Abstract
A Fruit and Vegetable Prescription program (12–16 weeks, 2018–2020) was implemented at community health centers serving rural communities in the northwestern United States. The impact of the program on type 2 diabetes control was evaluated. Reductions in mean hemoglobin A1C were statistically significant (P < .01). The percentage of participants with critically high blood glucose levels (A1C > 9%) decreased from 76% (114/151) to 41% (62/151; P < .01). The findings mirror those of similar programs. The sustainability of these beneficial interventions, however, relies on improved access to preventive care. (Am J Public Health. 2022;112(7):975–979.https://doi.org/10.2105/AJPH.2022.306853)
For low-income populations, Fruit and Vegetable Prescription (FVRx) programs address some barriers to adopting healthy eating patterns.1 The Wholesome Wave FVRx program empowers participants to select healthy options and thereby helps to manage nutrition-related chronic diseases.2
INTERVENTION AND IMPLEMENTATION
Employing social cognitive theory, the program promotes behavioral change by increasing access to produce and fostering self-efficacy (individual or group appointments, cooking classes). In addition, produce prescriptions (vouchers or gift cards) are a complementary treatment for managing chronic disease.1–3 This model has been found to effectively improve diabetes control.1,2 The Navajo FVRx program exemplifies how to tailor the Wholesome Wave model for specific communities.1 Of note, residents of rural communities experience high rates of obesity and physical inactivity coupled with poor dietary choices; thus, the need for FVRx programs for this population emerges.4
Implementation strategies were tailored to the population. Food security status was evaluated using the validated Hunger Vital Sign.5 Counseling with registered dietitian nutritionists (RDNs), behavioral health counselors, or pharmacists was offered. Household size determined the value of the monthly FVRx vouchers supplied (1 person = $10; ≥ 8 people = $50). Vouchers were redeemable at neighborhood grocery stores and a mobile farmers market set up in a clinic parking lot once a week. One Federally Qualified Health Center offered participation incentives ($5 gift card, raffle to win cookware). Given the severe sociodemographic barriers of the population, program completion was defined as attending at least one activity and at least 1% voucher redemption.
PLACE, TIME, AND PERSONS
Across 12 to 16 consecutive weeks (2018–2020), the Wholesome Wave program2 was implemented at Federally Qualified Health Centers in rural Idaho and Oregon. Health care providers enrolled an unblinded, convenience sample of high-risk adults (positive diabetes diagnosis, hemoglobin A1C above normal limits).
PURPOSE
This project evaluated the efficacy of an FVRx program to improve diabetes control among rural, low-income adults (at or below the poverty level) with severe sociodemographic barriers to optimal health. Statistical analysis employed SPSS version 27 (SPSS Inc, Chicago, IL).
EVALUATION AND ADVERSE EFFECTS
Of the 333 adults (aged ≥ 18 years) enrolled, 52% (172/333) completed the program (Appendix A, available as a supplement to the online version of this article at http://www.ajph.org). A1C data were missing from 12% (21/172) of the completers, and postintervention A1C data were not available for those who did not complete the program. Analysis of postintervention data therefore included 151 records of program completers.
The attrition rate for incentivized participants was 30% (17/57), compared with 48% (161/333) overall. Ridberg et al. reported a 1% to 26% attrition rate for a comparable population and program3; however, severe sociodemographic barriers and COVID-19 yielded a high attrition rate for this program. Limited funding and clinician time inhibited quickly pivoting in-person activities to an online format. In addition, access to technology was a challenge for participants, so program activities were disbanded. Vouchers and educational materials were mailed to participants. The vouchers, however, required in-person redemption, and grocery stores were operating under new processes and with limited staff and resources. Also, given the stay-at-home order, many of the participants were reluctant to go out to redeem the vouchers.
The mean participant age was 53.7 ±10.5 years (range = 22–96 years). Participant household sizes ranged from one to 10 people (mean = 3.5 ±2.2 people). More than one quarter (40/151, 27%) lived in two-person households; three participants lived in 10-person households. Food insecurity was prevalent (91/151, 60%). Participants were primarily Caucasian/White and Latinx/Hispanic (80/151 [53%] and 56/151 [37%], respectively). The mean baseline A1C for participants was 10.3 ±2%. None of the participants had preintervention A1C readings within normal limits or the controlled range.
