Abstract
Health systems are increasingly facilitating access to healthy food for their patients through Food is Medicine (FIM) programs, primarily through voucher-based produce prescription programs (PPRs). While some clinics have hosted community-supported agriculture (CSA) programs, there is little research on programs that include both self-referred and provider-referred participants. Likewise, there is minimal literature on the relationship between health care–based food assistance programs and length of participation. This study sought to build on nascent FIM literature by exploring dietary intake and food security impacts of a health care–based CSA program with dual referral pathways and investigating the impact of dosage. We collected 164 pretest—posttest surveys. Findings demonstrated increased fruit and vegetable intake and improved household food security following program participation, consistent with previous research. We found no correlation between change in food security status and length of participation. We found a nonsignificant relationship between fruit and vegetable intake and length of participation. These findings support health care–based food assistance programs as a strategy to positively impact key drivers of health. Further research is needed on the relationship between dosage and program outcomes.
Keywords
Background
Research has demonstrated health benefits of both fruit and vegetable (F&V) consumption (Devirgiliis et al., 2024; Wang et al., 2014) and consistent food security (Arenas et al., 2018; Gundersen & Ziliak, 2015). In response, health systems are increasingly facilitating access to healthy food for their patients through Food is Medicine (FIM) programs (Jury et al., 2025; Mozaffarian et al., 2024; Tohit et al., 2025). While health care–based food assistance programs can take multiple forms, including onsite pantries or food boxes, F&V vouchers, also known as produce prescription programs (PPRs), have predominated (Rudel et al., 2023; Veldheer et al., 2020). Fewer health care organizations have utilized a subsidized community-supported agriculture (CSA) model, providing local farm produce for free or at a subsidized rate. Initial studies have shown improved F&V intake and household food security for health care–based CSA participants (Berkowitz et al., 2019; Izumi et al., 2020).
Within FIM literature, there is little research on the impact of program dosage, whether defined as volume of food, degree of participation, or program length. Xie et al. (2021) found higher utilization of a produce voucher correlated with greater spending on F&V but did not find a relationship between produce spending and diabetes control. The present study sought to document food security and F&V intake changes for a health care–based CSA program and explore the relationship between length of program participation and program outcomes for both self-referred and provider-referred participants. Understanding the relationships of referral pathway and program dosage to outcomes is crucial for designing effective programs and informing program budgets.
Method
Intervention
The health system operated a program with two referral pathways at two hospital locations. Community members self-referred into the program by completing a social needs questionnaire related to food security, nutrition security, and ability to meet basic needs on the program’s website. Health care providers referred patients based on a food security screener and, if desired, additional criteria such as household income, clinical risk factors, or undergoing cancer treatment. Participants received free produce boxes, valued at $32 per box, with vegetables, some fruits, and, occasionally, dried beans or local grains. Produce was sourced from 24 local farms and distributed weekly over 36 weeks between April and December. Each week, members had the option to pack their own share on a designated day or collect a prepacked share over a 5-day period. The weekly staffed distribution day included an opportunity to swap items if desired, but the box contents were the same regardless of the pick-up method and many members utilized both options throughout the season. As program capacity allowed, new members were enrolled at the beginning of each of the three 12-week sessions.
Study Design and Sample
This study used a single-group, pretest–posttest design with the 442 members across two hospital locations in 2024. Although the intervention occurred in two different locations, the program administration was essentially identical beyond the volume of participants. Participant demographics were also similar across locations. The proportions of women, Whites, households speaking English at home, and households with children younger than 18, as well as the mean age and weeks of participation, differed between the two sites by between 1.3% and 9.8%.
Of the 313 individuals who participated through the end of the program, 164 completed both baseline and postintervention surveys, yielding a 52% response rate. Survey completion was incentivized with drawings for $25 Visa gift cards and cooking equipment. Study procedures, which were implemented as part of program improvement processes, were deemed exempt by the University of Michigan Institutional Review Board.
Data Collection
Team members distributed the presurvey in person during orientation events. As new members joined the program, they were asked to complete the presurvey. Presurveys were collected between April 3 and October 9, 2024. Most presurveys were self-administered and completed on paper. If requested, a member of the data collection team administered the survey orally. The postsurvey was first administered in person at distribution events near the end of the season. The postsurvey was then sent out electronically to members who did not complete the survey in person. Postsurveys were collected from November 20 to December 26, 2024. Surveys were available in English and Spanish.
Average daily intake in cups of fruit, vegetables, and fruit and vegetables combined was captured through the Dietary Screener Questionnaire assessing consumption frequency in the last 30 days (National Cancer Institute, 2021). Food security was measured using the 6-item USDA Food Security Survey module. The presurvey captured participant demographics and household composition. The postsurvey included reflections on changes in cooking and eating patterns since participating in the program.
Data Analysis
Data analyses were conducted with IBM SPSS Statistics Version: 28.0.1.0 using established scoring procedures for food security and F&V intake measures. Descriptive statistics were calculated for demographics and program participation characteristics. To assess the influence of program length, we compared participants in two of the three 12-week sessions (24 weeks or fewer) with participants of all three sessions (25 weeks or more) as well as by 5-week increments above 20 weeks. Paired sample t-tests were used to assess change between baseline and postintervention. The significance level was set at p < .5.
Results
Sample Demographics
The analytical sample that completed both pre- and postsurveys consisted of 164 program members. The majority of program members (n = 105, 64%) self-referred into the program while the remaining individuals (n = 59, 36%) were referred by their health care provider. Table 1 shows the sample characteristics. The majority of respondents (n = 129, 80.1%) identified as women and as White (n = 98, 60.1%). Twelve respondents indicated seven different languages, other than English, as the language spoken at home. Respondents ranged in age from 22 to 99. Thirty-nine percent of respondents had kids younger than 18 in the household. Participants’ residency spanned across 47 zip codes. Across the 36-week program season, respondents’ length of program participation ranged from 6 to 36 weeks, with more than half (n = 88, 54%) participating for the entire program.
