Evaluation of the 2021–2024 Implementation of the Fresh Start Enhanced Food Is Medicine Intervention Among Rural, Underinsured People With Type 2 Diabetes

Lauren R. Sastre, PhD, RDN, LDN¹; Brandon Stroud, MPH, RDN^1,2; Raven Breinholt, MPH¹; Alessio Fratacangeli, BS¹; Sarah Elliot, BS¹; Amy Turner, BS, RN¹; Bhibha M. Das, PhD, MPH³; Sissy Lee-Elmore, MBA, MPA⁴; Yolanda Cristiani, RN⁵; Jennifer A. Buxton, PharmD, CPP⁶; Belen Rogers, MPH, RDN⁷ (View author affiliations)

Suggested citation for this article: Sastre LR, Stroud B, Breinholt R, Fratacangeli A, Elliot S, Turner A, et al. Evaluation of the 2021–2024 Implementation of the Fresh Start Enhanced Food Is Medicine Intervention Among Rural, Underinsured People With Type 2 Diabetes. Prev Chronic Dis 2026;23:250471. DOI: http://dx.doi.org/10.5888/pcd23.250471.

PEER REVIEWED

Abstract

Purpose and Objectives

We aimed to evaluate the feasibility, acceptability, and impact of the Fresh Start produce prescription (FSPRx) intervention among rural, underinsured people with type 2 diabetes.

Intervention Approach

The 20-week FSPRx intervention was implemented in 15 counties in eastern North Carolina to provide 5 to 7 pounds of produce at 9 group classes and individual, telephone-based health coaching.

Evaluation Methods

We evaluated feasibility by reviewing program records, acceptability by reviewing surveys and interviews, and impact by reviewing validated pre–post questionnaires on knowledge, skills, and behavior and retrospective medical record review (glycated hemoglobin A_1c [HbA_1c]). Quantitative data were analyzed by using descriptive statistics, the Wilcoxon signed-rank test, paired sample t tests, repeated-measures analysis of variance (ANOVA), and multiple linear regression. We analyzed interview transcripts by inductive content analysis for themes.

Results

Approximately half of enrolled participants (N = 414) completed pre–post measures (n = 225) and actively participated (n = 170). Participants were majority female (65.4%) and were White (36.5%) or Black (34.3%). Program satisfaction was 96.5%. Participation was moderate for coaching encounters and class attendance (mean [SD], 3.1 [4.5] encounters; 2.0 [2.9] classes, respectively). Food literacy and consumption of fruits, vegetables, and whole grains increased significantly. Glycemic control (mean HbA_1c) improved significantly, with the greatest decreases among participants with 4 or more classes and 4 or more health coaching encounters (0.44% in the full analytic sample). Group classes were the strongest predictor of improved HbA_1c (B = −0.07, P = .06). Interest was high for flexible remote education (program notebook, 97.1%).

Implications for Public Health

Strengths of the FSPRx intervention were acceptability and impact, while feasibility was moderate. Although participation in group classes was the strongest predictor of improved HbA_1c, attendance was limited. Research is warranted to identify optimal integration of educational and behavioral support in Food Is Medicine programs, especially for rural populations.

Top

Introduction

Approximately 14.7% of US adults live with diabetes (1). Diabetes is associated with an increased risk of blindness, kidney disease, hospitalization, and annual health care costs up to $12,022 per person (1). Successful diabetes self-management necessitates addressing a range of individual behaviors that depend on resources internal and external to the health care system. While preventing diabetes is a critical role of public health, tertiary preventive programming is also warranted to reduce comorbidities and complications.

Diabetes self-management education (DSME) improves glycemic control, emphasizing regular glucose monitoring, healthy lifestyle, mental health, medication adherence, screenings, and medical care (2). The benefits of maintaining a healthy lifestyle, especially a nutrient-dense, carbohydrate-controlled dietary pattern can improve glycemic control and reduce complications and comorbidities (3–11). Despite the benefits of DSME and nutrition on glycemic control, not all people have adequate access to DSME or healthy food.

People residing in rural areas, particularly the rural southeastern US, are more likely than others to have diabetes while also having limited access to health care and health-promoting resources such as nutritious food (12–17). Food insecurity, defined as limited access to healthy food, is a concern for people with diabetes because it is associated with low diet quality, high carbohydrate consumption, and reduced capacity to maintain glycemic control (18–22). To address poor diet quality and related health outcomes, Food Is Medicine (FIM) initiatives, such as produce prescription (PRx) programs, connect people with healthy food (eg, fresh fruits and vegetables) and have demonstrated potential to improve glycemic control (23–26). While Food Is Medicine programs are promising, few have been implemented in rural areas or the rural southeastern US, and they have not included diabetes or food-centered education (23–26). Integration of food literacy education may be warranted because food literacy (eg, nutrition knowledge, use of food labels, budgeting, meal planning, culinary skills) predicts diet quality (27,28).

Comprehensive Food Is Medicine interventions that moderate the complex barriers to diet quality and glycemic control (eg, access to DSME, food literacy education) should be implemented to enhance public health efforts to improve health outcomes among people with diabetes. The Fresh Start produce prescription (FSPRx) program was developed to address gaps in previous Food Is Medicine PRx interventions while addressing the needs of rural eastern North Carolina communities, which are characterized by a high prevalence of type 2 diabetes, food insecurity, and limited socioeconomic and health care resources (26,29–33).

Top

Purpose and Objectives

The purpose of this study was to evaluate the feasibility, acceptability, and impact of the FSPRx intervention among rural, underinsured people with type 2 diabetes in 15 rural counties in eastern North Carolina. The FSPRx built on evaluation findings of a pilot PRx intervention, which identified the need and desire for more direct, intensive education and behavior support, thus leading to the integration of group classes and 1-on-1 telephone-based health coaching (34,35). The objectives of this study were to

1. Evaluate feasibility through the evaluation of group class attendance logs, health coaches’ records of encounters, and produce distribution logs
2. Evaluate acceptability through group class surveys, end-of-program surveys, and interviews exploring experience, perceived benefits, and satisfaction; and identify preferences and needs for future iterations
3. Evaluate the impact on food literacy and diet quality and clinical outcomes (glycated hemoglobin A_1c [HbA_1c]) through validated preintervention and postintervention surveys and a retrospective medical record review.

