SUPERSTARS: Supermarket Support for a Primary Care Healthy Food Prescription

Project Goal and Specific Aims The Project goal is to implement and rigorously evaluate an innovative healthy food prescription program, Superstars, among low-income rural primary care patients, paired with an incentive to use a supermarket shelf-tag-labeling system that supports healthier choices at the point of purchase. Research findings will be used to engage health systems, nutrition educators, and consumers to improve nutrition security, population health, and health equity.

Research Aim 1: Assess the program's impact on participants' nutrition security, as measured with the validated 4-item Gretchen Swanson screening tool.

The investigators hypothesize that participants in the intervention arm will have greater improvement in their nutrition security than those in the control arm at the end of the 4-month intervention.

Research Aim 2: Assess the program's impact on participants' supermarket purchases and diet, and explore the program's impact on health using the MaineHealth practice sites' electronic health record.

The investigators hypothesize that participants in the intervention arm will have greater improvements in healthy supermarket purchases and in overall dietary quality than those in the control arm at the end of the intervention.

The investigators hypothesize that there will be a small improvement in blood pressure levels among intervention participants compared to control participants after the intervention.

Education and Extension Aim 3: Use the research findings to encourage health systems, nutrition educators and consumers to adopt evidence-based strategies that improve nutrition security, population health, and health equity.

The investigators will accomplish the project Aims with the guidance of a stakeholder advisory committee which will be convened, virtually, three times (yearly) beginning month 3 of grant funding. The investigators will offer each committee member a one-time $300 stipend for their participation. Advisors will include representatives from each of the following sectors: healthcare system, supermarket retail, nutrition education, nutrition policy, and patients meeting study criteria (representing end users). The advisory committee will review project plans, provide input on project activities before they are finalized, problem-solve issues that arise during implementation, help interpret study findings, and advise during implementation of education and extension activities and products.

1. Methods Aims 1 and 2 Study Design: The investigators will use a randomized controlled study design to test the effectiveness of the Superstars intervention among 500 low-income MaineHealth primary care patients who do most of their shopping at Hannaford Supermarket.

Setting and Target population: At least 5 MaineHealth practice sites in midcoast Maine have been identified for participation. Each of the sites is within 7 miles of a Hannaford Supermarket. Given the sparse number of supermarkets in midcoast Maine, the presence of a Hannaford store within a few miles of the practice site provides ample opportunity for the store to be the primary shopping venue for the patient. The Hannaford supermarkets all use the Guiding Star shelf-tag labels and will redeem the primary care healthy food prescriptions that encourage purchasing the healthy 2- and 3-starred foods and beverages. All participants will be made aware of other Hannaford wellness initiatives including online resources ("MyHannaford.com"), on-site registered dietitians, and signage to highlight healthful and affordable recipes, including Close to Home, Fresh Magazine, and School Tour Program.

Recruitment and enrollment: Patients receiving care from participating MaineHealth primary care practices will be screened for insurance status eligibility. In addition, practice teams will be trained by study staff to discuss the program with eligible patients to boost enrollment at the time of office visits. To be eligible, patients must be 21 years or older and either be insured with Mainecare (Maine Medicaid) or be provided MaineHealth Assisted Care. Patients who meet the age and insurance eligibility criteria will be either provided a flyer about the program, called, or contacted by mail, or sent the equivalent information about the program through the mail by project staff. If a patient indicates they are interested in participating in the study, they will be contacted by study staff who will assess remaining eligibility criteria: 1) Primary household shopper, 2) Hannaford as the primary (> 50% of grocery purchases) weekly shopping venue and 3) participating in the loyalty program for at least 6 months, 4) having 2 or more visits recorded in the EHR in the past 2 years.

For eligible patients, study staff will obtain informed consent. Participants will then be sent a link to compete an enrollment (baseline) survey that includes questions about sociodemographic characteristics, outcome measures (nutrition security, food security) and other measures such as knowledge, attitudes, beliefs and skills related to nutrition, healthful shopping, food preparation, and depression. In addition, participants will be asked if they provide permission to access their MaineHealth electronic health record for blood pressure (BP) and body mass index (BMI).

An exit survey, administered immediately post intervention, will include all questions on the enrollment survey except basic demographics. For intervention group participants, the exit survey will also include questions related to satisfaction with the intervention and any barriers to using the program that may have been experienced.

The investigators expect enrollment of 500 patients to take between 2 and 4 months and at least 95% (N=475) to actively use the loyalty program based on our experience conducting research in similar settings.