Characteristics of the Study Sample
Food Security Outcomes
On a scale of 0 (high food security) to 6 (very low food security), the average food security score improved from 3.1 before the program to 2.3 after the program. The change was statistically significant in a paired sample t-test at the 0.01 level (p < .001). Both self-referral and provider-referral participants had similar food security improvements. Overall, 20.4% of respondents were considered food secure prior to the program compared with 45.1% afterward. Self-reported outcomes also indicated improvements in food security, with 97.6% of participants agreeing they had greater access to high-quality F&V through the program and 86.6% of participants agreeing they could afford to buy more of the food they needed.
F&V Intake Outcomes
Table 2 shows that average intake of fruit, of vegetables (excluding french fries), and of F&V combined (excluding french fries) all increased between baseline and postintervention. The change was statistically significant for all three measures. Changes were similar for self-referral and provider-referral participants. The largest effect size was seen for F&V combined, with a Cohen’s d value of 0.76. The mean difference translates to participants eating approximately an additional one third cup of F&V each day as a result of program participation. While 73% of participants increased F&V intake over the course of the program, only 15% reported reaching the USDA recommended consumption of 3.5 to 5 daily cups of F&V at postintervention.
Changes in Dietary Intake
Self-reported changes in dietary patterns aligned with these findings. The large majority of participants agreed they had tried new F&V (n = 157, 95.7%), had eaten more servings of F&V (n = 155, 95.1%), and had eaten a greater variety of F&V (n = 154, 93.9%).
Influence of Program Participation Length
There was no correlation between change in food security status and the number of weeks in the program. Regarding F&V intake, the average intake measures for fruit, for vegetables, and for F&V combined were consistently higher for participants in the two categories with the greatest number of weeks than in the two categories with the fewest weeks (see Figure 1). When comparing participants with 25 weeks or more in the program to those with 24 weeks or fewer, the difference in F&V intake at postintervention was just above statistical significance in an independent sample t-test (p = .052).
Average Daily Fruit and Vegetable Intake in Cups by Length of Program Participation
Discussion
This study heeds the call to advance the evidence base for FIM (Jury et al., 2025), demonstrating that a health care–based food provision program can reduce food insecurity and improve dietary behavior for self-referral and provider-referral participants in a CSA-based intervention. The proportion of participants experiencing food insecurity dropped significantly following participation in the program. Over the same time period, participants reported consuming approximately an additional one third cup of F&V each day, a meaningful increase toward dietary guidelines. These findings are corroborated by other studies on similar interventions (Garner et al., 2023; Hager et al., 2023; Izumi et al., 2020; Slagel et al., 2023; Xie et al., 2021).
Despite the clear improvements in food security and F&V intake, at postintervention, more than half of participants reported experiencing food insecurity and consuming less than the recommended minimum daily intake of 3.5 cups of F&V. These findings both reinforce the value of the program in mitigating critical needs and point to the importance of understanding remaining barriers to reaching food security and optimal F&V intake. Program data indicate that health challenges, available time, and transportation limitations were the most frequent barriers to accessing the produce boxes (Program Evaluation Group, 2025). Previous qualitative research has shown that unreliable transportation and habitual food preparation methods and preferences can limit participants’ ability to benefit from FIM interventions (Parks et al., 2025).
The literature on the relationship between program dosage, defined as either quantity of food or length of program participation, and health outcomes is far less robust. While we cannot conclude from our findings that additional time in the program correlates with increased F&V intake, the evidence points to the possibility of a relationship that merits further investigation. Studies with larger samples, more consistent distribution across length of program participation, data on share pick-up rates, and a design that also captures factors that moderate participation and outcomes, including the quantity of food received relative to household size, could help elucidate this relationship.
This study has several limitations. First, the pre–post design does not allow us to rule out factors outside the intervention that may have influenced food security and F&V intake. Second, we do not know the extent to which the respondents in the study sample were comparable to nonrespondents or to program participants who ceased participation before the season end. Third, the relatively small sample size and uneven distribution across weeks of participation limited the statistical power of the analysis. Finally, the dietary screener questionnaire used, while well-validated, is not as robust as objective measures of dietary intake and potentially subject to social desirability bias.
Implications for Practice
Adequate F&V intake and household food security are associated with improved health outcomes (Devirgiliis et al., 2024; Gundersen & Ziliak, 2015). This study indicates that a health care–based CSA can affect positive change on key drivers of health, demonstrating the value of FIM to providers seeking to deliver holistic care. Health care–based food provision also has the potential to build trust in the health care system, facilitating primary health care utilization. The similarity in outcomes for self-referral and provider-referral participants shows the value of multiple pathways to enrollment as a strategy to expand program reach and mitigate access barriers.
Implications for Research
More research is needed on the effect of program dosage to help inform both cost-effectiveness (Jury et al., 2025) and program design. Studies with larger samples and randomized control trial designs are especially critical. Finally, more research is needed on factors that hinder or facilitate program participation among low-income households and moderate outcomes.
Footnotes
Authors’ Note:
The authors wish to thank the funders, farmers, and participants who make this program possible and Laura Clinton, Amelia Popowics, and Shalonda Williams for support with data entry. Food and other program costs are supported by Trinity Health Michigan in collaboration with philanthropic and healthcare funders. The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by The Michigan Health Endowment Fund.