Top

Intervention Approach

FSPRx was designed to support communities in eastern North Carolina and address problems (food insecurity, poverty, diabetes, lack of access to health care) identified in previous studies (26,29–33).

Building on a previous PRx intervention

FSPRx built on our experience from the 24-week, delivery-based program, Healthy Eating and Active Lifestyle to Enhance Diabetes management (HEALED) (34,35). HEALED demonstrated significant improvement in HbA_1c levels among participants; however, low levels of engagement in indirect, passive nutrition and health education (eg, handouts) resulted in no effect on food literacy (34,35). Participants reported a desire for intensive, direct health and nutrition education with an emphasis on culinary support (34,35). FSPRx educational content, messaging, and approaches were guided by the Health Belief Model (37), which has been shown to improve dietary intake and diabetes self-management (38). Education was delivered through in-person group classes followed by 1-on-1 telephone-based health coaching support. The full, evidence-based theoretical model of the FSPRx intervention targeted established mediators on health outcomes (39) through evidence-based approaches tailored to meet community needs (Figure 1).

Figure 1.
Theoretical model for the Fresh Start produce prescription intervention, or FSPRx, rural eastern North Carolina, 2021–2024. The determinants of health were based on Hood et al (39). [A tabular description of this figure is available.]

Eligibility, clinical partners, recruitment, and enrollment

Eligibility criteria for FSPRx included 1) being an adult aged 18 to 64 years without health insurance, 2) a medical diagnosis of prediabetes or type 2 diabetes, and 3) ability to provide informed consent in English or Spanish. Clinical partnerships through the North Carolina Association of Free and Charitable Clinics Access East were essential to recruitment and evaluation. Implementation was staggered with 2 clinics in year 1 (2021–2022), 4 clinics in year 2 (2022–2023), and 5 clinics in year 3 (2023–2024). Clinical partners supported patient referrals through 1) providing lists of all active patients with diabetes, 2) reviewing electronic health records to flag patients who could be referred to the FSPRx, 3) following up to connect these patient referrals to the research team, and 4) recruiting patients in face-to-face encounters in waiting rooms. Patients were scheduled for in-person enrollment appointments on site at the locations of clinical partners. During enrollment, patients provided written informed consent, were assigned an anonymous identification code (ID), completed a preintervention survey, and were encouraged to attend at least 4 group classes and participate in at least 4 health coaching encounters.

Community-based group education classes and partners

During the 20-week intervention, 9 in-person group classes were provided every other week in the evening at local community sites (eg, rural churches, a local community center). Community partners provided education. Group class included 1) hands-on DSME and food literacy education, 2) physical activity demonstrations, and 3) cooking demonstrations with taste testing. Details on curriculum are available elsewhere (40).

Health coaching — recruitment, training, and support

One-on-one behavioral support and goal setting were provided by health coaches, who were volunteer undergraduate students trained and supervised by faculty researchers with expertise in nutrition and exercise science. Training for health coaches included knowing and understanding nutrition and physical activity guidelines, motivational interviewing, and SMART (specific, measurable, achievable, relevant, timebound) goal setting. Goals focused on increasing intake of fruits, vegetables, and whole grains, with a focus on nonstarchy vegetables; decreasing intake of fried, high-sugar, and high-sodium foods; and increasing physical activity. Frequency of health coaching encounters was determined by participants, with a target of at least 1 weekly telephone check-in. Details of encounters (date, summary, and notes on goal setting) were recorded in a Research Electronic Data Capture (REDCap) tool built for FSPRx.

PRx and food partners

Participants were provided a mixed selection (15–25 lb) of fresh produce at each of 9 group classes. This design was built on findings from HEALED, which demonstrated that exposure to a range of produce along with nutrition support (eg, education on health benefits) increases acceptance, willingness to try new foods, and dietary diversity (34,35). Produce was provided at no cost in collaboration with the Food Bank of Central and Eastern North Carolina’s Greenville branch, which provided more than 10,000 pounds of produce during the program, with supplemental produce directly from farms in partnership with the Society of St. Andrew.

Top

Evaluation Methods

We conducted a mixed-methods implementation evaluation. The intervention and all materials were approved after review by the University and Medical Center Institutional Review Board at East Carolina University (UMCIRB: #21–001619). All data were collected and managed in REDCap, a HIPAA–secure, web-based application designed to support data capture for research studies. Program participation and outcomes were tracked and recorded by the participants’ IDs assigned at enrollment. We summarized our evaluation measures and timeline (Figure 2).

Figure 2.
Annual activities for the Fresh Start produce prescription intervention, or FSPRx, rural eastern North Carolina, 2021–2024. Abbreviations: ECU, Eastern Carolina University; NCAFCC, North Carolina Association of Free and Charitable Clinics. [A tabular description of this figure is available.]

Objective 1: Evaluation of feasibility

We evaluated feasibility by assessing attendance at group classes, provision of produce, and encounters with student health coaches. All tracking used participant IDs.

Objective 2: Evaluation of acceptability

To examine acceptability, we administered surveys at the end of each group class and a final postintervention survey via telephone. Most questions used a 5-point Likert scale (eg, strongly agree, agree, neutral, disagree, strongly disagree). Group class surveys focused on satisfaction, knowledge gained, and changes in behavior, and open-ended questions explored favorite parts of classes, areas for improvement, and suggestions. The final postintervention survey, administered at the end of each intervention cycle, included closed-ended questions on overall program satisfaction, satisfaction with each of the 3 intervention components, use of recipes and produce, motivation for enrolling, and perceived health and nutrition impact. Open-ended questions explored changes in vegetable preferences, barriers to using produce, favorite aspects, and areas for improvement.

