Supermarket and Grocery Store–Based Interventions to Promote Healthful Food Choices and Eating Practices: A Systematic Review

Introduction Increasingly high rates of obesity have heightened interest among researchers and practitioners in identifying evidence-based interventions to increase access to healthful foods and beverages. Because most food purchasing decisions are made in food stores, such settings are optimal for interventions aimed at influencing these decisions. The objective of this review was to synthesize the evidence on supermarket and grocery store interventions to promote healthful food choices. Methods We searched PubMed through July 2012 to identify original research articles evaluating supermarket and grocery store interventions that promoted healthful food choices. We categorized each intervention by type of intervention strategy and extracted and summarized data on each intervention. We developed a scoring system for evaluating each intervention and assigned points for study design, effectiveness, reach, and availability of evidence. We averaged points for each intervention category and compared the strength of the evidence for each category. Results We identified 58 articles and characterized 33 interventions. We found 7 strategies used alone or in combination. The most frequently used strategy was the combination of point-of-purchase and promotion and advertising (15 interventions); evidence for this category was scored as sufficient. On average, of 3 points possible, the intervention categories scored 2.6 for study design, 1.1 for effectiveness, 0.3 for reach, and 2 for availability of evidence. Three categories showed sufficient evidence; 4 showed insufficient evidence; none showed strong evidence. Conclusion More rigorous testing of interventions aimed at improving food and beverage choices in food stores, including their effect on diet and health outcomes, is needed.


Introduction
Obesity, overweight, and health outcomes associated with poor nutrition (1) represent a significant economic and social burden in the United States. Annual medical costs attributed directly to obesity and overweight were estimated at $147 billion in 2008 (2). Public health researchers and practitioners are working to identify evidence-based interventions to promote more healthful eating practices. The Dietary Guidelines for Americans 2010 recommend stronger environmental strategies for improving the population's eating practices, including interventions to influence food purchasing behaviors in stores (3).
Supermarkets play an important role in food purchasing (4); consumers averaged 2.2 trips per week to the supermarket in 2010 (5). They also represent an optimal setting for interventions aimed at improving food purchase POP interventions typically entail the use of food demonstrations, taste testing, signs, labels, and other printed materials highlighting healthful food choices or describing recipes with the goal of influencing purchasing decisions toward more healthful options (4). Interventions based on pricing use reduced prices and coupons to promote healthful options (4,10). Interventions based on increased availability work to provide more healthful food choices (4). Promotion and advertising strategies use games, newspaper inserts, multimedia advertising, supermarket tours, and other activities to promote the purchase of more healthful foods (4).
The first author extracted the following data for each intervention: the theory on which the intervention was based (eg, social cognitive), intervention setting, location, year in which an article was published; description of intervention activities and duration; study design (eg, concurrent comparison group, prospective measurement of outcomes); and main outcomes measured.
The main outcomes were awareness and use, sales data, customers' knowledge and beliefs (14), preferences, intentions, and process measures (15-17). Awareness and use refers to the percentage of surveyed customers who noticed the intervention materials and believed their purchasing decision was influenced by them. Sales data refers to objective measurements of customers' purchasing decisions for a category of food or item. Preferences serve as a predictor of target food consumption (18,19). Intentions refer to behavioral intentions to prepare, select, and consume more healthful foods (19). Process measures included reach, dose, and fidelity; reach is the number of target audience members exposed to any component of the intervention (20), dose is the number of times each target audience member was exposed to any intervention component (21), and fidelity is the extent to which an intervention was implemented as planned (15). When interventions reported on fruit and vegetable intake, fat intake, or dietary intake, we also extracted these data.
We created an assessment schema on the basis of accepted terminology in the Guide to Community Preventive Services (Community Guide) (22) and other definitions (20) and categories (15). For each of 3 characteristics -study design, effectiveness, and reach -the first author assigned points to each intervention according to her assessment. Study design was scored as 1, 2, or 3 points, according to the suitability of study design to determine effectiveness (15,22). Greatest suitability (3 points) was defined as a study that had a concurrent comparison group and prospective measurement of outcomes. Moderate suitability (2 points) was defined as retrospective designs or studies that had multiple pre-or postmeasurements but no concurrent comparison group. Least suitability (1 point) was defined as before-after studies that had no comparison group or studies in which outcomes were measured in a single group at the same point in time. We did not assign zero points for study design.
Effectiveness was scored as 0 to 3 points, according to effectiveness of the intervention's main outcome measures (eg, awareness and use). Studies reporting a 70% to 100% increase pre-to posttest or between comparison and intervention groups in outcomes (eg, awareness and use) of the intervention were assigned 3 points. Studies reporting a 26% to 69% increase received 2 points. Studies reporting a 1% to 25% increase received 1 point. A score of 0 was assigned if no difference in outcomes was reported between study groups. Most studies reported effectiveness as awareness or use of their intervention (20). If awareness and use were not reported, we assessed the intervention's main outcome measures (eg, knowledge and beliefs, sales data, preferences, intentions, fruit and vegetable intake, fat intake, dietary intake) and used the same scoring. When an intervention had no effect on awareness or use but had a significant effect on sales data or preferences or intentions, we scored the alternative outcomes.
Reach was also scored as 0 to 3 points. According to the RE-AIM evaluation framework (20), reach is determined by dividing the number of intervention participants by the number of people in the targeted population. For interventions reaching 70% to 100% of the population, 3 points were assigned; for 26% to 69%, 2 points; for 1% to 25%, 1 point; and for 0%, 0 points.
For each intervention category (eg, POP), we calculated the average number of points for study design, effectiveness, and reach. We used this average as 1 of 2 subscores.
For each intervention category, we assessed the availability of data using the following scoring system. We gave 3 points to categories that included 10 to 30 interventions, 2 points to categories that included 2 to 9 interventions, and 1 point to categories used by only 1 intervention. These points represented the second subscore and functioned as an indicator of the amount of evidence available.
We calculated a summary score (range, 0-27) for each intervention category by multiplying the 2 subscores. We created 3 classes of evidence on the basis of the summary score: insufficient (0-9 points), sufficient (10-18 points), and strong (19-27 points). These classes were based on Community Guide designations (22). For all strategies combined, the scores for all 5 categories (ie, availability of evidence, study design suitability, effectiveness, reach, and overall summary) scores were summed and then divided by the number of categories.
Interventions were organized into the following categories (Table 1): POP (n = 6); POP and pricing (n = 1); POP and promotion and advertising (n = 15); POP, pricing, and promotion and advertising (n = 4); POP, increased availability of healthful foods, and promotion and advertising (n = 3); POP, pricing, increased availability of healthful foods, and promotion and advertising (n = 2); and pricing and promotion and advertising (n = 2).

