Impact of a Farmers' Market Nutrition Coupon Program on Diet Quality and Psychosocial Well-being Among Low-income Adults: A Randomized Controlled Trial

Background: The British Columbia (BC) Farmers' Market Nutrition Coupon Program (FMNCP) is a healthy eating initiative funded by the BC Ministry of Health. The program provides low-income households with $21/week in coupons to purchase healthy foods at farmers' markets and supportive nutrition skill-building activities. It is not clear if the BC FMNCP is achieving its aims of improving the diet quality, well-being and health of low-income adults, as its outcomes have not been investigated, and studies of similar programs are limited by weak designs that cannot demonstrate causality. This randomized controlled trial will assess the impact of the FMNCP on the overall diet quality (primary outcome), diet quality subscores, mental well-being, sense of community, experiences of food insecurity, risk of malnutrition (secondary outcomes) and subjective social status (exploratory outcome) of low-income adults immediately post-intervention and 16 weeks post-intervention. A qualitative investigation will also explore mechanisms of action and strategies to maximize positive program impacts.

Methods: Low-income adults (≥ 18 years) from up to 15 rural and urban communities will be randomized to a FMNCP intervention (n=132) or a no-intervention control group (n=132), with a 1:1 allocation ratio. An independent researcher from the Clinical Research Unit at the University of Calgary will generate a blocked randomization sequence that stratifies participants into blocks according to sex (male, female), geographic location (rural, urban), pregnancy (yes, no) and breastfeeding (yes, no). In the existing BC FMNCP, community partners distribute one to two sheets of coupons per week (each sheet contains $21 in coupons) to program participants for a total of 16 sheets. Coupons can be used over 16-20 weeks to purchase fruits, vegetables, dairy, meat/poultry/fish, eggs, nuts, and cut herbs at participating BC farmers' markets. However, to allow sufficient time to recruit participants for this study, community partners will distribute 16 coupon sheets to the FMNCP group over 10-15 weeks (households with 5-8 individuals will receive 32 coupon sheets). To ensure participants receive all 16 coupon sheets, community partners will provide two coupon sheets per household during the first 1-6 weeks of the intervention. Participants in the FMNCP group will also be invited to participate in nutrition skill-building activities (e.g., cooking classes). At baseline (0 weeks), immediately post-intervention (10-15 weeks) and 16 weeks post-intervention (26-31 weeks), participants will access a pilot-tested web-based platform to report sociodemographics, health-related variables, mental well-being, sense of community, experiences of food insecurity, risk of malnutrition and subjective social status. Dietary intake will be assessed via two 24-hour dietary recalls at each time point using the Automated Self-Administered Dietary Assessment Tool for Canada (ASA24-Canada-2018). Diet quality scores will be calculated using the Healthy Eating Index-2015. In addition, immediately post-intervention only, participants will report whether they received FMNCP coupons and attended nutrition skill-building activities (to assess contamination of the control group), how often and how much of their own money was spent at farmers' markets during the intervention period and the types of foods purchased.

Data analysis: Descriptive analyses will be conducted to examine participant characteristics by group at each time point. Characteristics of study completers (i.e., provided data 16-weeks post-intervention) and non-completers will also be compared.

Analyses will be intention-to-treat, in which participants will be analysed within the groups to which they were randomized regardless of adherence (e.g., failure to redeem coupons) or dropout. The analyses will include all participants who provided data at baseline. Repeated measures mixed-effect regression will assess differences in mean HEI-2015 scores, HEI-2015 subscores, mental well-being, sense of community, and subjective social status between the FMNCP and control groups immediately post-intervention and 16-weeks post-intervention. Repeated measures multinomial logistic regression will be used to assess differences in the odds of experiencing household food insecurity and risk of malnutrition for the FMNCP group compared to the control group immediately post-intervention and 16-weeks post-intervention. Statistical models will include intervention group (FMNCP vs control), time from baseline, intervention-by-time interaction, blocking variables (i.e., sex, rural/urban, pregnancy, breastfeeding), baseline values of the outcome, and household size as fixed effects covariates. Participant-specific (i.e., repeated measures) variations in outcomes will be modeled using random effects. Models will also include covariates specific to each outcome to increase the precision of estimates (86). For the primary outcome of overall diet quality, models will include the following: children living in the home (yes, no), sex, age, BMI, marital status, race/ethnicity, perceived health, smoking, day of data collection, and educational level. Adjusted group differences (i.e., FMNCP group vs control group) in outcomes will be estimated using 95% confidence intervals and corresponding p-values.

Subgroup analyses will examine whether the impact of the intervention on primary and secondary outcomes differs according to age group or sex. Dose-response analyses will examine whether the impact of the BC FMNCP on overall diet quality depends on the number of coupons redeemed and the number of nutrition skill-building activities attended. Interactions will be retained in statistical models if p<0.10. Analyses will be conducted in Stata (v15.1, Stata Corp, TX, USA), with p<0.05 indicating statistically significant differences between groups. Data interpretation will jointly consider effect sizes, confidence intervals and statistical significance.