The healthy food marketing strategies study: design, baseline characteristics, and supermarket compliance

Abstract

Identifying effective strategies to promote healthy eating and reduce obesity is a priority in the USA, especially among low-income and minority groups, who often have less access to healthy food and higher rates of obesity. Efforts to improve food access have led to more supermarkets in low-income, ethnically diverse neighborhoods. However, this alone may not be enough to reduce food insecurity and improve residents’ diet quality and health. This paper summarizes the design, methods, baseline findings, and supermarket in-store marketing strategy compliance for a randomized trial of the impact of healthy food marketing on the purchase of healthier “target” food items. Thirty-three supermarkets in low-income, high-minority neighborhoods in the metropolitan Philadelphia area were matched on store size and percentage of sales from government food assistance programs and randomly assigned to the intervention or control group. Healthy marketing strategies, including increased availability of healthier “target” products, prime shelf-placement and call-out promotion signs, and reduced availability of regular “comparison” products, were implemented in 16 intervention stores for an 18 month period for over 100 individual food items. Six product categories were studied: bread, checkout cooler beverages, cheese, frozen dinners, milk, and salty snacks. The primary outcome measure was weekly sales per store in each product category for 1 year preintervention and 18 months during the intervention. Compliance with the marketing strategies was assessed twice per month for the first 6 months and once a month thereafter. Store and neighborhood characteristics were not significantly different between control and intervention stores. Intercept surveys with customers to assess shopping habits and grocery marketing environment assessments to examine the food promotion environment were completed in the same six food categories. In intercept surveys, 51.0% of shoppers self-identified as overweight and 60.6% wanted to change their weight. Shoppers who typically purchased one type of food over another commonly did so out of habit or because the item was on sale. Findings revealed that preintervention sales of healthier “target” or regular “comparison” items did not differ between intervention and control stores for 1 year prior to intervention implementation. Rates of compliance with the healthy marketing strategies were high, averaging 76.5% over the first 12 months in all 16 stores. If healthy in-store marketing interventions are effective in this scaled-up, longer-term study, they should be translated into wider use in community supermarkets.

Implications

Practice: This is a real-world pragmatic trial of a relatively low-cost, broadly feasible intervention strategy that—if effective—could be widely translated and implemented in large supermarkets.

Policy: Because bricks and mortar supermarkets remain the primary location where most Americans purchase food for themselves and their families, this study has broad potential for wide implementation and policy support if the results support the long-term effectiveness of placement and promotion strategies.

Research: Given the rapid deployment of healthy food access approaches across the USA without clear evidence on effectiveness, it is critical to conduct rigorous research using objective data to assess whether supermarkets in low-income neighborhoods can achieve positive health effects that may reduce health disparities in chronic diseases.

Introduction

As the prevalence of obesity has increased over the past two decades [1], public health experts have begun to focus on the environments that shape overeating, unhealthy food choices [2], and food insecurity. Food insecurity, defined as a lack of access to adequate or healthful food because of limited money or other resources, impacts 11.8% of American households [3,4]. Food-insecure households are at increased risk for a variety of poor health outcomes, including obesity and related comorbidities [5–7].

Retail grocery stores, the primary locations for food purchases, are pivotally positioned between consumers and the products they eat and are an opportune place to help influence food choices to favorably affect energy balance [8]. Efforts to increase access to supermarkets in disadvantaged urban and rural communities hold promise for promoting healthier diets [9]. But it is not clear what effect improved access to both healthy and unhealthy products will have on diet, food insecurity, or obesity. Several studies have found that simply having access to a neighborhood supermarket did not impact residents’ diet quality or obesity rates [10–13], while others found positive effects [14]. These data suggest the possibility that access, while necessary, may not be sufficient to drive healthier choices.

The development and/or reestablishment of supermarkets in low-income, ethnically diverse neighborhoods provides a unique opportunity to design and evaluate various strategies to promote the purchase of healthier products [8,15]. Increased availability of healthier foods, if translated to purchases that alter the products in people’s homes, has the potential to positively affect diet quality among those at greatest risk for obesity—low-income, ethnic minorities in food-insecure households. Yet, there is limited public health research examining how in-store marketing efforts affect the purchasing of healthier foods.