In years 1 (2022) and 2 (2023), we conducted semistructured interviews. Patients were invited to be interviewed when the postintervention surveys were collected. All interviews were recorded via Rev (Rev.com, Inc) and transcribed verbatim. Interviews ranged from 10 to 60 minutes; most lasted 30 minutes. At the end of year 3, in a formative evaluation, we conducted a telephone survey among participants to examine interest in new models of implementation, education approach, and resources (eg, program notebook, social media, website) and potential additional resources or referrals (eg, mental health, dental, primary care, physical therapy).

Objective 3: Evaluation of impact on nutrition and health outcomes

The preintervention survey (in person at enrollment) and the final postintervention survey (via telephone) were identical and included the following components: 1) questions on sociodemographic characteristics, 2) the validated Hunger Vital Sign 2-question screener for food insecurity risk (41), 3) the validated Rapid Eating and Activity Assessment for Participants food frequency questionnaire (42) to examine dietary patterns, and 4) a validated 19-item tool to assess food literacy (43). This tool scored answers to questions of perceived self-efficacy on a scale of 1 to 7; scores were summed for food literacy subconstruct scores (eg, budgeting, shopping, meal planning, resourcefulness) and for a total food literacy score (ranging from 19 to 133) (36,43). HbA_1c data were collected by clinical staff during routine clinical visits, extracted through retrospective record review, and reported to the nearest 0.1%, with pre–post measures collected within 3 months before and after the intervention.

Analysis

Quantitative data analysis

We used descriptive statistics (frequencies, percentages, and mean [SD]) to describe sociodemographic characteristics, group class attendance, health coaching encounters, and closed-ended survey responses. We performed bivariate analyses, tabulating program participators (patients with full preintervention and postintervention data) versus nonparticipators (patients with preintervention data, no group class attendance, and ≤1 health coaching encounter) across the following characteristics: sex, age, race, ethnicity, use of SNAP (Supplemental Nutrition Assistance Program), employment status, food pantry use, food security status, preintervention total food literacy scores, and HbA_1c We used Pearson χ² tests, t tests, and Mann–Whitney U tests to examine potential differences (eg, baseline data, diet quality) between participators and nonparticipators that might bias study outcomes.

To measure impact, we used Wilcoxon signed-rank tests to examine changes in diet quality (n = 143). We used paired sample t tests to examine changes from preintervention to postintervention in total food literacy scores (n = 142) and HbA_1c (n = 225). We used McNemar tests to examine the proportion of people who moved from uncontrolled HbA_1c (≥7.1) to controlled HbA_1c (<7.0) from preintervention to postintervention. A multiple linear regression model was then used to examine predictors of postintervention HbA_1c while controlling for baseline HbA_1c and demographic covariates (age, sex, race, and ethnicity). Independent variables included health coaching and group class participation (coded as continuous variables reflecting number of completed sessions). To evaluate whether improvements differed by program engagement level, we conducted a repeated-measures analysis of variance (ANOVA) with time (pre vs post) as the within-subjects factor and program engagement group (health coaching and group class participation targets) as the between-subjects factor. Participants were categorized as low engagement (≤4 group class attended and ≤4 health coaching encounters), health coaching–only target met (≥4 health coaching encounters but ≤4 group class attended), group class–only target met (≥4 group class attended), or both targets met (≥4 health coaching encounters and ≥4 group classes). The threshold 4 was selected because it was an identified threshold to achieve behavior change in a previous study (44). Participants were aware of this target and encouraged both during enrollment and throughout the program to attend 4 group classes and have at least 4 encounters with their health coach.

Statistical significance was set at P < .05. All quantitative data were analyzed using SPSS version 31.0 (IBM Corporation). A total of 414 participants enrolled in the program. Sociodemographic data were available for 315 participants (76.1%). HbA_1c values were available preintervention and postintervention for 225 participants. We conducted HbA_1c analyses for both the full analytic sample (n = 225) and a meaningful program engagement subsample (n = 170) to capture both overall program effectiveness and outcomes among participants with meaningful engagement. The full analytic sample included all participants with available preintervention and postintervention data, regardless of participation level, including people who did not attend any group classes or participated minimally (eg, 1 health coaching encounter). We restricted the meaningful program engagement analysis to participants who demonstrated true program engagement, defined as attending 2 or more health coaching encounters or 1 or more group classes. Participants with missing sociodemographic data were retained in analyses if participation or outcome data were available. The full analytic (n = 225) and meaningful engagement (n = 170) paired samples were sufficient to support the planned inferential analyses and detect meaningful changes in HbA_1c; therefore, the sample sizes were adequately powered to produce meaningful results. Paired sample t tests and McNemar tests were conducted on both the full analytic and meaningful engagement sample. While multiple linear regression analyses were conducted on only the full analytic sample, we conducted repeated-measures ANOVA on the meaningful-engagement sample to examine the effect of participation on HbA_1c.

Qualitative data analysis

We used inductive content analysis to review semistructured interviews to identify key themes and subthemes related to participant experiences and program perceptions. Members of the research team (L.R.S., B.S., S.E., A.T.) independently reviewed and coded 3 data-rich transcripts and developed independent preliminary codebooks. The team met to discuss the codebooks and created a single consensus codebook. Then, each research team member individually coded each transcript and met to discuss any discrepancies to reach a consensus on the final codes for each transcript.

Top

Results

A total of 414 participants enrolled in FSPRx from 2021–2024. Of these, 315 (76.1%) provided baseline sociodemographic data. Of the 315 participants with baseline sociodemographic data, 65.4% were female, 36.5% White, 34.3% Black, 53.3% unemployed, and 44.4% food insecure (Table 1). Mean (SD) age was 52.3 (9.1) years. People with uncontrolled HbA_1c (vs uncontrolled) at baseline were significantly more likely to participate in the program (χ²₁ = 4.5, P = .03). Females (vs males) were also more likely to participate (χ²₁ = 4.33, P = .03). We observed no other differences in participation by sociodemographic characteristics.

