Food Is Medicine, but Are Produce Prescriptions?

The 21st century in the U.S. has seen a groundswell of interest in so-called food-is-medicine interventions in health care practice and policy. Novel endeavors have formed across the country to narrow the chasm between health care and food systems. Bringing these unusual partners to the table, with food representing the largest sector of our economy (1) and health care spending accounting for nearly one-fifth of the gross domestic product, offers the potential to radically advance population health objectives and strengthen the economy. From the food lens, local food systems and food equity are among the perspectives represented. From the health care lens, the intersection of food and health is well aligned with the triple aim of offering food rather than (or in addition to) medicine to improve individual experience with care, addressing nutrition insecurity as a root cause of chronic disease to improve the health of populations, and reducing per capita costs of care with averted hospitalizations and health care expenditures. The most evidence exists for medically tailored meals, which are fully prepared and tailored to specific needs for individuals with serious health conditions, such as diabetes (2–7). Simulations modeling national implementation of medical tailored meals indicate that implementing food is medicine for Americans with diabetes could be associated with $19.3 billion in averted health care expenditures in 1 year, with a net policy cost savings of almost $11 billion (4).

Additional food-is-medicine models include medically tailored groceries, which are selected by a nutrition professional, and produce prescriptions (Fig. 1) (2). Great heterogeneity exists particularly for produce prescription. Newman et al. (8) surveyed the landscape of U.S. produce prescription programs prior to the coronavirus disease 2019 (COVID-19) pandemic and confirmed that diabetes is the most common eligibility criterion, followed by economic risk or food insecurity. Most programs included a health care visit (87%) and delivered nutrition education (83%) that covered chronic disease management, increasing/limiting nutrients, and food resource management, delivered by a registered dietitian via counseling or classes. Generally, the intervention design findings are consistent with a scoping review of prescription programs (9). Most produce prescription programs measure food and vegetable intake, food insecurity, HbA_1c, and participant feedback. However, evidence for produce prescriptions and medically tailored groceries remains limited.

In this issue of Diabetes Care, Hager et al. (10) retrospectively evaluated changes in HbA_1c and other health outcomes in a 6-month produce prescription program for patients from two clinics in Hartford, CT. Eligible patients were identified and referred by clinicians based on uncontrolled diabetes (HbA_1c >6.5%) and residence in lower-income zip codes. Notably, enrollment into the produce prescription program began in November 2019 to March 2020, at which point all in-person nutrition education was cancelled due to the COVID-19 pandemic and no adaptations were described. Program implementation continued through October 2020. Wholesome Wave, a large, nonprofit organization operating across many U.S. states (wholesomewave.org), administered the program. Participants received $60 per month (regardless of family size) for 6 months in paper vouchers, in clinic or by mail (exclusively by mail after March 2020), to purchase fruits and vegetables at a local grocery retail chain. Nutritional education offered included registered dietitian-led grocery store tours focused on label reading and using the vouchers at checkout and a launch event on 2 November 2019.

For evaluation of the program, patients had at least one HbA_1c measurement during the program period using electronic health record data. A control group of nonparticipants with baseline and follow-up HbA_1c measures was created using overlap weight propensity scores to estimate the probability of being enrolled in the program based on sociodemographics, comorbidities, health care utilization parameters, HbA_1c, systolic blood pressure, diastolic blood pressure, and BMI. The authors examined change in HbA_1c from baseline to 6 months after enrollment, using generalized estimating equations with overlap weights from propensity scores. All HbA_1c measurements taken within the 6-month window were used for the primary analysis. Secondary outcomes included changes in systolic blood pressure, diastolic blood pressure, BMI, inpatient hospitalizations, and emergency department admissions. After weighting for propensity score overlap weights for the 252 produce prescription participants and 534 nonparticipant control individuals, mean age was 60.4 years, mean baseline HbA_1c was 8.6%, 61.7% were female, 81.0% were Hispanic, and 46.4% had Medicaid. Change in HbA_1c was not different between participants and nonparticipants (−0.11% vs. −0.24%; difference of 0.13%; 95% CI −0.05%, 0.32%), and neither were changes in secondary outcomes. The authors provide some data on process measures, with 90% of received produce dollars in aggregate redeemed, using the grocery store’s retail sales data. Only 5% of all participants attended the in-person class at the launch event, and 9% attended one grocery store tour.

Several limitations exist. First, this is a retrospective study, and there remains potential unmeasured confounding despite use of overlap propensity score weights. As data collection was not protocolized, the primary outcome used follow-up HbA_1c values collected any time in the first 6 months and baseline values collected any time in the prior year. In addition, there were no data on fruit and vegetable intake, since these variables are not collected routinely in electronic health records. Less than one-half of the treatment group received Medicaid, a proxy indicator for lower-income status, and no data on food insecurity was collected. Regardless, the authors provide important data suggesting that provision of produce prescriptions alone may not improve glycemic control, as the vast majority of patients received no nutrition education. Smaller produce prescription programs with more engagement (e.g., community health worker and cooking demonstrations) have reported pre–post improvements in HbA_1c (11,12). It is also possible that the COVID-19 pandemic affected the intervention, although sensitivity analysis did not suggest HbA_1c improvement prior to March 2020.

There is an urgent need to understand how to best structure food-is-medicine interventions and their effects on health outcomes and health care expenses. Reviews have found that while there was some evidence for improving fruit and vegetable consumption and reducing food insecurity, the overall quality of food-is-medicine studies was weak, as many studies lacked control or comparison groups, were nonrandomized, and had small sample sizes and incomplete outcome data, often using nonvalidated measurement tools (8,9). Transportation access has been recognized as a very important barrier, as have nutrition literacy and stigma. In the spirit of cross-sector partnerships, novel models are needed that connect public and private community resources for nutrition education, likely a critical ingredient (Fig. 1). Given the mandate and capacity to serve lower-income populations in communities and clinics, Supplemental Nutrition Assistance Program Education (SNAP-Ed) may be a key resource. Early cross-sector produce prescription models have shown promise in improving dietary intake and, importantly, food resource management practices to support food security (13). While having the health care team directly involved in providing healthy foods to patients seems intuitive, additional research is needed to understand how best to facilitate communication between health care teams and patients and the effect of these interactions. With renewed energy and interest from the government, businesses, philanthropists, and the medical community, many exciting large initiatives are in the works to address the role of nutrition and food insecurity in health (14). Future large, well-designed randomized controlled trials will be needed to determine optimal strategies and solidify evidence in support of various food-is-medicine interventions.