A Healthy Retail Intervention in Native American Convenience Stores: The THRIVE Community-Based Participatory Research Study

Abstract

Objectives. To assess a healthy retail intervention in Tribal convenience stores in Oklahoma.

Methods. We adapted healthy retail strategies to the context of 8 Tribally owned stores. We assessed individual- and store-level outcomes in a cluster-controlled intervention trial (April 2016–June 2017). We measured fruit and vegetable intake, store environment perceptions, and purchases before and after the intervention among a cohort of 1637 Native American shoppers. We used mixed-effects linear regression to estimate pre- to postintervention changes in and between groups.

Results. We followed 74% of participants (n = 1204) 9 to 12 months. Intervention and control participants perceived healthier stores after intervention. Higher shopping frequency was related to purchases of fruits, vegetables, and healthy items.

Conclusions. Intervention exposure was associated with healthy purchasing but not fruit and vegetable intake. Research is needed to further assess impacts of environmental interventions on intake.

Public Health Implications. As the first healthy retail intervention in Tribally owned stores, our results contribute evidence for environmental and policy interventions to address obesity in Tribal Nations. Multicomponent interventions, led by Tribal leaders from diverse sectors, are needed to create healthy environments and sustainable improvements in Native American health.

Community food environments with few or no supermarkets and numerous fast food and convenience stores are associated with obesity.^1–3 Native Americans are more likely to be obese than are Whites and have rates of diabetes and hypertension that exceed those of the US general population.⁴ However, few studies have examined the food environments of Native American communities, which include rural, reservation, and urban communities. A study of 2 Southwestern reservations found that supermarkets were scarce and Native Americans were dependent on convenience stores that stocked unhealthy foods and no produce.⁵ Other studies of rural and reservation communities found poor access, high cost, and poor quality of fresh produce as barriers to vegetable and fruit consumption.^6,7 Urban-dwelling Native Americans also appear to encounter barriers to healthy eating. Although research is limited, a study by Dammann and Smith found that Native Americans living in the inner city had poor access to grocery stores and shopped at corner stores, purchasing only essential items at inflated prices.⁸

Interventions to improve community food environments are recommended to address obesity.^9,10 Encouraging supermarkets to locate in low-income communities is helpful, but a more feasible approach is to implement “healthy makeovers” in existing stores to improve healthy food access.^11,12 Although results are mixed, recent systematic reviews conclude that healthy retail interventions encourage customers to buy and eat healthy foods, particularly interventions that combine lowering prices with easy access and engaging promotion strategies.^13,14

To our knowledge, only 2 published studies, both in reservations, have tested the efficacy of healthy retail interventions in Native American communities.^15,16 These studies included cooking demonstrations, taste tests, and education. The interventions increased knowledge and healthy food purchasing and, in the second trial, noted a trend toward lower body mass index among intervention group participants although results were not significant.¹⁶ The study cites the absence of convenience store participation as a limitation to broader intervention reach.¹⁶ Although both studies generated important knowledge, the diversity of Native American communities requires that healthy retail interventions be adapted for community context.¹⁷

Tribal Health and Resilience in Vulnerable Environments (THRIVE) is a 5-year participatory research study to improve Tribal food environments by implementing healthy makeovers in rural Tribally owned convenience stores in the Chickasaw and Choctaw Nations of Oklahoma. The study is grounded in social cognitive theory, which posits that behavior change is influenced by the interaction of personal, behavioral, and environmental factors.¹⁸ To our knowledge, this is the first healthy retail intervention study to be implemented in Tribally owned convenience stores.