POP and promotion and advertising
The level of evidence for this category (24,25,29,(34)(35)(36)(39)(40)(41)49,50,(53)(54)(55)(56)(57)(62)(63)(64)(65)(66)(67)(68)(69)(70)(71)(72)(73)(74) was sufficient. Only 7 interventions reported objective store sales data (34)(35)(36)40,50,53,57). Among them,5 (34,36,40,50,53,57) showed increased sales of featured items and 1 intervention (35) showed no change. Only 3 interventions (34,49,50) cited a theoretical model as a framework; 1 of these interventions (49,73,74) included a policy component. The intervention "1% Or Less" (57) reported that 90% of people randomly selected for a telephone survey postintervention were aware of the campaign. The 5 A Day program's use of audio communications was implemented as planned during the study period (49) POP, increased availability of healthful foods, and promotion and advertising The level of evidence found for this category (26,28,46,48,76,77) was sufficient. In these interventions, staff worked closely with community members to determine which foods contributed most to total fat and overall calories and identified culturally relevant foods to promote. Researchers also worked with food store owners and managers to stock promoted foods and then advertised these products to consumers, thus simultaneously addressing the supply and the demand sides of healthful eating. Marshall Islands Healthy Stores (28) reported high levels of consumer exposure to the mass media components. Two interventions included self-reported purchasing data for evaluation purposes and reported a positive intervention effect. Two of 3 studies reported moderate to high fidelity; Marshall Islands Healthy Stores reported on logistical difficulties with program written materials. All of these interventions drew on social cognitive theory and included assessments of knowledge, beliefs, and intentions. One intervention (46) relied also on the social marketing framework (51). All interventions targeted low-income or racial/ethnic minority populations and were tailored to the communities in which they were implemented with culturally relevant materials and messaging.
POP, pricing, promotion and advertising, and increased availability of healthful foods The level of evidence for this category (17,37,45,(78)(79)(80) was sufficient. Baltimore Healthy Stores (79) was a feasibility trial and not intended to reach a large number of consumers; however, the intervention reached 5% of the target population. This study collected weekly data on store sales of promoted foods (37); weekly sales of promoted foods increased in intervention stores only when stocking improved. Both programs were implemented with high fidelity. Evaluation of the other intervention included self-reported food purchases of promoted foods (45). The social cognitive theory informed both interventions, and one (45) drew also on the theory of planned behavior, reporting increased caregiver food-related knowledge but not increased intentions to purchase healthful foods.

