randomized design, conducted from May 2012 through September 2012, to test the feasibility of our intervention. All procedures were reviewed and approved by the Wake Forest School of Medicine institutional review board (IRB). The program, Farm Fresh Healthy Living, was developed, administered, and evaluated by a partnership of university researchers, a community nonprofit agency, and a noncertified organic farm.
Participants were clients of a nonprofit [501(c)3] community action agency in Forsyth County, North Carolina, who lived throughout the 413 square miles of the county. Over two-fifths (41.4%) of the county population is nonwhite or Hispanic (19). The county forms part of the Winston-Salem Metropolitan Statistical Area and has a population density of 859.2 persons per square mile. The agency conducts programs focused on family financial self-sufficiency and housing to help families with household incomes below 200% of the federal poverty guidelines to achieve self-sufficiency. Eligibility criteria were being 1) female, 2) at least 18 years old, 3) head of the household with at least 1 minor child, and 4) able to speak and understand English.
Typical randomization procedures were modified after extensive discussions with the agency leadership and personnel, who were reluctant to recruit women for a study in which they could possibly receive free food and then be randomized to a control condition. To alleviate this concern, a procedure to preserve the random assignment but allow recruitment for a specific condition (intervention or control) was devised and approved by the IRB. Agency personnel first produced a list of all eligible clients (N = 93), which was randomized by the university researchers to 2 lists. A random procedure was used to classify 1 list as intervention and the other as control. Names on each list were then placed in random order. Agency personnel trained by the researchers and using scripts approved by the IRB called participants in the order listed until 25 on each list had been recruited. Both intervention and control scripts recruited women to participate in research “designed to learn more about foods families in [the agency’s] programs eat, where they get these foods, and some of the factors involved in their choosing these foods.” Both scripts described the evaluation questionnaires. The intervention script described participation in the CSA program; the control script did not. Recruiters kept records of the disposition of each person on the list. From 29 potential intervention participants contacted, 1 was ineligible and 2 refused to participate (refusal ratio = 7%). From the list of 26 potential control participants contacted, 1 person was ineligible and none refused to participate. All who agreed to participate signed an authorization to have their contact information given to the researchers.
Study intervention
The Farm Fresh Healthy Living Program provided intervention families with a CSA share from the farm, consisting of a three-quarter bushel box containing a 12- to 15-pound share of seasonal fruit and vegetable items, once each week for 16 weeks from May through August 2012. Boxes were available for pick-up every Thursday between 12:00 pm and 6:00 pm from the agency offices. Late pick-ups were accommodated with advance notice when possible. The box contained at least 2 simple recipes focused on the week’s produce, with emphasis on items that may have been unfamiliar to participants. The CSA shares were discounted by the farm and purchased by the community agency using private donations. During the program, the intervention group was offered 5 evening education and skill-building sessions: 3 cooking classes conducted by local North Carolina Cooperative Extension staff and based on the Cook Smart, Eat Smart curriculum (20); a tour of the participating farm; and a grocery store tour with a dietitian focused on healthful eating on a budget.
Study evaluation and data analysis
Participants in intervention and control groups were interviewed twice by telephone: at baseline before the study began and again at follow-up after program completion. Interviews lasted 25 to 30 minutes and were conducted by an interviewer not connected with the project. This interviewer obtained verbal informed consent before collecting any study data. Participants received a $10 gift card for each interview completed. Both interviews obtained data for outcome analysis. The baseline interview included questions on personal characteristics; the follow-up interview included questions for process evaluation.
The independent variable was assignment to intervention or control group. Covariates collected included the following demographic characteristics: age, marital status, number of people in the household, years of education, race/ethnicity, and employment status. Self-efficacy for fruit and vegetable consumption was measured using a 5-item scale, ranging from “not at all confident” to “very confident” (21).
Two primary outcomes were measured. Home availability of fruits and vegetables (F&V availability) was measured using a checklist including fruit (14 items) and vegetables (25 items) (22). Items in any form — fresh, frozen, canned, or dried — were counted. Total F&V availability was obtained by summing all items; possible values ranged from 0 to 39. Fruit and vegetable intake (F&V intake) of the study participant was measured using the Centers for Disease Control and Prevention’s Behavioral Risk Factor Surveillance System (BRFSS) (23). Four questions queried the number of times in the past week a respondent had consumed fruit, dark-green vegetables, orange-colored vegetables, or other vegetables. The measure of total F&V intake was constructed by summing times per week for all 4 questions.
Process data from study records included the number of times produce was picked up during the program and the number of classes and events attended (range, 0–5). Process data from intervention participants included their reports of barriers to program participation, disposition of the food received, future willingness to provide partial payment for the program, and overall program assessment. Other process data came from debriefing meetings with agency, farm, and program staff.
Demographic characteristics of the intervention and control groups were compared by using counts and percentages. Differences were assessed by using χ2, Fisher exact, and t tests, as appropriate. General linear models predicting follow-up outcomes were constructed, adjusting for age, education, baseline values of the outcome of interest, and self-efficacy for fruit and vegetable consumption. P values equal to or less than .05 were considered significant, and all calculations were performed using SAS Version 9.2 (SAS Institute, Inc, Cary, NC).