The Chickasaw and Choctaw Nations, once reservations, are now classified by the US government as Tribal Jurisdictional Areas and are among the largest sovereign indigenous Nations in the United States.¹⁹ The Nations together constitute one quarter of the state of Oklahoma’s population and have a combined population of more than 70 000 Native Americans residing in the Tribal boundaries.¹⁹ The poverty rate for all residents in the Chickasaw and Choctaw Nations is 15.3% and 20.7%, respectively,¹⁹ and both Nations have high rates of diabetes (25%), obesity (56%), and hypertension (48%).²⁰ Access to healthy foods is limited; 56% of Native Americans report inadequate food quantity.²¹ More than half (56%) of Tribal members travel more than 20 miles round trip to purchase groceries²² (this is a food desert indicator²³), and 65% shop for food at a convenience store once or more per week.²²

The THRIVE study is led by a steering committee comprising university health researchers, Tribal employees (most of whom are also Tribal citizens), and leaders from the health, commerce, and government sectors of both Nations. The steering committee guided and implemented all aspects of the study.²² We present the primary results from the healthy retail intervention trial. Data are available on request from the Chickasaw and Choctaw Nations. The survey materials (Appendices A and B) and data analysis code (Appendix C) are available as supplements to the online version of this article at http://www.ajph.org. When presenting study data, the names of the Tribes are not identified, per Tribal preference.

METHODS

Convenience stores provide a convenient location to quickly purchase food and gasoline and are an important economic resource for Tribes.²⁴ In 2014, the year of the most recent available data, there were 293 Tribally owned convenience stores in 25 states, and these numbers are increasing.²⁴ The Chickasaw and Choctaw Nations own more than a dozen convenience stores throughout rural Oklahoma. These stores are similar to non-Tribal convenience stores in size and scope, but differences exist. All employees are Tribal employees and all revenue generated from the stores is used for operations, health, and social services in both Nations. Tribal citizens receive discounts on all purchases. The stores sell widely marketed commercial foods and snacks. Additionally, the Tribal stores have tables and chairs, casinos, and “smoke shops,” where commercial tobacco products are sold without state or county taxes.

Intervention Description

Using a 2-step participatory research process described elsewhere,²² the THRIVE study sought to increase the availability, variety, and convenience of healthy foods and implement placement, promotion, and reduced pricing interventions in the stores. To select the healthy products, we assessed the Tribal stores to identify ready-to-eat foods high in vegetables and fruits. We also assessed foods not currently sold in the stores but available through the food distributors that procure foods from large-scale commercial farms. Although the Nations have policies to procure local and Tribally produced foods, no local farms can supply the quantity needed for the stores, and thus local foods were not available. Similarly, although a number of the potential and ultimately selected intervention foods contained fruits and vegetables traditional to Choctaw and Chickasaw people (e.g., corn, tomatoes, and beans), traditional meals (e.g., Banaha and Tanchi Labona) could not be included because of preparation and transportation barriers.

Once we compiled potential foods, we conducted taste tests and focus groups with Native American shoppers to finalize intervention foods and understand shopper preferences, which guided the promotion, placement, and pricing strategies. Each Nation identified 11 meals and 20 snacks to promote.²² Healthy snacks contained 200 or fewer calories with 35% or less of the calories from fat; healthier meals contained 500 or fewer calories with 30% or less of the calories from fat, per the guidelines of the Nutrition Environment Measurement Survey (NEMS).²⁵

We placed open-air coolers stocked with intervention foods in the stores and displayed fresh fruit baskets near store entrances and other prominent locations. The healthy foods were marketed and promoted with in-store signs and displays, including Native American language signs. Both Nations offered combination meals (e.g., a wrap with a side fruit or vegetable and water) priced at 30% below the sum of the individually priced items.²²

Trial Design

We used a cluster-controlled trial design with treatment condition allocated at the store level (Figure 1).^26,27 We recruited Native American shoppers at each store to assess individual-level changes. In each Nation, 2 stores received the intervention and 2 stores served as control stores (n = 8). We selected the stores on the basis of their similarity in size, sales patterns, sociodemographic characteristics, and geographic distance from one another to avoid contamination (i.e., at least 60 miles apart).