The largest number of community-based intervention studies in retail food stores has been in small stores (i.e., corner stores and bodegas)—and most have been pre–post studies and relied on manager and consumer self-reports of environmental and behavior change [16]. Previous supermarket interventions focused on healthier items have used point-of-purchase (POP) approaches (nutrition education posters, shelf-tags, and pop-out flyers) to increase awareness about the health attributes of targeted products [17]. POP interventions have had mixed results on the sale of healthier products; some studies report increased sales [18–22], while others found no change [18,23–30]. While price discounts (e.g., coupons) can be effective to increase sales [31–33], they require substantial ongoing investment. Overall, the existing literature on promoting healthier purchases in supermarkets have been conducted in middle-class areas, leveraged the health attributes of products, or used price discounts that are costly to sustain.

New research in low-income, high-minority neighborhoods that use representative data for shopper populations from store sales records, and supplements sales data with observations and surveys, can provide important new insights on how healthy food marketing strategies work. Such research can also shed light on how interventions in supermarkets affect diet quality in food-insecure groups. We previously conducted a large randomized pilot study in eight supermarkets, where we tested in-store marketing of healthier food options in five food and beverage categories, using simple placement and product availability strategies based on traditional marketing approaches. Those strategies significantly increased the sales of healthier, lower-calorie items in the milk and frozen dinner categories and of water in both the soda aisles and checkout beverage coolers, in intervention stores compared to control stores, during a 6 month intervention period [34]. The present study builds on the pilot study, with a larger sample and a longer-term intervention, thus investigating the scalability and sustainability of healthy in-store marketing strategies.

In this paper, we describe the design, methods, and baseline findings, as well as first year supermarket compliance with in-store marketing interventions for the scaled-up study. The primary aim of the study was to evaluate, in a randomized controlled trial design, the effects of in-store healthy food marketing strategies on sales of specific healthier items in six product categories (milk, frozen dinners, beverage checkout coolers, bread, salty snacks, and cheese). The main study hypothesis that sales of targeted healthier products would be significantly higher in the intervention stores after the intervention, and that the changes would be sustained for the duration of the intervention, drove the study design and methods described here. With respect to the baseline data reported in this paper, we hypothesized that there would be equivalence between intervention and control group stores with regard to store and shopper characteristics and that in-store marketing intervention compliance during the first 12 months would be as good or better than what was achieved in the pilot study.

Methods

Overall research strategy and conceptual framework

A randomized controlled trial design was used to test the impact of healthy food marketing strategies to promote the purchase of healthier “target” food items in six product categories: bread, checkout cooler beverages, cheese, frozen dinners, milk, and salty snacks. Thirty-three supermarkets in low-income, high-minority neighborhoods in the metropolitan Philadelphia area were randomly assigned to the intervention or control group. Healthy food marketing strategies, including prime shelf-placement and call-out promotion signs, increased “target” (healthier) product shelf space, and reduced regular “comparison” product shelf space, were implemented in 16 intervention stores for an 18 month period. The primary outcome measure was weekly sales per store in each product category for 1 year preintervention and 18 months during the intervention.

The foundational theoretical framework for this study was an ecological model of health behavior, which involves examining multiple levels of influences on food choice and its determinants [2,35,36]. Ecological models of health behavior emphasize the environmental and policy contexts of behavior while incorporating social and psychological influences. Ecological models lead to the explicit consideration of multiple levels of influence, thereby guiding the development of more comprehensive interventions [37]. These models focus on the relationships between people and their environments [2,35,36]. Key variables measured and analyzed in this study were at the organizational and individual levels. The study does not explicitly address the interpersonal and community levels, which are outside the scope of this study, and it is not a test of the full model. This study also draws on concepts from social marketing and consumer behavior, which blend psychology and economics to help understand behavior [38,39].

