Bigger Produce Prescriptions Drive Real Gains in Food Security, Trial Shows

In the rolling agricultural heartland of California’s Central Valley, where the fields produce much of the nation’s fresh food even as the people living beside them struggle to afford it, a deceptively simple question has just been answered with unusual rigor: how much money does it take for a produce prescription to actually work? A randomized community trial published in the Journal of General Internal Medicine has found that the answer matters far more than most program designers have assumed. When researchers from Tufts University, the University of California San Francisco, and the University of North Carolina at Chapel Hill gave community members monthly produce prescription benefits of $40, $80, or $110 for six months, they observed a clear dose-response relationship: the larger the benefit, the more benefits people redeemed, the more food-secure they became, and the more fruits and vegetables they ate. The finding, led by Ryan M. Kane of Tufts’ Food is Medicine Institute together with Hilary K. Seligman and Seth A. Berkowitz, is the first head-to-head comparison of multiple produce prescription doses, and it arrives at a moment when insurers, health systems, and Medicaid programs across the United States are racing to scale food is medicine interventions without solid evidence about how to size them.

The stakes of this technical question are enormous. Poor nutrition and food insecurity are estimated to account for $1.1 trillion in annual healthcare spending and lost productivity in the United States, a figure that reflects the deep entanglement of diet with chronic diseases ranging from diabetes to cardiovascular disease. Produce prescriptions, often abbreviated PRx, are a flagship intervention of the food is medicine movement: they subsidize the purchase of fruits and vegetables for individuals facing food insecurity, poor nutrition, or diet-related disease, with the goal of lowering financial barriers to healthy food and making the healthy choice the default choice. Yet despite growing enthusiasm, the field has been haunted by an unstandardized implementation question. The dose, or monetary amount of the benefit, has typically been set by budgetary constraints rather than by evidence, leaving researchers uncertain how the amount of money provided relates to the amount of benefits redeemed, to food security, and to actual fruit and vegetable consumption—the key process outcomes on the pathway toward improved health.

To bring standardization to dose selection, the research team turned to an unlikely but policy-relevant guide: the USDA’s Thrifty Food Plan, the same federal benchmark used to calculate food assistance benefits. The Thrifty Food Plan estimates the cost of consuming the recommended amount of fruits and vegetables, and it can be adjusted over time for food price inflation and scaled to household size, making it an evidence-based anchor for prescription amounts. In this trial, the three doses—$40, $80, and $110 monthly—were defined as 10%, 20%, and 30% of the Thrifty Food Plan’s four-person household fruit and vegetable budget. The study was conducted in Stockton, California, in partnership with El Concilio, one of the largest community-based nonprofit social service providers in the Central Valley, whose staff guided participant engagement, recruitment, reminders, and follow-up. The produce prescription cards themselves were provided by Vouchers 4 Veggies, a San Francisco-based organization established in 2015 that has run citywide produce prescription programs for a decade.

The trial design was a single-site, unblinded, open-label, three-arm, prospective randomized community-partnered study. After completing a baseline survey, 242 adult clients of El Concilio were randomized 1:1:1 to receive $40, $80, or $110 in monthly electronic benefits for six months, loaded onto a reusable card mailed to each participant. The benefits could purchase fresh fruits and vegetables, in person or online, only at major area retailers with groceries—Walmart, Safeway, Dollar General, Family Dollar, and Instacart among them—and did not roll over from month to month. To ensure equity in the total benefit each participant received, the researchers added a graded bonus payment at study completion: $420 for the $40 group, $180 for the $80 group, and $0 for the $110 group, bringing everyone’s total to $660. The study population was predominantly female (97.1%) and Hispanic (82.6%), with a mean age of about 35 years, and carried a heavy burden of hardship: 64.9% reported food insecurity and 49.8% reported fair or poor self-rated health, far above the general US population rates of 13.5% and 14.8% respectively.

The primary outcome was the benefit redemption rate—the percentage of available benefits actually spent—supplied directly by the card provider. Secondary outcomes were validated survey measures: the 18-item household Food Security Survey Module for food insecurity and the National Cancer Institute’s Dietary Screener Questionnaire for fruit and vegetable intake. Surveys were administered in English or Spanish at baseline, three months (the primary endpoint), and six months. Notably, the researchers did not require participants to report food insecurity as an eligibility criterion, on the reasoning that produce prescriptions may benefit individuals whether or not they report food insecurity. The statistical analysis followed the intention-to-treat principle and confronted a practical challenge: loss to follow-up was greater than anticipated, with survey response rates of 71% at three months and 70% at six months. To address this, the team used targeted minimum loss-based estimation, or TMLE, a modern causal inference method that combines an outcome model with a model of the joint probability of exposure and remaining uncensored—elements similar to g-computation, propensity score methods, and inverse probability of censoring weighting. In randomized trials, TMLE has been shown to offer substantial efficiency gains while remaining unbiased, and by accounting for censoring it can handle potentially differential loss to follow-up that might otherwise create post-randomization bias.

