Abstract Body (Do not enter title and authors here): Background: The US population eats less than recommended amounts of fruits and vegetables (F&V) despite their health benefits. This randomized clinical trial tested whether healthy food subsidies, produce placement (choice architecture), and/or loss-framed text message reminders about subsidy balances (salience reminders) increased F&V purchases in an online grocery store over 3 months.
Methods: Participants aged 30–70 years with obesity and diabetes, and online shopping experience were randomized to one of five arms: 1) control ($20 monthly coupon); 2) subsidy ($160 monthly for F&V); 3) subsidy+salience (weekly text messages loss-framed about subsidy balance); 4) subsidy+choice architecture (F&V appear first on storefront); or 5) subsidy+salience+choice architecture. Post-randomization they received an invitation to activate their online accounts. Total dollars spent on F&V via this online account was the primary outcome. Intention-to-treat analyses (ITT; n=210) included all randomized participants; “per protocol” analyses included those who activated their accounts (n=187). The analysis used linear mixed effects models.
Results: Participants were 61% female, 75% Black, with mean BMI of 38.0 (SD: 7.4) kg/m². Under ITT, the average monthly spend was 1) control: $16.2 (SD=$27.1), 2) subsidy alone: $104.1 (SD=$67.3), 3) subsidy+salience: $120.0 (SD=$60.4), 4) subsidy+choice architecture: $94.4 (SD=$65.2), and 5) subsidy+salience+choice architecture: $106.3 (SD=$66.4). Under ITT, the subsidy increased mean dollars spent on F&V by $93.5 (95% CI=$73.7, $113.3, p<.001) but among arms with the subsidy, there were no statistically significant differences (p>.05). Among those who activated their online accounts, the mean amount spent on F&V was $128.3 (SD=$51.5) vs. $106.7 (SD=$$62.7) for the salience/non-salience arms, an increase of $19.0 (95% CI=$3.7, $34.4, p=.015). Salience messages appeared to be effective by encouraging individuals to use their online accounts: in arms without salience, participants made no purchases on their accounts in 20% of months versus 9% for arms with salience. In contrast, choice architecture vs. no choice architecture groups had similar spending on F&V (difference of -$11.3, 95%CI=$-26.9, $4.2, p=0.15).
Conclusion: Providing a subsidy substantially increased F&V purchasing. Once participants activated accounts, choice architecture did not result in noticeable improvement whereas reminders of the subsidy balance increased F&V spending.