Recruitment and randomization of stores

Eligible stores were at least 35,000 square feet in size and located in urban or suburban communities with at least 50% of households below the state median income. Stores that received the pilot study intervention were excluded. In order for a store to participate, stores had to be willing to cooperate with the study randomization, assessment, and intervention procedures (if assigned to the intervention arm). Before randomization, stores were required to complete a corporate agreement to participate in the study and provide weekly sales data before and during the study period. Randomization was completed using matched pairs of stores to ensure a balance of relevant store-level factors (chain, region, and percentage of sales of Special Supplemental Nutrition Program for Women, Infants and Children (WIC) and Supplemental Nutrition Assistance Program (SNAP) food assistance programs). Randomization assignments were completed by the study biostatistician (K.H.M.) after enrollment of at least four stores. The target sample size of 32 stores was determined based on the minimum detectable effects (a small effect size of 1 total unit difference between groups) for three products included in the pilot study, assuming 80% power for testing the six product category outcomes between groups.

Interventions

Interventions in this study focused on two key strategic elements of the marketing mix: placement and promotion [38]. Strategies were initially identified through an integrative literature review of retail marketing strategies [40] and were refined in consultation with supermarket managers for the pilot study upon which this research builds [34]. Feasibility and sustainability in the retail setting were important considerations, thus pricing strategies, such as coupons and discounts, were not included. Products were selected as healthier items based on caloric content (i.e., lower-calorie products) in the milk, frozen dinners, beverage checkout coolers, salty snacks, and cheese sections of the stores and for whole-grain content in the bread section. As in the pilot study [34], product categories were selected for their high sales volume and potential for significant positive changes, as well as for price neutrality of the healthier target products in each category, that is, that the healthier product did not cost more than the comparison product (Table 1). Prior to implementing the interventions, research staff assessed target and comparison product placement and availability on store shelves in each intervention and control store. They then created store-specific planograms for intervention stores to improve the shelf placement and availability of target items while reducing the availability of comparison items by approximately 30%. Store staff implemented and maintained the planogram changes with assistance from research staff. Placement interventions in the dairy section focused on promoting lower-calorie milk (skim, 1%, and 2%) while diminishing the presence of whole milk. The visual order of the milk displays was changed, and the number of facings, or the fronts of packages the consumer can see on the shelf, of whole milk were decreased by 30% while increasing the facings of the lower-calorie milk. In the other product categories (frozen dinners, beverage checkout coolers, bread, salty snacks, and cheese), target products were moved to eye level and the number of facings was increased. Promotion strategies included placing call-out signs produced by research staff with templates provided by each supermarket chain alongside the targeted products. The signs were on bright colored paper, laminated, with the product indicated in large text. No additional messages, health information, or claims were provided. Signage was rotated monthly (e.g., new colors/different target products) to increase the chances that customers would notice them [8,39]. Milk taste tests [41] were conducted at intervention stores each month. An example of the intervention in beverage coolers is shown in Fig. 1.

Table 1

Intervention products

Category	Target (healthier) products	Comparison products
Milk	Skim, 1%, 2%	Whole
Beverages	Water, diet beverages	Sweetened drinks/soda
Sliced bread	100% whole wheat	White bread
Shredded cheese	Mozzarella and low-fat cheddar	Regular cheddar
Salty snacks	Pretzels	Regular potato chips
Frozen dinners	Single entrees <300 calories	No comparison

Fig 1

Sample intervention: checkout cooler.

Measures

For primary outcome measures, stores provided weekly sales data for each of the selected products. The data included units sold of all selected products (e.g., higher-calorie versions and lower-calorie alternatives) in ounces or packages. All sizes of the selected products were included. These data were collected for 1 year preintervention and the 18 months during the intervention and were transmitted to the research team as Excel files. An indicator of healthy sales movement was created for each product category based on the proportion of sales for the healthier option compared to total sales in the product category, summarized as a percentage. This indicator could not be calculated for the targeted frozen dinners because there was no comparison or “less healthy” frozen meals matched to those products.