Randomization: The investigators will use a randomized controlled study design to conduct the study with 500 participants from at least 5 MaineHealth primary practice sites in midcoast Maine. Once enrolled, half of the participants will be randomly assigned to receive the 4-month intervention while the other half will receive regular care.

All participants will receive 1) a 5% discount on purchases from Hannaford Supermarkets if they use their loyalty program at check-out or for use shopping online through Hannaford.com (to allow purchase tracking) during the entire 4-month study period, and 2) information about how to access Hannaford Supermarkets nutrition resources available through the Hannaford Supermarkets website. These resources include recipes, and nutrition and cooking classes. Hannaford Supermarkets also have dieticians on staff as a resource for customers.

Intervention Arm: In addition to the above discount and information, those in in the intervention arm will further receive a $40 monthly discount to purchase 2- and 3-star products in any Hannaford Supermarket. Once enrolled, the primary care practice will send a healthy food prescription to intervention participants via email and/or text message. The prescription will focus on dietary guidance based on MyPlate and the Dietary Guidelines and will explain how participants can use the Guiding Stars to help them improve their diet and choose the incentivized items.

The electronic discounts will be applied automatically at checkout or may be used shopping online through Hannaford.com when the loyalty program is used (shopper enters cell phone number or scans loyalty card at checkout). Participants will be sent reminders (via email or text, as preferred) by study staff to use their monthly 2- and 3-star discount.

Control Arm: Control arm participants will be provided regular care in addition to the 5% discount on all purchases from Hannaford Supermarkets each time they use their loyalty program at check-out and nutrition information, as described above.

Aims 1\&2 Measures

Outcome Measures:

Nutrition Security (Aim 1-Primary Outcome): The validated Gretchen Swanson 4-question Nutrition Security measure will be included on both enrollment and exit surveys. Response categories to all 4 questions range from 0 (Always) to 4 (Never). The measure's score is the mean of the 4 responses. A higher score indicates a greater degree of Household Nutrition Security, which means the household is able to acquire healthful foods without resource limitations or worry. "Low" scores are 2.00 or below. The investigators will use a continuous mean score (between 0.00 and 4.00), with the lower the score indicating more food insecure.

Food Security (Aim 1- Secondary Outcome): The investigators will use the validated 6-question USDA Household Food Security Module to measure food security on both enrollment and exit surveys. The exit survey will be modified to reflect food security over the 4 months of the intervention rather than the past year. The investigators will use participants' raw scores on the 6-question survey to classify participants as having high (0-1), low (score of 2-4), or very low (score of 5-6) food security. The enrollment and exit surveys will be programmed into REDCap, and a link to complete the surveys will be sent to those who agree to participate in the study. Study staff will communicate with participants twice weekly to remind them to complete the surveys for 3 weeks or until the surveys are completed.

Dollars spent per shopping month on 2- and 3-star foods (Aim 2- Primary Purchase Outcome): The investigators have acquired monthly Guiding Stars datasets on all products sold at Hannaford since March 2016 and will continue to receive these monthly datasets through the end of our follow-up period. The investigators will merge the sales data with Guiding Stars nutrition data on product UPC over the entire project period. Guiding Stars captures all nutrition information on products sold in stores where its labeling system is used, including Hannaford, and uses an algorithm to rate the nutritional quality of these foods; healthier foods receive 1, 2, or 3 stars (good-better-best) while less healthy foods receive 0 stars. The Guiding Stars data include products' UPCs, all major macronutrients (e.g., saturated fat, sugar) and micronutrients (e.g., vitamin D, iron), product serving size, and star rating. These data are updated monthly to include products newly brought to market, remove discontinued products, and re-rate products that have been reformulated. In an analysis of Hannaford sales from 2018-2022, the investigators determined that 31% of items sold were 2- or 3-star foods. The investigators will examine changes in total purchases (servings and dollars) of 2- and 3-starred foods.

Dollars spent on specific food group purchases (Aim 2- Secondary Purchase Outcome): The investigators have developed a classification system that groups each product into 1 of 34 mutually exclusive food groups.34 The investigators will examine changes in total purchases (servings and dollars) of several healthier groups, including fresh, frozen and canned fruits, vegetables, beans, and nuts. The investigators will also examine changes in purchases of unhealthy groups that tend to be calorie-dense, convenient, and less expensive than healthier options like fresh produce and nuts. These include sugary drinks, packaged "convenience" foods (e.g., macaroni and cheese, chicken nuggets), processed meats (e.g., deli meat), pizza, sweet/salty snacks (e.g., chips), baked goods (e.g., cookies, cakes), frozen desserts (e.g., ice cream), and candy.