Feasibility

Engagement was moderate across programmatic components; of 414 participants, 51.2% (n = 212) attended at least 1 group class, and 39.4% (n = 163) completed at least 1 health coaching assessment and a follow-up encounter, with 2 or 3 group classes or encounters most common. The mean (SD) number of group classes attended was 2.0 (2.9) classes, while the mean (SD) number of health coaching encounters attended was 3.1 (4.5). One-quarter (25.1%; n = 104) met the intervention target of attending 4 or more group classes; 30.2% (n = 125) attended 4 or more health coaching encounters, and few participants attended all 9 group classes (4.3%; n = 18).

Acceptability

For measures of acceptability, overall satisfaction was 96.5%, with most participants who answered the question on acceptability (n = 146) indicating “very satisfied” (75.3%; n = 110) or “satisfied” (21.2%; n = 31); the remaining 3.4% (n = 5) of participants were neutral. Nearly all group class survey responses indicated that participants learned something new (99.6%; 730 of 733) and increased their confidence and skills (98.5%; 719 of 730). More than half reported purchasing new foods introduced in class (68.0%; 397 of 584), most commonly because of their health benefits (82.8%; 309 of 373) or taste (69.2%; 258 of 373), and most used a recipe at home from a cooking demonstration (72.2%; 419 of 580). Most used all or most of the produce provided (90.6%; 116 of 128) and used program recipes at home (75.0%; 108 of 144). Most participants strongly agreed or agreed that the program improved their blood glucose control (75.3%; 110 of 146), overall diet quality (73.1%; 106 of 145), willingness to try new foods (66.2%; 96 of 145), and access to fresh produce (65.9%; 87 of 132), and that it helped them to become more physically active (56.5%; 82 of 145). Themes and complementary quotes developed from qualitative interviews also support the acceptability of the program among participants (Table 2).

Impact

Participants’ food and nutrition knowledge, skills, self-efficacy, and dietary behaviors improved significantly. Diet quality shifted toward increases in healthy food consumption and decreases in processed food consumption. Pooled results across all 3 years showed that of 143 participants, the proportion consuming whole grains 3 to 4 days per week increased from 21.7% (n = 31) to 40.6% (n = 58), while those reporting consuming 5 to 7 days per week increased from 16.1% (n = 23) to 21.0% (n = 30) (z = 3.53, P < .001). Fruit consumption 3 to 4 days per week increased from 30.8% (n = 44) to 39.2% (n = 56) and 5 to 7 days per week increased from 24.5% (n = 35) to 37.8% (n = 54) (z = 4.85, P < .001). Vegetable consumption 5 to 7 days per week increased from 26.6% (n = 38) to 51.7% (n = 74) (z = 4.93, P < .001).

Of 143 participants, those reporting rarely or never consuming processed meats increased from 39.2% (n = 56) to 47.6% (n = 68) (z = −2.74, P = .006). Participants reporting rarely or never consuming sweets increased from 41.3% (n = 59) to 62.2% (n = 89) (z = −4.63, P < .001). Participants reporting rarely or never consuming sugar-sweetened beverages increased from 55.9% (n = 80) to 69.2% (n = 99) (z = −4.32, P < .001). Significant reductions were also observed in fried foods: participants reporting rarely or never consuming them increased from 33.6% (n = 48) to 46.9% (n = 67) and those reporting consuming them 3 to 4 days per week decreased from 19.6% (n = 28) to 12.6% (n = 18) (z = −2.65, P = .008). Regular chips and cracker consumption also significantly decreased, with the proportion of participants reporting rarely or never consuming increasing from 33.6% (n = 48) to 46.9% (n = 67) and consumption 3 to 4 days per week decreasing from 17.5% (n = 25) to 7.0% (n = 10) (z = −2.29, P = .02). Similar trends were observed when examining yearly stratified outcomes (Table 3A)

Among the 142 participants with complete food literacy data preintervention and postintervention, we found a mean increase of 7.7 points from 92.6 (SD, 21.6) points preintervention to 100.3 (SD, 21.4) points postintervention (t = −4.40, P < .001). We observed significant improvements in year 1, from 85.9 (SD, 20.3) points preintervention to 99.6 (SD, 16.8) points postintervention (t = −3.71, P < .001) and year 2, from 89.1 (SD, 19.5) points preintervention to 100.9 (SD, 19.5) postintervention (t = −5.64, P < .001) (Table 3B).

Improvements in HbA1c

Improvements in food and nutrition knowledge, skills, self-efficacy, and dietary behaviors translated into significantly improved clinical outcomes through improved glycemic control (ie, HbA_1c) (Table 4). Across the full analytic sample (n = 225), we found a mean decrease in HbA_1c of 0.44%, from 8.06% (SD, 1.96%) preintervention to 7.62% (SD, 1.89%) postintervention (t = 4.19, P < .001). Consistent with these findings, we found a significant shift toward greater glycemic control in the proportion of participants with controlled HbA_1c, from 40.4% (n = 91) of participants preintervention to 50.2% (n = 113) postintervention (P < .001). Results of multiple linear regression examining the effect of health coaching and group class participation on postintervention HbA_1c, controlling for baseline HbA_1c and sociodemographic characteristics, were significant (F_7,178 = 22.46; P < .001; adjusted R² = 0.45). Results for group class participation alone (B = –0.07, P = .06) indicated that for every 1 group class attended, HbA_1c decreased by an average of 0.07%, although this result was not significant. Among participants who met meaningful-engagement targets and had data on preintervention and postintervention HbA_1c (n = 170), we found a mean decrease in HbA_1c of 0.53%, from 8.17% (SD, 2.03%) preintervention to 7.63% (SD, 1.90%) (t = 4.10, P < .001). We also found a significant improvement in the proportion of participants with controlled HbA_1c, from 37.1% (63 of 170) preintervention to 48.2% (82 of 170) postintervention (P < .001). In repeated-measures ANOVA, we found that participants who did not meet either program target (≥4 health coaching encounters or ≥4 group classes attended [n = 47]) had a mean (SD) HbA_1c reduction of 0.53% (1.83%), participants who met only the health coaching target (n = 40) had a mean (SD) reduction of 0.15% (1.61%), and participants who met only the group class target (n = 29) had a mean (SD) reduction of 0.47% (1.13%). These results highlight that all program services showed positive trends in clinically significant HbA_1c improvements. However, participants who met both the health coaching and group class targets (n = 54) achieved the largest mean (SD) reduction, 0.86% (1.64%), demonstrating a dose–response benefit on HbA_1c for participants with higher levels of engagement.