The results, adjusted for age, race, ethnicity, income, household size, and baseline outcome values, revealed a dose-response pattern across all three key outcomes. At three months, the benefit redemption rate was 60.2% (95% CI 51.9–68.6%) in the $40 group, 67.6% (95% CI 59.9–75.3%) in the $80 group, and 70.4% (95% CI 63.0–77.9%) in the $110 group. Food insecurity told an even starker story: 74.3% (95% CI 62.0–86.6%) of the $40 group reported food insecurity at three months, compared with 66.2% (95% CI 50.1–82.2%) of the $80 group and just 47.4% (95% CI 31.6–63.1%) of the $110 group. Mean fruit and vegetable consumption, excluding fries, climbed from 2.51 cups per day (95% CI 2.31–2.72) in the $40 group to 2.66 (95% CI 2.38–2.95) in the $80 group and 2.89 (95% CI 2.61–3.17) in the $110 group. When the researchers compared the $110 dose with the $40 dose, the differences were statistically significant for food insecurity (p = 0.01) and fruit and vegetable intake (p = 0.03) at three months, and for the benefit redemption rate (p = 0.04) at six months. Comparisons between adjacent doses—the $80 versus the $40, or the $110 versus the $80—often did not reach statistical significance, a pattern the researchers interpreted through the lens of a trial designed to detect signals rather than provide definitive testing between every pair of arms.

Perhaps the most surprising finding concerned the researchers’ own hypothesis. They had expected that the $40 group would show the highest redemption rate, reasoning that a smaller benefit would be easier to spend fully, and that a ceiling effect might emerge as dose increased. The data pointed the opposite way: redemption rate was, if anything, lower for the $40 group, suggesting that a $40 monthly incentive may be insufficient to motivate the time-cost of utilization—the effort of traveling to a store and shopping for produce against the reward of a modest subsidy. Meanwhile, redemption rates were similar in the $80 and $110 groups, but the absolute dollar amounts redeemed were higher for the $110 group: over the first three months, participants in the $40 group redeemed a mean of $72.79 (95% CI $62.26–$83.32), the $80 group $162.61 (95% CI $144.02–$181.20), and the $110 group $232.79 (95% CI $208.65–$256.93), with all pairwise comparisons statistically significant at p < 0.001. This suggests that people continue to redeem a similar percentage of benefits even as the dollar amount rises, without evidence of a ceiling effect for the amount of produce purchased among the doses studied.

The team went beyond the three fixed doses to estimate exploratory dose-response curves, exploiting variation in household size across the randomized groups. They converted each household’s randomly assigned benefit into an equivalized amount, accounting for economies of scale by dividing by the square root of household size—a four-person household with a $40 benefit, for example, would provide $20 equivalized dollars. Treated in $10 increments, the equivalized dose showed a dose-response relationship with benefit redemption rates, and there was no apparent plateauing at higher doses. Additional outcomes, including ease of following a healthy diet, self-rated health, use of CalFresh (SNAP), and the CDC’s Measuring Healthy Days module of self-reported healthy days, did not differ across groups—a result the researchers considered unsurprising given that this was a community-based trial among relatively young participants without a diet-related disease focus or a structured nutrition program. Subgroup analyses limited to individuals with food insecurity were very similar to the overall trial, and no harms or adverse events were reported.

The trial adds to a sparse randomized evidence base. A recent systematic review of food is medicine randomized controlled trials, published as a scientific statement from the American Heart Association, identified only three early-stage produce prescription trials, with sample sizes of 78 to 128 participants. Two focused on patients with diabetes and provided very low doses—a $6 one-time benefit and $10 weekly for about two months—while the third, the largest and longest, enrolled community health center patients with overweight or obesity and delivered a weekly community-supported agriculture box equivalent to a $300 dose over 24 weeks, significantly increasing Healthy Eating Index scores by 4.3 points and lowering food insecurity. The new trial’s standardized Thrifty Food Plan approach to dose selection permits equivalent future doses to be set for any given level of food prices, adjusted for inflation and local cost of living—a technical feature that could prove as influential as the dose-response finding itself. The strong community partnership with El Concilio enabled rapid enrollment of a predominantly Hispanic sample with a high percentage of Spanish-speaking participants (40%), demographic groups often poorly represented in clinical trials.

The study’s limitations are acknowledged candidly by its authors. As the first trial to compare multiple produce prescription doses head-to-head, there was minimal prior data to guide dose selection and power calculations, the sample size was relatively small, and attrition introduced uncertainty despite the robust statistical approaches. Whether the results generalize beyond predominantly female, Hispanic, young adults with high burdens of food insecurity and fair or poor health remains unknown. Still, the implications for policy are immediate. As insurers, health systems, and Medicaid Section 1115 demonstrations seek to implement produce prescription programs to improve patient health outcomes, the critical design element of dose can no longer be an afterthought set by budgetary convenience. The evidence now suggests that a $110 monthly dose—30% of the Thrifty Food Plan’s fruit and vegetable budget—can cut food insecurity nearly in half relative to a $40 dose in a heavily burdened community, while driving greater redemption and consumption. For a movement betting that food can be medicine, the prescription is becoming clearer: the strength of the dose may determine the strength of the cure.

Subject of Research: Produce Prescription Dose Response: A Randomized Community Trial

Article Title: Produce Prescription Dose Response: A Randomized Community Trial

Article References: Kane, R. M., Perez-Velazco, X., Levi, R., Goldberg, A., Kuhn, M., Mehri, N., Seligman, H. K., & Berkowitz, S. A. (2026). Produce Prescription Dose Response: A Randomized Community Trial. Journal of General Internal Medicine. https://doi.org/10.1007/s11606-026-10773-0

Image Credits: AI Generated

DOI: 10.1007/s11606-026-10773-0

Keywords: Produce, Prescription, Dose, Response, Randomized, Community, Trial, scientific research