Secondary measures included compliance monitoring for the interventions and shopper intercept surveys. Compliance checks were conducted twice a month for 6 months and monthly thereafter (in intervention stores only). Compliance with the planned placement and promotion strategies was rated on a scale of 0%–100% at each compliance check using a protocol and formulas developed in the pilot study and with established interrater reliability. For example, if the planogram specified that a healthier target product was to be placed at eye level with 50% more facings than the regular (comparison) product, the compliance check examined how well the placement adhered to the plan. Intercept surveys, adapted from Davis et al. [42], were conducted with consumers who did most of their shopping at the participating store, at baseline and 12 and 18 months in all stores, with a convenience sample of 50 surveys per store per occasion. These brief 25-item surveys assessed customers’ background characteristics, shopping patterns, such as the number of shopping trips per week, and whether they used a shopping list and/or coupons. The surveys also assessed buying habits for each targeted product category, including the reasons for selecting a specific type or brand and whether they would try a different type or brand in the future. The 12 and 18 month intercept surveys assessed whether shoppers noticed the placement and promotion strategies for the healthier target products that were the focal points of the intervention in the stores.

Additional measures included store and neighborhood characteristics, including neighborhood race and income, proximity to public transportation and other food retail outlets, and indicators of crime that might affect shoppers’ experiences (robbery and assault). We also tracked the overall food promotion environment in the supermarkets, using the Grocery Marketing Environment Assessment tool, based on the Nutrition Environment Measures Survey in stores, a valid and reliable, widely used assessment of healthful food choices in retail settings [15,43,44] [data not reported here].

Statistical analysis

Statistical analyses reported here examine the equivalence of stores and shoppers in the outlets randomized to intervention and control arms of the study. We computed descriptive statistics on store and neighborhood characteristics, baseline data from intercept surveys, intervention compliance during the first year of the intervention, and weekly sales data for the selected products during the first year of the intervention.

We compared the mean sales (total movement and average weekly movement) during the preintervention period across intervention and control stores. In addition, we compared, within intervention group, the mean sales of target and comparison products. The distribution of each of the outcomes was right skewed, so a flexible family of models was chosen. Specifically, we applied marginal longitudinal models (or generalized estimating equations) with a gamma distribution and log link. Each model included sales data for all target and comparison products within a category with fixed effects for product type (e.g., whole wheat vs. white bread), intervention group (intervention vs. control), time in weeks, and the interaction between the three main effects. The interaction allowed us to test the difference between the sales for intervention stores versus control stores and the difference between the sales in target versus comparison products. In addition, a structured exchangeable covariance structure was included to account for the correlation of paired stores. Model estimate means and 95% confidence intervals were estimated from appropriate contrasts. For frozen dinners, separate models were fit for each type since not all stores carried all product types. In addition, a composite of only the products sold in all stores (turkey dinners and lasagna) was derived and analyzed. We used a p-value <.05 for assessing statistical significance. Models were fit using SAS version 9.4 (SAS Institute, Cary, NC).

Results

Recruitment results

Recruitment was completed in two waves due to management changes in one previously committed supermarket chain, which led to their decision not to participate in the study. The second chain agreed that its stores would participate nearly 1 year after the first wave of recruitment began. Thirty-three supermarkets were randomized in the trial, one more than planned.

Store and neighborhood characteristics

Of the 33 supermarkets randomized in the trial, 16 were intervention stores and 17 were control stores. Stores were located in Southeastern Pennsylvania, Delaware, and southern New Jersey (Fig. 2). The average store size was 60,501 square feet (range 36,000–87,000) and stores had been open for an average of 24 years. Sixteen percent of the stores’ sales were WIC/SNAP, with a range from 2.5% to 56.5%. Using the location of the stores based on their census tracts, we determined the store’s neighborhood characteristics (Table 2). Across all stores, neighborhoods were 65.6% white, 24.1% black, and 5.8% Hispanic. Median household income averaged $52,325 per year and, on average, 11.7% of residents were below the poverty level. All but 3 of the 33 stores had access to a public transit stop within 0.25 miles of the store. There was an average of 15.2 supermarkets/small groceries/convenience stores within 2 km of a study store. In store neighborhoods, robbery averaged 271 per 100,000 population (above the national rate of 98) and aggravated assault averaged 351 per 100,000 (above the national rate of 249) [45]. Baseline percentage of healthy sales movement ranged from 15% to 60%. None of the store or neighborhood characteristics were significantly different between control and intervention stores.