Formative evaluation

Of the 119 individuals enrolled in year 3 of the program, 70 (58.8%) completed the formative evaluation. Interest in remote education strategies was high for synchronous (97.2%), asynchronous (92.7%), and self-guided program notebooks for education and goal setting (97.1%). Participants also expressed strong interest in online cooking (92.9%), physical activity demonstrations (91.4%), and educational videos via social media (89.7%). Expanded clinical and behavioral health services, including physical therapy (83.1%), dental (83.1%), and mental health care (75.4%), were highly desired (Table 5).

We summarized our successes, lessons learned, and overall implications (Figure 3).

Figure 3.
Successes and lessons learned during the Fresh Start produce prescription intervention, or FSPRx, rural eastern North Carolina, 2021–2024, and implications for Food Is Medicine interventions. [A tabular version of this figure is available.]

Top

Implications for Public Health

The purpose of this study was to evaluate the feasibility, acceptability, and impact of the implementation of the FSPRx intervention among rural, underinsured people with type 2 diabetes. To our knowledge, our evaluation of FSPRx is one of the most comprehensive evaluations of a Food Is Medicine PRx model. This study contributes examples of implementation and evaluation approaches for future Food Is Medicine PRx interventions.

Participants were highly satisfied with the FSPRx intervention. The mixed-methods evaluation showed not only significant improvements in diet quality, food literacy, and health outcomes among participants but also why and how behavior changes were made (eg, enjoying new favorite vegetables, using recipes from a group class cooking demonstration at home, reading food labels while grocery shopping,). Participants who met optimal targets (≥4 group class and ≥4 health coaching encounters) almost doubled mean declines in HbA_1c compared with HEALED, which lacked intensive, direct education and behavioral support (0.86% vs 0.47%). Furthermore, among participants who did not meet optimal goals, mean declines of HbA_1c were similar (0.44% FSPRx vs 0.47% HEALED), further demonstrating the value of enhanced education and behavioral support.

The context for these behavior changes is important. Participants reported through interviews how they integrated behavior changes into their daily lives, with most describing application of knowledge and skills gained from group class and health coaching support. Group classes, which focused on skill building and increasing health and food literacy and self-efficacy through interactive discussions and activities, physical activity, and cooking demonstrations, were the strongest predictors of clinical improvement. People who attended at least 4 group classes had the greatest improvement; attendance at each group class was associated with a 0.07% decline in HbA_1c, although this was not significant. Additionally, because group class attendance was the most effective program component, future Food Is Medicine models may consider emphasizing education and collaboration with community partners with expertise in nutrition, food literacy, and culinary knowledge (eg, Cooperative Extension, Expanded Food and Nutrition Education Program [45], Cooking Matters [46]). Research is warranted to identify the most efficient and effective behavioral and educational supports to maximize the impact of Food Is Medicine on health outcomes.

Glycemic control improved during the intervention; however, the feasibility of group education was moderate to low, with only one-quarter of participants meeting the target of attending at least 4 group classes, despite high levels of satisfaction with the intervention. Although the association between the number of group classes attended and postintervention HbA1c did not reach statistical significance in adjusted regression analyses, participants who met both health coaching and group class engagement targets achieved the largest HbA_1c reductions, suggesting a potential dose–response benefit of higher overall program engagement. It is important to recognize the unique barriers faced by rural, underresourced communities in attending in-person (eg, transportation) or virtual (eg, internet access) classes layered on top of additional common barriers (eg, scheduling, time). Our formative evaluation allowed us to identify potential areas for improving access to future program. Interest was highest in a self-guided educational notebook and moderately high in virtual options ranging from social media to synchronous or asynchronous online virtual classes. Research is needed to implement and evaluate more accessible educational approaches in rural and underresourced communities.

Limitations

Our evaluation has several limitations. First was the collection of HbA_1c data via retrospective medical record review and reliance on patients’ access and adherence to health care during the intervention. Many participants did not have full preintervention and postintervention HbA_1c data, which precluded evaluation of the full sample. Future programs should seek to ensure collection of clinical outcomes during enrollment and postintervention, potentially collecting these data directly instead of relying on patients’ encounters with clinical partners. Second, participants were recruited from and enrolled at clinical partners during face-to-face visits, which resulted in limited attendance and no-shows for approximately half of those scheduled for appointments. Future recruitment and enrollment may benefit from more flexible virtual or telephone-based enrollment. Third, participants were a unique rural, uninsured population with type 2 diabetes, which may limit generalizability of our findings to other populations. Connecting with underinsured or uninsured individuals through a wider network of community partners might strengthen future iterations of this program. Fourth, development and maintenance of undergraduate health science students as health coaches may not be possible for other Food Is Medicine programs; we had access to a large pool of student volunteers. Fifth, the pre–post design cannot determine causality, and while the FSPRx program demonstrated strong effectiveness on participant’s nutrition knowledge, behavior, and health, further research is needed to investigate and establish causality.