POP, pricing, and promotion and advertising
Although this category had insufficient evidence (27,30,31,42,75), it had the highest score for effectiveness. One intervention reported on successful nationwide dissemination of supermarket tours (31,75). Two interventions referenced a theoretical model. One (30) was based on the theory of planned behavior and the other (42) on social marketing principles. Two (27,31) interventions included self-report purchasing habits.

Pricing and promotion and advertising
The level of evidence for this category (33,43,81) was insufficient. One intervention's (43) evaluation relied on customer self-report food purchasing data, and the other (33) reported on participant recruitment.

POP and pricing
The level of evidence for this category (38) was insufficient. Self-reported shopping and dietary habits (fruit and vegetable intake) were the primary outcomes, but the authors were unable to systematically compare these data between intervention and comparison stores because of inconsistent monitoring. The intervention was based on the consumer information processing model.

Discussion
The average level of evidence for the interventions summarized in this review nears a sufficient designation but was insufficient overall. Our review generated 4 main findings. First, demand-side interventions (ie, those using POP or promotion and advertising strategies) represented most of the evidence for the study period. Second, evaluation of food store interventions in the early literature emphasized awareness of the interventions, use of the interventions, or both among target populations while excluding other important measures. Store-based strategies evolved to address the supply side, using such strategies as pricing (27,43) and increased availability of healthful foods (28,46,76,37), while continuing to increase demand by using the 2 previous approaches (POP and promotion and advertising). Innovative supply-side interventions were mostly implemented in smaller stores and could be part of a strategy for working in multiple aspects of the food environment. Reporting on process measures such as dose and fidelity is an additional strength of these interventions, allowing for elucidation of the most active components of an intervention. The transition from demand-to supply-side strategies suggests maturation in public health planning and evaluation increasingly based on social and behavioral theoretical models and addresses barriers some communities face when trying to access healthful foods. In communities with limited access to healthful foods, combining culturally sensitive demand-and supply-side strategies is effective in promoting positive food-related behaviors. Third, our review suggests that mass media campaigns accompanying POP interventions (57,(70)(71)(72)) have been effective population-level approaches to influence consumers' decisions on purchasing lower-fat beverages. Finally, the limited use of randomization in food store intervention design reflects the difficulties inherent in applying this design to communitybased health promotion interventions and the greater suitability of quasi-experimental designs.
We found limited evidence on the effect of the interventions on customer purchasing behavior. Eight (34,36,37,40,50,53,57,59) of 13 interventions collecting store sales data demonstrated an increase in targeted product purchases, and 13 interventions presented self-reported data on purchasing behaviors. One of these, a 12-week childfocused intervention (50) yielded a significant increase in the proportion of sales of featured items to total store sales. The intervention displayed fruit, vegetable, and healthful snack samples in a low-to-the-ground kiosk. Similarly successful interventions targeting low-income populations and drawing on social cognitive theory (46) and social marketing (79) were more likely to include components such as taste tests and focus on purchasing of more healthful items, fruit and vegetable acquisition, and food preparation.
Strengths of our study include the up-to-date systematic analysis of 58 articles identified through a comprehensive database and consideration of previous reviews (4,(9)(10)(11)82). In a departure from an earlier review (4), we assessed each intervention by strategy or combination of strategies used, and we developed new categories to describe the simultaneous use of more than 1 strategy.
Our study had several limitations. In an attempt to be comprehensive and include all evidence available to date, we included 7 studies from the late 1970s and the 1980s; data from older studies may not be relevant to today's food environment. Findings for unsuccessful campaigns are less likely to be published in publications searchable in PubMed. We searched for published studies rather than nonpublished reports or gray literature because published articles tend to have more standardized information on setting, study design, evaluation methods, and results. We searched PubMed only. Highly controlled marketing experiments (83-87) may offer additional insights on effective strategies, but we did not include them because of their less direct translation into community-based interventions. Systematically assessing effectiveness was challenging because of the diversity in community-based interventions; many of the studies were conducted in other settings, such as small stores (37) and schools (48). Only 1 reviewer classified and scored the intervention; the classification and scoring were not verified by a second reviewer. In assessing study design, we did not differentiate between studies that used randomization and studies that did not use a control group. The differences in effectiveness suggest our estimates on levels of evidence may overestimate the actual effectiveness of food store interventions, because some of the results we assessed may have reflected baseline differences between treatment and control groups beyond the interventions implemented. Finally, because the availability of evidence was calculated as 1 of 2 subscores, newer intervention strategies will, by default, given our methods, have less research data available. Yet, some of these may offer promise given the quality of the evidence and their significant results (42,46,79).
This review focused on supermarket and grocery store interventions. Food store interventions represent only 1 level of approach among many levels -from the individual to policy (6). Increasingly, public health agencies such as the Centers for Disease Control and Prevention (88) are encouraging local communities to incorporate policy-level approaches to improve access to healthful foods (89). Task forces (90) and state and local food policy councils (91,92) have been proposed as critical elements of such efforts. These organizations leverage public incentives to help obtain financing through such mechanisms as tax exemptions, Community Development Block Grants, state grants such as the Pennsylvania Fresh Food Financing Initiative (93) (upon which Healthy Food Financing Initiative federal efforts have been based), and loans to supermarkets in underserved communities. They also ensure that funded stores participate in state food assistance programs (89,92). State and local government can, among other activities, expedite approval processes to stimulate supermarket development or encourage pedestrian-friendly development to help patrons avoid transportation barriers (89,92). Neighborhood retail analysis, incentives for energy-efficient equipment and systems, and incentives for locally grown products are other policy approaches.
Practitioners need access to up-to-date evidence when approaching grocery and supermarket owners or managers to implement interventions. Food stores want practical strategies that will change consumer behavior (94); they also need a return on investment for increasing access to more healthful foods. Some in-store efforts such as 500 Club and Footsteps to Health (www.getactivelacrosse.org/lacrosse/) complement larger environmental change, and policy initiatives (www.healthinpractice.org/ and SOS Shopping Matters), such as nutrition benefit interventions (eg, Supplemental Nutrition Assistance Program Education enacted by Healthy, Hunger-Free Kids Act of 2010) to emphasize obesity prevention as well as nutrition education and are consistent with the socioecological model (6,7), which posits that multilevel interventions addressing the connections between people and their environments maximize the effect of interventions at each level.
Our systematic review of supermarket and grocery store interventions to promote healthful eating suggests that interventions combining demand-as well as supply-side strategies have sufficient evidence to influence customers and management toward more healthful food purchases. The most effective strategies should be combined, and more rigorous evaluation designs should be used. Recent reports of the relationship between the food environment and health outcomes provide impetus for interventions to target food deserts (95-99) and represent an opportunity to add to evidence (12). Consistent with the socioecological model, public health practitioners are encouraged to use multilevel interventions, including policy and environmental change strategies, and to examine health outcomes during evaluation of these interventions.
Glanz K, Hewitt AM, Rudd J. Consumer behavior and nutrition education: an integrative review. J Nutr Educ 1992;24 (5)     Effectiveness 18% to 43% of shoppers were exposed to intervention activities; no evidence of effect on consumption of fruits and vegetables