Table 2

Store and neighborhood characteristics

Characteristic	Control n = 17	Intervention n = 16	p-value
	M (SD) or %	M (SD) or %
Store characteristics
Store size sq. ft, mean (SD)	61,608.3 (11,867.6)	59,320.3 (9,640.9)	.56
WIC/SNAP sales, %	17.1	14.8	.63
Years store has been open	24.8 (15.7)	22.4 (12.8)	.64
Neighborhood characteristics
Race/ethnicity
White, %	58.2	73.4	.14
Black, %	29.5	18.4	.27
Hispanic, %	7.2	4.4	.09
Median household incomea	49,007.5	55,849.1	.26
Living below poverty level, %	13	10.3	.51
Proximity to transport and other stores
Public transportation stop available within 0.25 miles of store? %	94.1	87.5	.51
Total number of supermarkets, small grocery stores, and convenience stores within 2k of HRS2 store, mean (SD)	17.7 (21.1)	12.4 (15.3)	.42
Crime
Robbery rate per 100,000 pop, mean (SD)	298.8 (196.9)	241.6 (174.4)	.38
Aggravated assault rate per 100,000 pop, mean (SD)	363.5 (195.6)	338.0 (166.3)	.69
Healthy sales movement (%)
Target beverages	33.6	34.9	.64
Target bread	16.7	15.1	.53
Target cheese	46.3	47.2	.83
Target milk	58.8	59.9	.58
Target salty snacks	40.6	39.4	.64

SD standard deviation.

^aMedian household income for states in the study: PA $56,951; NJ: $76,475; DE: $63,036.

Fig 2

Enrolled locations (N = 33).

Shopper intercept surveys

At baseline, 50 shopper intercept surveys were completed at each store for a total of 1,650 surveys. Shopper demographics and highlights of responses are summarized in Table 3. Most shopper characteristics were similar between those at control and intervention stores. Respondents were an average (standard deviation) age of 55.2 (16.7) years with almost half (47%) of all surveyed shoppers between 50 and 69 years old. The majority of shoppers were female at 71.6%. Fifty-nine percent of shoppers identified themselves as non-Hispanic white, 31% non-Hispanic black, and 5% other or mixed race. Four percent identified as Hispanic ethnicity. There were more blacks and fewer whites in the Control stores (p < .01). Thirty-five percent of shoppers surveyed had a high school education or less.

Table 3

Baseline characteristics of shoppers from intercept interviews (N = 1,650; some missing data)

Demographic characteristicsa	Control (n = 850)	Intervention (n = 800)	p-value
Age, mean (SD)	55.57 (16.18)	54.78 (17.15)	.64
Female, n (%)	603 (71.8)	562 (71.4)	.87
Race/ethnicity, n (%)			<.01
Hispanic	30 (3.7)	35 (4.5)
Non-Hispanic Black	302 (37.1)	197 (25.5)
Non-Hispanic White	433 (53.1)	508 (65.8)
Other	50 (6.1)	32 (4.1)
Education level, n (%)			.50
High school graduate/GED or less	285 (33.6)	287 (36.0)
Some college or technical school	276 (32.5)	260 (32.6)
College graduate or more	287 (33.8)	251 (31.5)
SNAP recipient, n (%)	142 (16.8)	156 (19.6)	.15
Household size, mean (SD)	2.68 (1.56)	2.87 (1.79)	.02
Health-related characteristics
Perceived weight status, n (%)			.12
Underweight or right weight	376 (47.0)	384 (50.9)
Overweight	424 (53.0)	370 (49.1)
Wishes to change weight, n (%)	521 (62.1)	470 (59.1)	.22
Shopping habits
Shopping frequency, %			.90
Shop more than once a week	45.1	46.3
Shop once a week	32.4	30.7
Shop less than once a week	22.6	23
Shopping list use, n (%)			.07
Never	16.4	14.6
Rarely	14.6	13.4
Sometimes or more	69.0	72.1
Purchase products not on list, n (%)			.12
Never	0.9	1.6
Rarely	6.2	5.1
Sometimes or more	92.9	93.3
Usually buys a particular brand/type (%)
Milk	83.3	86.4	.10
Bread	61.6	60.9	.21
Cheese	44.8	43.7	.91
Salty snack	44.7	43.4	.85
Beverages	39.0	41.1	.40
Frozen dinners	23.1	22.5	.10
GED General Education Development test (high school diploma equivalency)