Conclusion

This research suggests the value of targeted food-centered nutrition and health education, grounded in theory to complement the provision of a PRx to play an essential role in improving diet quality, nutritional status, and health in a rural, underresourced population with type 2 diabetes. The FSPRx model is tailored to overcome barriers that lower-income populations commonly experience (eg, food insecurity, inadequate transportation, internet barriers, lower health literacy, limited space/equipment for cooking and physical activity). Our evaluation approach may be useful for evaluating programs designed for other lower-income and underresourced populations. Satisfaction with our program (acceptability) and impact on nutrition and health were strong; however, attendance and interaction (feasibility) were moderate to limited. Future iterations of the FSPRx program will prioritize developing and investigating approaches to increasing access to educational resources and behavior support to improve engagement and participation. Lastly, a large network of clinical and community partnerships was and will continue to be critical to the implementation of FSPRx. Research is warranted to identify the effective and efficient behavioral and educational supports to optimize impact among program participants as well as to establish robust best practices for the implementation and evaluation of Food Is Medicine programs.

Top

Acknowledgments

This study was funded by The Duke Endowment and with support from the Department of Nutrition Science, College of Allied Health Sciences at East Carolina University. The authors declare they have no conflict of interest. No copyrighted material, surveys, instruments, or tools were used in the research described in this article.

The authors thank the North Carolina Association of Free and Charitable Clinics and the clinic directors, medical providers, and staff who collaborated with us to connect us with medically underinsured/underserved patients; the patients who participated in this intervention and shared their lived and program experiences; the student volunteers from East Carolina University who worked with us to develop recipes and educational materials and supported group classes, collection and distribution of produce, and provision of health coaching; the Department of Nutrition Science, especially the chair, Dr Michael Wheeler, for strong departmental support and funding; the churches and community partners across rural eastern North Carolina who provided space for classes; the Food Bank of Central and Eastern North Carolina, especially the Greenville Branch, which supported the program with fresh produce at no cost to participants for all 3 program years; the administrative staff and leadership at the College of Allied Health Sciences for their support, especially Dr Heather Wright and Michael Hale, for identifying and allocating space for expanding the Farm2Clinic within the College of Allied Health Sciences; and the Brody School of Medicine at East Carolina University, especially Dr Paul Shackleford, for providing mobile resources (health sciences mobile unit/trailer and vans).

Top

Author Information

Corresponding Author: Lauren Sastre, PhD, RDN, LDN, Department of Nutrition Science, East Carolina University, East Carolina University Health Sciences Building, Ste 2435 F, Greenville NC 27834 (sastrel18@ecu.edu).

Author Affiliations: ¹Department of Nutrition Science, College of Allied Health Sciences, East Carolina University, Greenville, North Carolina. ²Department of Health Education and Promotion, College of Health and Human Performance, East Carolina University, Greenville, North Carolina. ³Department of Kinesiology, College of Health and Human Performance, East Carolina University, Greenville, North Carolina. ⁴WATCH Clinic, Goldsboro, North Carolina. ⁵Hope Clinic, Bayboro, North Carolina. ⁶Cape Fear Clinic, Wilmington, North Carolina. ⁷Food Bank of Central & Eastern North Carolina, Raleigh, North Carolina.