At least once during the intervention, most participants (127/151 [84%]) met with an RDN (82/151 [54%] individually and 47/151 [31%] by group appointment). A small percentage attended appointments with behavioral health specialists (15/151 [10%]) or pharmacists (4/151 [3%]). Nearly half (69/151 [46%]) attended at least one of 12 cooking classes. Actual produce purchased ranged from 4% to 100% of the dollar amount of vouchers supplied (mean voucher redemption rate = 60% ±28%).
Results of paired t tests showed statistically significant reductions in A1C readings (95% confidence interval) by sociodemographic factors between program completers and noncompleters. Postintervention A1C readings for more than 13% of participants were within normal limits or the controlled range (5/151 and 14/151, respectively). The percentage of participants who started with critically high A1C readings was reduced by more than one third (114/151 [76%] preintervention; 62/151 [41%] postintervention; P < .01). Participants aged 30 to 59 years and those from two- and three-person households experienced significant reductions in A1C (P < .01). Food-insecure participants experienced a greater beneficial change in A1C than food-secure participants (1.8 ±2.4 and 0.8 ±2.2, respectively). Mean reductions were statistically significant for all voucher redemption rates (P < .01) and for both incentivized and nonincentivized participants (P < .01; Table 1).
TABLE 1—
Significant Reductions in Hemoglobin A1C: Federally Qualified Health Centers in Rural Idaho and Oregon, 2018–2020
| Preintervention Noncompleters (n = 161)a | Preintervention Completers (n = 172)a | Postintervention Completers (n = 151)a | Postintervention A1C Change (n = 151)a | |||||
| No. (%) | Mean (SD) | No. (%) | Mean (SD) | No. (%) | Mean (SD) | Mean (SD) | P | |
| Age, y | ||||||||
| 30–39 | 7 (4) | 11.2 (1.7) | 14 (8) | 11.6 (2.4) | 11 (7) | 7.4 (1.8) | 3.1 (2.6) | < .01 |
| 40–49 | 46 (29) | 11.3 (1.7) | 44 (26) | 10.6 (1.6) | 38 (25) | 8.9 (1.9) | 1.8 (2.1) | < .01 |
| 50–59 | 53 (33) | 10.5 (1.5) | 59 (34) | 10.1 (1.9) | 55 (36) | 9.0 (2.1) | 1.3 (2.5) | < .01 |
| 60–69 | 34 (21) | 10.8 (1.9) | 41 (24) | 10.1 (1.8) | 38 (25) | 9.0 (1.9) | 0.7 (2.1) | .04 |
| Race/ethnicity | ||||||||
| Hispanic/Latinx | 71 (44) | 11.1 (1.7) | 56 (33) | 10.5 (1.7) | 56 (37) | 9.1 (1.9) | 1.7 (2.4) | < .01 |
| White/Caucasian | 70 (44) | 10.4 (1.6) | 102 (59.3) | 10.2 (1.9) | 80 (53) | 8.8 (1.9) | 1.3 (2.3) | < .01 |
| No. of persons in household | ||||||||
| 1 | 20 (12) | 10.8 (1.7) | 22 (13) | 10.1 (2.0) | 22 (15) | 8.9 (1.9) | 0.7 (2.2) | .03 |
| 2 | 45 (28) | 10.8 (1.7) | 40 (23) | 10.1 (2.0) | 40 (27) | 8.9 (1.9) | 0.7 (2.2) | < .01 |
| 3 | 17 (11) | 10.5 (1.9) | 30 (17) | 10.1 (1.8) | 30 (20) | 8.8 (2.0) | 2.1 (2.2) | < .01 |
| Food security status | ||||||||
| Secure | 48 (30) | 10.8 (1.5) | 42 (24) | 10.3 (1.8) | 42 (29) | 9.1 (1.9) | 0.8 (2.2) | .01 |
| Insecure | 77 (48) | 10.8 (1.7) | 103 (59.9) | 10.3 (1.9) | 91 (60) | 8.8 (2.0) | 1.8 (2.4) | < .01 |
| Voucher usage (% redeemed) | ||||||||
| 0–19.9 | . . . | . . . | . . . | . . . | 17 (11) | 8.3 (0.9) | 2.0 (2.3) | < .01 |
| 20–39.9 | . . . | . . . | . . . | . . . | 22 (15) | 9.0 (1.9) | 1.6 (2.3) | < .01 |
| 60–69.9 | . . . | . . . | . . . | . . . | 33 (22) | 9.4 (2.3) | 0.7 (2.3) | < .01 |
| 70–79.9 | . . . | . . . | . . . | . . . | 33 (22) | 8.7 (1.9) | 1.8 (2.4) | < .01 |
| 80–100 | . . . | . . . | . . . | . . . | 46 (31) | 8.8 (2.2) | 1.2 (2.4) | < .01 |
| Incentive for participation | ||||||||
| Incentivized | 0 (0) | . . . | 57 (33) | 9.6 (1.6) | 40 (27) | 8.5 (1.9) | 0.9 (1.8) | < .01 |