With respect to perceived weight status, 51.1% considered themselves to be overweight and 60.6% wanted to change their weight, mainly to lose weight. On average, 46% of shoppers reported shopping at their grocery store more than once a week, a majority used a shopping list at least some of the time (71%), and 18% received SNAP benefits. Shoppers were asked to report whether or not they usually buy a particular brand or type of our six product categories, and the reasons why. The most important reasons for usually buying the same product type, across all product categories, were that it was a habit, that someone in the household likes it, or that it was on sale. Eighty-five percent of respondents answered yes for milk and 61% for bread. About 44% answered yes for salty snacks and cheese. The type and brand mattered the least for beverages and frozen dinners at 40% and 23%, respectively. As for placement and promotion of products, 61% reported that they notice the products located at endcaps but only 43% said that they notice special promotions and sales while shopping. About half of the shoppers noticed the drinks in the checkout coolers.

Average weekly sales in the 52 weeks preintervention

Table 4 shows the average weekly sales volume in intervention and control stores, for each of the target (healthier) products and comparison products (see Table 1), in the preintervention phase. Analyses revealed that, with the exception of some of the frozen dinners, there were no significant differences between intervention and control in baseline volume sales of the products included in the study. Statistical modeling for the main results will account for those few products where baseline differences need to be considered in main outcome analyses.

Table 4

Average weekly sales 52 weeks preintervention

			Average weekly sales
Category	Product	Store Type	Mean	95% CI	p-value
Bread (ounces)a	Wheat	Intervention	2,289.10	1,868.4–2,804.5	.14
		Control	2,792.60	2,100.3–3,713.3
	White	Intervention	17,349.70	12,425.8–24,224.8	.94
		Control	17,183.80	12,100.0–24,403.6
Frozen dinnersb (boxes)	Pasta	Intervention	17.2	14.1–20.9	.02
		Control	23.4	17.0–32.4
	Mac and cheese	Intervention	6.5	4.6–9.2	.85
		Control	6.7	4.7–9.7
	Chick nuggets	Intervention	22.7	16.7–30.9	.05
		Control	31.9	24.2–42.1
	Turkey	Intervention	9.9	7.7–12.7	.04
		Control	12.3	10.0–15.2
	Lasagna	Intervention	8.3	5.6–12.4	.15
		Control	10.3	7.1–14.9
	Compositec	Intervention	18.1	13.3–24.5	.04
		Control	22.3	17.1–29.2
Salty snacks (ounces)	Pretzel	Intervention	2,102.90	1,495.5–2,957.1	.43
		Control	1,878.20	1,422.2–2,480.4
	Chip	Intervention	2,962.70	2,262.4–3,879.9	.33
		Control	2,663.30	2,126.6–3,335.5
Cheese (ounces)	LF cheddar/Mozzarella	Intervention	2,233.60	1,670.0–2,987.3	.71
		Control	2,162.20	1,670.3–2,798.9
	Cheddar	Intervention	2,726.50	1,989.2–3,737.2	.15
		Control	3,255.60	2,144.4–4,942.7
Beverages (ounces)	Water	Intervention	1,569.10	1,278.3–1,926.1	.08
		Control	2,134.10	1,455.1–3,129.9
	Diet/unsweetened	Intervention	3,061.00	2,286.4–4,098.0	.61
		Control	2,857.60	2,340.0–3,489.7
	Regular	Intervention	8,898.90	7,030.6–11,263.8	.18
		Control	10,460.80	7,830.3–13,975.0
Milk (ounces)	Skim	Intervention	23,154.50	17,221.8–31,130.9	.57
		Control	21,364.60	16,721.0–27,297.7
	1%	Intervention	42,110.30	33,140.1–53,508.6	.89
		Control	41,603.50	33,298.2–51,980.4
	2%	Intervention	81,456.20	63,058.0– 105,222.5	.83
		Control	83,036.80	67,938.7–101,490.1
	Whole	Intervention	96,971.70	77,940.8–120,649.5	.56
		Control	101,389.60	83,632.5–122,916.9