Top

References

1. Centers for Disease Control and Prevention. National diabetes statistics report. Accessed August 17, 2025. https://www.cdc.gov/diabetes/php/data-research/index.html
2. National Institute of Diabetes and Digestive and Kidney Diseases. Managing diabetes. Accessed August 17, 2025. https://www.niddk.nih.gov/health-information/diabetes/overview/managing-diabetes
3. Messina G, Alioto A, Parisi MC, Mingrino O, Di Corrado D, Crescimanno C, et al. . Experimental study on physical exercise in diabetes: pathophysiology and therapeutic effects. Eur J Transl Myol. 2023;33(4):11560. PubMed doi:10.4081/ejtm.2023.11560
4. Salama II, Sami SM, Salama SI, Abdel-Latif GA, Aboulghate A, Raslan HM, et al. . Impact of lifestyle modification on glycemic control and cognitive function among type II diabetes mellitus patients. Future Sci OA. 2023;9(1):FSO835. PubMed doi:10.2144/fsoa-2022-0060
5. AlShammari AZAM, Alhamad FF, Elbashir Abdelgadir ZE, AlShammari HM, Alahmadi M, Alharbi KF. Impact of lifestyle interventions on the progression of diabetic retinopathy in patients diagnosed with diabetes mellitus: a systematic review. Cureus. 2025;17(4):e82411. PubMed doi:10.7759/cureus.82411
6. Jing T, Zhang S, Bai M, Chen Z, Gao S, Li S, et al. . Effect of dietary approaches on glycemic control in patients with type 2 diabetes: a systematic review with network meta-analysis of randomized trials. Nutrients. 2023;15(14):3156. PubMed doi:10.3390/nu15143156
7. Pavlidou E, Papadopoulou SK, Fasoulas A, Papaliagkas V, Alexatou O, Chatzidimitriou M, et al. . Diabesity and dietary interventions: evaluating the impact of Mediterranean diet and other types of diets on obesity and type 2 diabetes management. Nutrients. 2023;16(1):34. PubMed doi:10.3390/nu16010034
8. Kobayashi Y, Long J, Dan S, Johannsen NM, Talamoa R, Raghuram S, et al. . Strength training is more effective than aerobic exercise for improving glycaemic control and body composition in people with normal-weight type 2 diabetes: a randomised controlled trial. Diabetologia. 2023;66(10):1897–1907. PubMed doi:10.1007/s00125-023-05958-9
9. Butayeva J, Ratan ZA, Downie S, Hosseinzadeh H. The impact of health literacy interventions on glycemic control and self-management outcomes among type 2 diabetes mellitus: a systematic review. J Diabetes. 2023;15(9):724–735. PubMed doi:10.1111/1753-0407.13436
10. Chrvala CA, Sherr D, Lipman RD. Diabetes self-management education for adults with type 2 diabetes mellitus: a systematic review of the effect on glycemic control. Patient Educ Couns. 2016;99(6):926–943. PubMed doi:10.1016/j.pec.2015.11.003
11. Almutairi N, Hosseinzadeh H, Gopaldasani V. The effectiveness of patient activation intervention on type 2 diabetes mellitus glycemic control and self-management behaviors: a systematic review of RCTs. Prim Care Diabetes. 2020;14(1):12–20. PubMed doi:10.1016/j.pcd.2019.08.009
12. Khavjou O, Tayebali Z, Cho P, Myers K, Zhang P. Rural–urban disparities in state-level diabetes prevalence among US adults, 2021. Prev Chronic Dis. 2025;22:E05. PubMed doi:10.5888/pcd22.240199
13. Hussain AS, De SK, Chatwin G, Shahzad M, Zulfiqar E, Govindasamy V, et al. . Regional and demographic variations in diabetes mellitus and myocardial infarction mortality among US adults: a retrospective observational analysis from 1999 to 2023. Int J Cardiol Cardiovasc Risk Prev. 2025;26:200456. PubMed doi:10.1016/j.ijcrp.2025.200456
14. Dugani SB, Mielke MM, Vella A. Burden and management of type 2 diabetes in rural United States. Diabetes Metab Res Rev. 2021;37(5):e3410. PubMed doi:10.1002/dmrr.3410
15. O’Hearn M, Lara-Castor L, Cudhea F, Miller V, Reedy J, Shi P, et al. ; Global Dietary Database. Incident type 2 diabetes attributable to suboptimal diet in 184 countries. Nat Med. 2023;29(4):982–995. PubMed doi:10.1038/s41591-023-02278-8
16. Byker Shanks C, Andress L, Hardison-Moody A, Jilcott Pitts S, Patton-Lopez M, Prewitt TE, et al. . Food insecurity in the rural United States: an examination of struggles and coping mechanisms to feed a family among households with a low-income. Nutrients. 2022;14(24):5250. PubMed doi:10.3390/nu14245250
17. Ramadurai V, Sharf BF, Sharkey JR. Rural food insecurity in the United States as an overlooked site of struggle in health communication. Health Commun. 2012;27(8):794–805. PubMed doi:10.1080/10410236.2011.647620
18. Hutchinson J, Tarasuk V. The relationship between diet quality and the severity of household food insecurity in Canada. Public Health Nutr. 2022;25(4):1013–1026. PubMed doi:10.1017/S1368980021004031
19. Casagrande SS, Bullard KM, Siegel KR, Lawrence JM. Food insecurity, diet quality, and suboptimal diabetes management among US adults with diabetes. BMJ Open Diabetes Res Care. 2022;10(5):e003033. PubMed doi:10.1136/bmjdrc-2022-003033
20. Berkowitz SA, Karter AJ, Corbie-Smith G, Seligman HK, Ackroyd SA, Barnard LS, et al. . Food insecurity, food “deserts,” and glycemic control in patients with diabetes: a longitudinal analysis. Diabetes Care. 2018;41(6):1188–1195. PubMed doi:10.2337/dc17-1981
21. Seligman HK, Jacobs EA, López A, Tschann J, Fernandez A. Food insecurity and glycemic control among low-income patients with type 2 diabetes. Diabetes Care. 2012;35(2):233–238. PubMed doi:10.2337/dc11-1627
22. Walker RJ, Garacci E, Ozieh M, Egede LE. Food insecurity and glycemic control in individuals with diagnosed and undiagnosed diabetes in the United States. Prim Care Diabetes. 2021;15(5):813–818. PubMed doi:10.1016/j.pcd.2021.05.003
23. Hager K, Du M, Li Z, Mozaffarian D, Chui K, Shi P, et al. . Impact of produce prescriptions on diet, food security, and cardiometabolic health outcomes: a multisite evaluation of 9 produce prescription programs in the United States. Circ Cardiovasc Qual Outcomes. 2023;16(9):e009520. PubMed doi:10.1161/CIRCOUTCOMES.122.009520
24. Bryce R, WolfsonBryce JA, CohenBryce A, Milgrom N, Garcia D, Steele A, et al. . A pilot randomized controlled trial of a fruit and vegetable prescription program at a federally qualified health center in low income uncontrolled diabetics. Prev Med Rep. 2021;23:101410. PubMed doi:10.1016/j.pmedr.2021.101410
25. Bourenane K, Emon N. Current state of food prescriptions used to treat cardiometabolic risk factors in the US adult population. Cureus. 2024;16(2):e53629. PubMed doi:10.7759/cureus.53629
26. Feeding America. 2022 Food insecurity in Food Bank of Central and Eastern North Carolina. 2023. Accessed May 20, 2026. https://map.feedingamerica.org/county/2022/overall/north-carolina/organization/food-bank-of-central-and-eastern-north-carolina
27. Müllertz ALO, Stjernqvist NW, Outzen MH, Bloch P, Elsborg P, Ravn-Haren G. A cross-sectional study of the association between food literacy and dietary intake among Danish adolescents. Appetite. 2024;200:107526. PubMed doi:10.1016/j.appet.2024.107526
28. Savarese M, Sapienza M, Acquati GM, Nurchis MC, Riccardi MT, Mastrilli V, et al. . Educational interventions for promoting food literacy and patient engagement in preventing complications of type 2 diabetes: a systematic review. J Pers Med. 2021;11(8):795. PubMed doi:10.3390/jpm11080795
29. Kind AJH, Buckingham WR. Making neighborhood disadvantage metrics accessible: the Neighborhood Atlas. N Engl J Med. 2018;378(26):2456–2458. PubMed doi:10.1056/NEJMp1802313
30. North Carolina Diabetes Advisory Council. North Carolina’s Guide to Diabetes Prevention and Management 2020. Accessed May 2, 2023. https://www.diabetesnc.com/wp-content/themes/dnc/assets/downloads/0321/NC_Guide_2020_PPT_for_webinar_3.23.21_final.pdf
31. Diabetes North Carolina. Diabetes management. SMES statistical map. 2017. Accessed June 1, 2026. https://www.diabetesnc.com/diabetes-management/management-facts
32. County Health Rankings & Roadmaps. North Carolina data and resources. 2025. Accessed May 20, 2026. https://www.countyhealthrankings.org/health-data/north-carolina/data-and-resources
33. Engel-Smith LN. NC’s rural provider shortage persists despite decades of efforts. NC Health News. September 8, 2020. Accessed October 23, 2023. https://www.northcarolinahealthnews.org/2020/09/08/ncs-rural-provider-shortage-persists-despite-decades-of-efforts
34. Stroud B, Jacobs MM, Palakshappa D, Sastre LR. A rural delivery-based produce prescription intervention improves glycemic control and stress. J Nutr Educ Behav. 2023;55(11):803–814. PubMed doi:10.1016/j.jneb.2023.08.006
35. Sastre LR, Stroud B, Smith E, Hendrix K, McBride O. Development and evaluation of the delivery-based HEALED Produce Rx Program for uninsured patients with diabetes in rural eastern North Carolina. Prev Chronic Dis. 2023;20:E51. PubMed doi:10.5888/pcd20.220384
36. Sastre LR, Stroud BJ. Fresh Start: a comprehensive pilot produce prescription program to improve food literacy and glycemic control in rural, uninsured patients. Am J Lifestyle Med. 2023;0(0):15598276231195776.
37. Rosenstock IM. The Health Belief Model and nutrition education. J Can Diet Assoc. 1982;43(3):184–192. PubMed doi:10.1177/15598276231195776
38. Zhang A, Wang J, Wan X, Zhang Z, Zhao S, Bai S, et al. . Community-based intelligent blood glucose management for older adults with type 2 diabetes based on the health belief model: randomized controlled trial. JMIR Mhealth Uhealth. 2025;13:e60227. PubMed doi:10.2196/60227
39. Hood CM, Gennuso KP, Swain GR, Catlin BB. County Health Rankings: relationships between determinant factors and health outcomes. Am J Prev Med. 2016;50(2):129–135. PubMed doi:10.1016/j.amepre.2015.08.024
40. Stroud BJ, Sastre LR. Evaluation of a produce prescription (PRx) program with food literacy and culinary medicine education for rural, uninsured patients with type-2 diabetes. Am J Health Promot. 2025;39(8):8901171251340385. PubMed doi:10.1177/08901171251340385
41. Cutts D, Cook J. Screening for food insecurity: short-term alleviation and long-term prevention. Am J Public Health. 2017;107(11):1699–1700.
42. Segal-Isaacson CJ, Wylie-Rosett J, Gans KM. Validation of a short dietary assessment questionnaire: the Rapid Eating and Activity Assessment for Participants short version (REAP-S). Diabetes Educ. 2004;30(5):774–781, 776, 778 passim. PubMed doi:10.1177/014572170403000512
43. Lavelle F, McGowan L, Hollywood L, Surgenor D, McCloat A, Mooney E, et al. . The development and validation of measures to assess cooking skills and food skills. Int J Behav Nutr Phys Act. 2017;14(1):118. PubMed doi:10.1186/s12966-017-0575-y
44. Jacobs M, Harris J, Craven K, Sastre L. Sharing the ‘weight’ of obesity management in primary care: integration of registered dietitian nutritionists to provide intensive behavioural therapy for obesity for Medicare patients. Fam Pract. 2021;38(1):18–24. PubMed doi:10.1093/fampra/cmaa006
45. US Department of Agriculture, National Institute of Food and Agriculture. Expanded Food and Nutrition Education Program. Page last updated April 21, 2026. Accessed June 1, 2026. https://www.nifa.usda.gov/grants/programs/capacity-grants/expanded-food-nutrition-education-program-efnep
46. Cooking Matters. Explore Cooking Matters core. 2026. Accessed June 1, 2026. https://cookingmatters.org