Point-of-Purchase and Promotion and Advertising
Special Diet Alert (34,62,63) Setting, location, and year 10 Intervention grocery stores in Washington, DC, and 10 control grocery stores in Baltimore, Maryland, 1992 Activities and duration Brand-specific shelf markers, take-away information booklets, radio and television spots; Effectiveness Improved nutrition knowledge and food use scores among those aware of the campaign; increase in purchase of targeted items Eat for Health (36,55,68,69) Setting, location, and year >100 Intervention supermarkets, Washington, DC; and 30 control supermarkets, Baltimore, Maryland, 1989-1994 Activities and duration Shelf labels; information guide; monthly bulletin; signs in produce sections; television, radio, and print advertising; and brief in-store videos; 24 months Design Prospective measurement with comparison group Main outcomes measured Awareness and use; knowledge and beliefs; fat intake; sales data; process measures Effectiveness Modest effect on food purchasing behaviors; improved nutrition knowledge, attitudes, and self-reported food purchasing behaviors; approximately 200,000 people exposed to campaign  Intentions Effectiveness 23% to 33% of 48 participants intended to alter dietary behavior in targeted direction; posttest purchase of low-fat dairy products, whole grain products, and polyunsaturated spreads was greater than intention to purchase at baseline; posttest purchase of legumes and tofu was less than intention to purchase at baseline 1% Or Less campaign (57,(70)(71)(72) Setting, location, and year 8 Intervention stores in Clarksburg, West Virginia, and Bridgeport, West Virginia, and 6 comparison stores in Wheeling, West Virginia, 1998-2005 Activities and duration Focused message was communicated through paid advertising, public relations activities, and community-based education programs: newspaper, radio, and television advertisements, press conferences, blinded taste tests at supermarkets; signs in dairy case; 2 months No effect of intervention on food purchases reported; mailed recruitment efforts (73% of total recruitment efforts) were more successful than community (20%) or in-store (7%) recruitment efforts Strategies were categorized as point-of-purchase, pricing, promotion and advertising, increased availability of healthful foods, and combinations thereof. Any results reported on reach are included in data on effectiveness. For main outcomes measured, process measures included reach, dose, and fidelity.
Year study was published; if several articles described the same intervention, the range of years of publication is provided.