^aUnits of bread were in 16 or 20 ounce packages.

^bEstimates of means and confidence intervals (CIs) for frozen dinners based on analyses stratified by product; all product types not at all stores.

^cSum of units sold for turkey and lasagna. Pasta, mac and cheese, and chicken nuggets were not carried in all stores.

Intervention compliance in the first 12 months

Overall compliance across product categories and stores averaged 76% (range 62%–88%). Scores varied over time between stores and product categories but did not decrease significantly through the year. Cheese and milk had the greatest mean compliance at 82%–85%. Bread had the lowest mean compliance at 65%. There were no individual stores that had continuously poor compliance.

Discussion

Current efforts to address obesity and food insecurity recognize the complexity of these widespread problems, and the contributions of many influences on diet quality need further study [3]. Our study begins to bridge this gap by examining how effectively in-store interventions increase sales of more healthful foods. Key strengths of our study include the randomized design, incorporating marketing tactics alongside sustained changes in product positioning, encouraging changes within already high-selling product categories, interventions developed in conjunction with retailers serving residents in low- to moderate-income communities, and the use of objective sales data as the main outcome. An important challenge in the study, identified by responses to the intercept surveys, is that many shoppers habitually buy a particular type or brand of the target products in the study, especially for milk and bread. Findings also suggest that different product categories may have varying levels of product loyalty and price sensitivity and, therefore, be more or less influenced by certain in-store marketing strategies.

The present study is the only one of its size that examines objective sales data (an objective measure; not self-report) to test the impact of targeted in-store promotion of healthier alternative products, wherein product placement and promotion are implemented. Other studies are limited in their emphasis on nutrition education only, specific geographies or products, and small store formats [46–50]. A review of studies that sought to improve healthy food purchases in supermarkets or grocery stores found that few interventions utilized randomized designs (6 of 33 studies) and nearly all conducted interventions for less than 12 months [51], with the exception of three information-only interventions [52–54].

Prior research has also emphasized the critical nature of cooperative partnerships with retailers in order to implement and maintain in-store marketing interventions and has identified such participatory and translational research partnerships as critical to maximizing public health impact [55,56]. The present study approach and implementation findings illustrate successful partnerships in that compliance across categories and stores over a 12 month intervention period were high (62%−88%), averaged 76.5%, and did not decline over time. While there is no benchmark in the literature to define supermarket compliance in a study like ours, other studies [57,58] and reviews [55] have reported challenges to achieving health-promoting changes in store food environments, including product availability and placement. We used compliance-enhancing monitoring and feedback to achieve a rate that was four percentage points (or 5.4%) higher than in our pilot study [34]. Also, our collaborative approach to intervention development and maintenance maximizes the potential for wider-scale program replication.

In partnership with retailers, our team met regularly before the intervention began to codesign a feasible approach to product placement and shelf-tagging strategies using retailer templates whenever provided. Further, the team provided regular updates to the stores and chains on implementation, sales data collection in order to maintain and continue to build strong, effective relationships with supermarket management and staff. A primary goal of translational research in public health is to identify new approaches to support population health for which wide-scale adoption and implementation is possible [59]. In alignment with new thinking about best practices in research translation, the present study sought to optimize the potential for replicability in its approach to working with retailers.