Dietary Intake (Aim 2-Secondary Outcome): The Automated Self-Administered 24-hour (ASA24) Dietary Assessment Tool will be used to calculate Healthy Eating Index (HEI-2015) overall and component scores. Scores range from 0 (least healthy) to 100 (healthiest). The ASA24, a web-based tool developed by the National Cancer Institute, enables automatically coded, self-administered 24-hour recalls. Respondents are guided through the recall using a modified version of the USDA Automated Multiple-Pass Method. A link to the ASA24 will be sent to the participant electronically on 2 random days (1 weekday and 1 weekend) to take the survey online. Participants will be asked to complete these within 3 weeks of receiving the link in order to receive the incentives. Reminder emails will be sent twice weekly for 3 weeks or until the recalls are completed.

Health (Aim 2- Exploratory Outcomes): Body Mass Index (BMI, kg/m2) and blood pressure (BP, mm Hg) measures will be obtained from the EHR for each participant at baseline and follow-up. Baseline measures will be the last BP or BMI recorded within 12 months before enrollment in the study, and follow-up measures will be the first BP or BMI recorded within 12 months after the 4-month study period. Changes in systolic and diastolic BP and in BMI will be assessed among participants who provide permission to access the health record. As an alternative method to assess BMI, participants will be asked to provide their height and weight on the enrollment survey and on the exit survey.

Other measures (Aims 1 and 2 covariates): Demographics (household size, number of children in the household, age, race/ethnicity, sex, and household income), prior participation in other nutrition education programs, and current participation in SNAP and other assistance programs, constructs related to healthy food utilization (knowledge, attitudes, beliefs and skills related to nutrition, healthful shopping, food storage, food preparation tools and food preparation). Questions measuring these constructs will be added to both enrollment and exit surveys with the exception of demographics which will only be asked at enrollment.

Aims 1 \& 2 Analysis and Interpretation

Nutrition Security (Aim 1-Primary Outcome): The investigators will use a linear regression model and a difference-in-differences (DD) approach to estimate the effect of the healthy food prescription on nutrition security score for the intervention vs. control group. All analyses will be intent-to-treat based on randomization. The investigators will descriptively present baseline demographic and health-related characteristics separately for each trial arm and will control for these characteristics as needed to correct for any chance imbalances between arms. The investigators will use a model of the form:

Nutrition security score = β0 + β1Prescription + β2Post + β3Prescription×Post In this model, Prescription is an indicator for being in the prescription group and Post is an indicator for observations after the intervention was implemented. β0 represents the mean nutrition security score for those in the control group at baseline, β1 represents the difference in baseline score for the prescription group vs. control group, β2 represents the pre-post change in the score for the control group and β3 represents the difference in score for those in the prescription group above and beyond the change for the control group and is therefore the main effect of interest. The investigators will have 80% power to detect a 0.25 increment change in score.

Food Security (Aim 1- Secondary Outcome): The investigators will use the same approach to estimate effects of the prescription intervention on food security score. The investigators anticipate similar power to detect changes in the food security score as the nutrition security score.

The investigators will compare baseline food insecurity to baseline nutrition insecurity rates among participants to assess this measure's usefulness to provide clinical decision support. If nutrition security is found effective, the single question clinical nutrition screener (question N3) could be recommended for inclusion in future clinical screening.

Dollars spent on 2- and 3-star foods (Aim 2- Primary Outcome): The investigators will also use a DD approach to estimate the effect of the prescription intervention on probability of purchasing a 2- or 3-star item. The investigators will compare changes in purchases from the 6 months before the intervention to the 4-month intervention period and 6-month period following the intervention. The analysis will therefore include up to 16 data points per participant. The investigators will use generalized estimating equations (GEE) with a log link and Poisson distribution, clustering on participant and estimating robust standard errors. The model will be:

Pr(Y=1) = β0 + β1Prescription + β2Post + β3Prescription×Post + β4Follow-up + β5Prescription×Follow-up In this model, Follow-up is an indicator for the 6-month period after the intervention ends. β0-β3 are interpreted similar as in Aim 1; β3 represents the pre-post difference in probability of purchasing a 2- or 3-star item for the prescription group above and beyond the pre-post difference for the control group and is the main effect of interest. β4 represents the difference in probability for the control group 6 months after the intervention ends and β5 represents the difference in probability for the prescription group in this 6-month period above and beyond the change for the control group. If β5 > 0, it suggests that those in the intervention group had improved purchases of healthy foods even after the intervention ended.