| Not incentivized | 161 (100.0) | 10.9 (1.7) | 115 (67) | 10.7 (1.9) | 111 (73.5) | 9.1 (2.0) | 1.5 (2.5) | < .01 |
| Level of blood glucose control (A1C) | ||||||||
| Within normal limits < 6.0) | 0 (0) | . . . | 0 (0) | . . . | 5 (3) | 5.6 (0.3) | 3.2 (1.9) | .02 |
| Controlled (6.0%–6.9%) | 0 (0) | . . . | 0 (0) | . . . | 14 (9) | 6.5 (0.2) | 3.8 (2.3) | < .01 |
| Uncontrolled (7.0%–8.9%) | 9 (6) | 8.3 (0.6) | 36 (24) | 8.4 (0.4) | 70 (46) | 7.9 (0.5) | 1.9 (1.8) | < .01 |
| Critically high (> 9.0%) | 152 (94.4) | 11.0 (1.6) | 114 (75.5) | 11.0 (1.7) | 62 (41) | 10.8 (1.6) | 0.1 (2.2) | .72 |
Note. Results are from paired t-tests.
Of the 333 enrolled, 48% (161/333) dropped out of the program and 52% (172/333) completed the program. A1C data were missing from 12% (21/172) of the completers. In addition, postintervention A1C data were not available for those who did not complete the program. Analysis of postintervention data included 151 records of program completers.
Race and ethnicity were collected on the basis of the categories used in the electronic health records of the community health centers.
Linear mixed effect models were used to explore the associations between the program components (predictors) and variations in A1C (outcome), given the sociodemographic differences for each participant (95% confidence interval). The sample sizes for participants meeting with behavioral health specialists or pharmacists were small (15 and 4, respectively); these program components were therefore not included. There were no significant main effects for dietary counseling by an RDN, attending a group session led by an RDN, or participating in a cooking class. Table 2 provides pre- and postintervention mean A1C readings for unadjusted models and for adjusted models for age and race/ethnicity. Adjusted analyses for the other sociodemographic variables are not reported as the models did not converge.
TABLE 2—
Pre- and Postintervention Mean A1C Readings, Unadjusted and Adjusted for Age and Race/Ethnicity: Federally Qualified Health Centers in Rural Idaho and Oregon, 2018–2020
| Unadjusted, Mean A1C (95% CI) | Adjusted for Age, Mean A1C (95% CI) | Adjusted for Race/Ethnicity, Mean A1C (95% CI) | ||||
| Preintervention | Postintervention | Preintervention | Postintervention | Preintervention | Postintervention | |
| RDN individual counseling | ||||||
| Yes | 10.51 (9.27, 11.74) | 10.40 (9.20, 11.66) | 10.81 (9.30, 12. 32) | 10.75 (9.24, 12.26) | 10.42 (8.14, 12.71) | 10.35 (8.05, 12.64) |
| No | 10.04 (8.70, 11.38) | 9.03 (7.69, 10.37) | 9.74 (8.18, 11.30) | 8.73 (8.18, 11.30) | 9.95 (7.23, 12.17) | 8.90 (6.72, 11.15) |
| RDN group session | ||||||
| Yes | 9.38 (9.95, 11.80) | 8.50 (6.07, 10.92) | 9.19 (7.31, 11.07) | 8.31 (6.43, 10.19) | 9.38 (6.95, 11.80) | 8.45 (6.08, 10.92) |
| No | 10.50 (8.35, 23.65) | 10.16 (8.01, 12.32) | 10.84 (9.47, 12.20) | 10.50 (9.14, 11.88) | 10.50 (8.35, 12.65) | 10.16 (8,01, 12.32) |
| Cooking class | ||||||
| Yes | 10.05 (8.70, 11.39) | 9.66 (8.32, 11.01) | 9.56 (7.56, 11.56) | 9.18 (7.17, 11.18) | 9.89 (7.66, 12.12) | 9.51 (7.27, 11.74) |
| No | 10.50 (9.26, 11.73) | 9.90 (8.66, 11.13) | 10.11 (8.22, 12.01) | 9.54 (7.84, 11.44) | 10.29 (7.80, 12.59) | 9.89 (7.39, 11.99) |
Note. CI = confidence interval; RDN = registered dietitian nutritionist. Results are from mixed effects models. The dependent variable is mean A1C; the fixed variables are time and program component; the random variable is voucher usage (ranged from 4% to 100%). Adjusted analyses are not reported for other sociodemographic factors because the models did not converge. No significant differences were observed (P < .05).