The findings from the baseline data collected for this study provide a firm foundation for the ultimate analysis of the effectiveness of the healthy in-store food marketing strategies, given the comparability of stores and shoppers in intervention and control conditions. Data from the baseline intercept interviews suggest important issues to consider in analyzing and interpreting the main results of the trial. The compliance findings attest to the high level of implementation of the in-store marketing interventions, another key element for assuring internal validity of the trial findings.

Data on the relationship between dietary quality and food insecurity are mixed, and suggest that the relationship is complex and may vary by gender, age, and level of food insecurity [6,7] and are likely moderated in part by participation in programs like SNAP [60]. Households with very low food security shop more often in smaller stores [61] and low food security has also been found to be associated with high levels of consumption of high-fat dairy products, salty snack foods, and sugar-sweetened beverages [62,63].

The recruitment period for the study was longer than expected and occurred in waves rather than all at once. Ultimately, the study enrolled stores across two large retail chains; however, in the early phases of the study, the research team met with a range of store chain sizes and formats, including smaller and limited-assortment stores. Despite our openness to alternative store formats and chain sizes, our preliminary visits to those stores revealed that they did not carry the variety of products and potential brands planned for the intervention, and the store owners’ commitment to participate was insufficient to enroll and randomize those stores. Thus, while the enrollment period was more time-intensive then initially expected, the process likely supported a clear understanding of project goals and operational needs and created a partnership that contributed to sustained collaboration; no stores dropped out once the study began.

Translational Implications

The study described in this article provides an important, and the most robust to date, test of a key premise of an ongoing national policy of financing supermarkets in disadvantaged and minority communities, intended to increase food access [64] and facilitate healthier diets. Given the rapid deployment of these approaches across the USA without clear evidence on effectiveness, it is critical to conduct a rigorous study using objective data to assess whether supermarkets in low-income neighborhoods can achieve positive health effects that may reduce health disparities in chronic diseases.

The study design provides an important example of a real-world pragmatic trial of a relatively low-cost, broadly feasible intervention strategy that—if effective—could be widely translated and implemented in large supermarkets. In 2018, 49% of shoppers viewed the supermarket as their primary store, and supercenters, while still popular, have declined in overall use since 2010 as shoppers’ primary store for groceries (27% in 2010 vs. 24% in 2018) [65]. Even with the advent of new retail food sources, for example, the rise of dollar stores’ food sales, delivery, and online purchasing, the bricks and mortar supermarket remains the primary location where most Americans purchase food for themselves and their families. This study demonstrates the feasibility of implementing healthy food marketing strategies in supermarkets, a finding which, if associated with improvements in healthful food sales, will justify the widespread implementation of these low-cost approaches. Future research should examine the applicability of similar methods to food pantry settings [66,67] while recognizing that the layout of food pantries may be less standardized than that of large supermarkets.

Therefore, this study has broad potential for wider implementation if the results support the long-term effectiveness of these placement and promotion strategies for increasing sales of healthier products in low-income and food-insecure neighborhoods.

Acknowledgments:

The authors thank Erica Davis, Matt Phillips, Brean Flynn Witmer, Steve Menkes, Vicky Tam, Julia Orchinik, and Sarah Green and the supermarket managers and staff

Funding:

This study was funded by grant number 1R01DK101629 from the National Institute of Diabetes, Digestive and Kidney Diseases of the National Institutes of Health.

Compliance with Ethical Standards

Conflicts of Interest: The authors declare that they have no conflicts of interest.

Authors’ Contributions: K.G. conceptualized and led the study and wrote the main draft of the paper. A.C. coordinated fieldwork and data collection. K.M. provided statistical expertise and conducted randomization of stores. P.K. conducted statistical analyses. D.W. led neighborhood characteristics analyses and assessments. D.P.G. managed the study operations. C.M.B. managed data cleaning and reduction for sales data. A.K. co-conceptualized the study and wrote significant sections of the paper. All authors reviewed, edited and approved the manuscript.

Ethical Approval: All procedures performed were in accordance with ethical standards. The study procedures were approved by the University of Pennsylvania IRB. This article does not contain any studies with animals.

Informed Consent: Informed consent was obtained from all individual participants and supermarkets that were included in the study.

Clinical Trials Registration: NCT02499211.