Assuming that in the proposed study the control group has a 30% probability of purchasing a 2- or 3-star item and that the within-shopper correlation is 0.2, the investigators would have 83% power to detect a 22% increase in probability of purchasing a 2- or 3-star item (equivalent to the intervention group's probability of purchasing a 2- or 3-star item increasing from 30% to 36.5%).

Food group purchases (Aim 2- Secondary Outcomes): The investigators will use a similar approach to estimate changes to secondary purchase outcomes except the GEE will use an identity link and normal distribution to estimate mean changes food group purchases.

Dietary Intake (Aim 2-Secondary Outcome): The investigators will compare dietary intake within individuals (pre- to post-intervention), and between individuals (intervention and control arms) using HEI-2015 scores. The HEI-2015 will be calculated based on 2 ASA24 dietary recalls completed at baseline and immediately post intervention. If a participant does not complete a second ASA24, the HEI-2015 will be calculated based on 1 ASA24. The HEI-2015 measures overall dietary quality, consistent with US Department of Agriculture guidelines, and scores range from 0 (least healthy) to 100 (healthiest). The mean score for US adults is 59. The investigators will estimate the effect of the healthy prescription intervention on HEI-2015 scores using a linear regression model of the same form as in Aim 1 (i.e., β3 will estimate the pre-post change in mean HEI-2015 score for the prescription group above and beyond that of the control group). With n=250 in each group and a SD in each group of 16.2, the investigators should have 87% power to detect a 4.5-point change in HEI-2015 score.

BP and BMI (Aim 2- Exploratory Outcomes): The investigators will conduct similar DD analyses using linear regression models to compare changes in BP and BMI using most recent data available from the EHR within 12 months before study enrollment and 12 months post intervention. The investigators will explore the effect of the healthy prescription intervention on changes in systolic BP, diastolic BP, and BMI in the same manner as for other outcomes.

Other Measures (Aims 1 and 2 covariates): Answers to questions measuring healthy food utilization constructs (knowledge, attitudes, beliefs and skills related to nutrition, healthful shopping, food storage, food preparation tools and food preparation) will be used to adjust for these variables if the investigators detect chance differences between study groups (i.e., if randomization doesn't work perfectly). They may also be used to detect "spillover effects". For example, if the investigators see changes in healthy purchases, the investigators could explore the additional benefit of improving nutrition attitudes. Further, if our results do not demonstrate improvements in nutrition security or purchases, the investigators will be able to explore whether other measures may have contributed.

Process measures: The investigators will measure program implementation and process by tracking all activities related to identifying eligible patient populations, delivering the intervention, and using the incentive. Outcomes such as purchases and diet are influenced by nutrition-related knowledge, attitudes, beliefs and skills which will be measured at baseline and immediately post-intervention. Satisfaction with the intervention and the coupon redemption process will also be explored in the exit survey to help us understand how these project components may have influenced outcomes and to help refine recommendations for future implementation. Issues with 2-and 3-star voucher redemption at the supermarket could also influence outcomes.

The investigators will conduct at least 5 qualitative key informant interviews with Hannaford Supermarket staff (e.g., assistant manager or cashier), including 1 employee from each of the stores in communities linked to the healthcare sites. Trained study staff will schedule and conduct these interviews over Zoom during year 2 of the grant while the intervention is ongoing. The interviews will explore the process of providing the healthy food incentives and tracking participants for the purpose of the study in order to understand any barriers encountered by participant shoppers or by the retailer. Zoom calls will be recorded, transcribed, and analyzed using a rigorous thematic content analysis approach to enrich and expand our quantitative findings. For example, these interviews may provide a deeper understanding of how the store process worked and may yield detailed suggestions for specific process improvements for future implementation.

Data collection and tracking systems will be put in place to capture process data which will be used to 1) make mid-term corrections on program implementation to ensure success; and 2) help explain outcome measures of success. These process measures, incorporated into each project component, will be used to assess fidelity of planned program implementation and assure the project stays on track for the duration of funding.

The Framework for Program Evaluation in Public Health will guide the evaluation process through the outlined 6 steps: 1) engage key stakeholders along the way, 2) describe the program, 3) focus the evaluation design, 4) gather credible evidence, 5) justify conclusions, and 6) ensure use and share lessons learned. Program staff will monitor evaluation data in an ongoing way at regular project meetings. The investigators will also schedule yearly meetings (and additional ad-hoc meetings, as needed) with the stakeholder advisory committee to interpret evaluation data and advise on making adjustments as needed. The analysis plan further elaborates how process evaluation will help to explain potential differences in research outcomes.