Program participation incentives have been found to promote short-term health behaviors.6 The incentives may therefore affect the long-term sustainability of A1C reductions for incentivized participants.
SUSTAINABILITY
In a review (100 articles), the authors concluded that cost–benefit analysis supports the efficacy of FVRx programs for improving health outcomes.7 Participation in this FVRx program significantly improved diabetes control among rural, low-income participants. The grant funding supporting the program, however, was short-term. Sustainability relies on expanded reimbursement for FVRx programs. Because the RDNs served as both program managers and health care providers, the programmatic and clinical successes are reflective of their ability to assume dual roles. Currently health insurance does not reimburse RDNs for preventive health services. The need for improved access to RDNs for these services precipitated.
PUBLIC HEALTH SIGNIFICANCE
For decades, the public health community has been discussing the unfavorable impact of the social determinants of health—including economic and food insecurity—on chronic disease prevalence and management. Pem and Jeewon found an association between food insecurity and increased risk of inflammatory diseases (e.g., diabetes).7 They note the need for programs to reduce the prevalence of food insecurity as a strategy to decrease the risk and severity of chronic diseases. Of note, FVRx programs were found to increase access to healthy foods and improve eating patterns among Supplemental Nutrition Assistance Program (SNAP) households.1
FVRx programs offer an evidence-based strategy for addressing food insecurity and access to care. The provision of a multicomponent FVRx program was associated with short-term reductions in blood glucose levels. By using dietary modifications to help manage diabetes, these programs help control the associated health care costs.6 Furthermore, given that vouchers were distributed on the basis of household size, this FVRx program allowed other household members to benefit from increased intakes of produce.1
ACKNOWLEDGMENTS
S. Ridinger obtained funding from the Idaho Primary Care Association through a RCHN Community Health Foundation funding opportunity.
CONFLICTS OF INTEREST
None of the authors have conflicts of interest to report.
HUMAN PARTICIPANT PROTECTION
The Idaho State University institutional review board deemed the analysis of the program data to be exempt from Title Code of Federal Regulations, Part 46[45CFR 46], as it was an outcomes assessment.
See also Kapadia, p. 962.
REFERENCES
- 1.Sundberg M, Warren A, VanWassenhove-Paetzold J, et al. Implementation of the Navajo fruit and vegetable prescription programme to improve access to healthy foods in a rural food desert. Public Health Nutr. 2020;23(12):2199–2210. doi: 10.1017/S1368980019005068. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Wholesome Wave 2014http://www.wholesomewave.org/resources/publications. Accessed December 31, 2020.
- 3.Ridberg A, Bell JF, Merritt KE, et al. A pediatric fruit and vegetable prescription program increases food security in low-income households. J Nutr Educ Behav. 2019;51(2):224.e1–230.e1. doi: 10.1016/j.jneb.2018.08.003. [DOI] [PubMed] [Google Scholar]
- 4.Trivedi T, Liu J, Probst J, et al. Obesity and obesity-related behaviors among rural and urban adults in the USA. Rural Remote Health. 2015;15(4):3267. doi: 10.22605/RRH3267. [DOI] [PubMed] [Google Scholar]
- 5.Gattu RK, Paik G, Wang Y, et al. The Hunger Vital Sign identifies household food insecurity among children in emergency departments and primary care. Children (Basel). 2019;6(10):107. doi: 10.3390/children6100107. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Vlaev I, King D, Darzi A, et al. Changing health behaviors using financial incentives: a review from behavioral economics. BMC Public Health. 2019;19(1):1059. doi: 10.1186/s12889-019-7407-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Pem D, Jeewon R. Fruit and vegetable intake: benefits and progress of nutrition education interventions—narrative review article. Iran J Public Health. 2015;44(10):1309–1321. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4644575/ [PMC free article] [PubMed] [Google Scholar]