Top

Tables

Abbreviation: SNAP, Supplemental Nutrition Assistance Program.
^a Of 414 Fresh Start participants, 315 (76.1%) provided baseline sociodemographic data. All values are number (percentage) unless otherwise indicated.

^a Information in brackets indicates identification of interview participant.

^a Based on Segal-Isaacson et al (42).
^b Determined by Wilcoxon signed rank test; P < .05 considered significant.

^a Determined by paired sample t test; P < .05 considered significant.
^b Based on Lavelle et al (43).

Abbreviation: HbA_1c, glycated hemoglobin A_1c.
^a Participants with complete pre–post HbA_1c data, regardless of engagement level.
^b Participants who met the minimum intended program dose (≥2 health coaching sessions or ≥1 group class).
^c Paired-sample t tests compared baseline and postintervention HbA_1c in each analytic sample. P values 2-tailed; significance set at α = .05.
^d Calculated by using paired mean difference; significance set at α = .05.
^e Control defined as HbA_1c <7.0%. Improved, uncontrolled at preintervention, controlled at postintervention; worsened, controlled at preintervention, uncontrolled at postintervention; controlled, controlled at preintervention and postintervention; uncontrolled, uncontrolled at preintervention and postintervention.
^f McNemar test assessed changes in the proportion of participants who achieved clinical control (HbA_1c <7.0%) from preintervention to postintervention. P values 2-tailed; significance set at α = .05.
^g Linear regression models examined the effects of program participation (group class and health coaching sessions) on the dependent variable, postintervention HbA_1c, while controlling for baseline HbA_1c, age, sex, and race and ethnicity; only significant predictors are shown.
^h Full analytic model: n = 186; R = 0.68; R² = 0.47; adjusted R² = 0.45; F_7,178 = 22.46, P < .001.
ⁱ Per-protocol model: n = 142; R = 0.66; R² = 0.44; adjusted R² = 0.41; F_7,134 = 14.90, P < .001.

Top