FIGURE 1—

Tribal Health and Resilience in Vulnerable Environments (THRIVE) Study Store Randomization, Recruitment, and Data Collection in Oklahoma Native Nations: April 2016–June 2017

In both Nations, we conducted baseline assessments of individual-level measures immediately before the intervention. On the basis of funding and staffing availability, the intervention length differed by Nation, lasting 9 months in Nation A and 12 months in Nation B. We assessed follow-up measures of individual-level outcomes on the same cohort of store shoppers after 9 and 12 months. Additionally, we measured exposure to placement, promotion, and pricing intervention components at follow-up with the intervention cohort.

Participant Recruitment and Consent

Both Tribal Nations advertised the study by posting flyers at convenience stores, direct e-mail, and social media. Tribal and university staff, all of whom were study executive committee members, conducted face-to-face recruitment at the 8 stores. Adults aged 18 years and older who met the following criteria were eligible to participate: (1) self-identified as Native American living in either Nation, (2) shopped at Tribal stores at least 3 times per week, and (3) had no plans to move in next 12 months. Eligible shoppers gave consent, completed the baseline questionnaire, and were given a $10 gift card. We contacted participants again via telephone at the end of the intervention to complete a follow-up questionnaire, and we mailed them a $20 gift card. There was no replacement for those lost to follow-up. Tribal and university staff routinely collected process evaluation measures throughout the intervention to assess fidelity.

Measures

Individual-level outcome measures.

The primary individual-level outcomes were changes in fruit and vegetable consumption from baseline to follow-up. Secondary individual-level outcomes included consumption of other foods, changes in perceived food environment, and recall of promotions and subsequent reported purchase of intervention items among intervention participants. With the exception of questions about exposure to intervention components, which we measured at follow-up only, we measured all outcomes before and after the intervention.

We used a brief dietary screener to assess 16 foods. Participants reported their intake for fruits, green salad, and other vegetables. Additionally, participants were asked to report intake of fried potatoes, high-fat meats (6 items), lean meats (2 items), and baked and nonbaked chips. We assessed intake for each item on the basis of intervals used in the National Cancer Institute Quick Food Scan, ranging from never to 2 or more times per day.²⁸ We used National Cancer Institute scoring procedures to convert categorical intake to a times per day variable. Although we also measured beverage intake, our focus was the food-related intervention and findings.

We used sections of the Perceived Nutrition Environment Survey²⁹ to assess participants’ perceptions of the food environment in study stores: 7 items from the “food shopping” section to assess healthy and unhealthy food promotion and placement and 7 items from the “restaurant/eating out” section to assess healthy and unhealthy food availability and cost. Participants responded using a 5-point Likert scale ranging from “strongly disagree” to “strongly agree.”

We administered a study-specific section of the survey, which included pictures of intervention signage and displays, to intervention participants to evaluate whether participants noticed signage and promotional displays of intervention foods and made related purchases. We included some signs not used in the intervention to assess acquiescence bias.

Store-level outcomes measures.

We adapted components of the NEMS tools²⁵ to assess objective changes in the nutrition environment of the stores before and after the intervention. Our adapted instrument, described in detail elsewhere,³⁰ measured the availability of healthy foods with an emphasis on ready-to-eat fruits and vegetables. Categories included fruits; vegetables; beverages; salads; sandwiches, wraps, and burritos; dairy items; baked chips; other healthy snacks; canned meats; low-fat baked goods; and cereals. In addition to the availability domain, the NEMS measured pricing, placement, and quality. We calculated the total scores and subscores for each category and domain at baseline and follow-up for each intervention and control store. Tribal and university staff administered the NEMS in all study stores before and after the intervention.

Statistical Analysis

On the basis of power calculations to detect an effect size of 0.2 with up to 20% dropout and 10% reduction in sample size because of clustering, the target population was 820 Native Americans per Nation (n = 1640 total). We used the t test and χ² test to compare demographic and health characteristics among the control and intervention groups. We used multilevel mixed-effects linear regression to examine pre- to postintervention changes in outcomes in and between groups. All mixed-effects models included random effects for individual and store to account for clustering at both levels. We fit in-group effect using models stratified by intervention group; models included a main effect term for time point only. Between-group effect models included main effect terms for intervention group and time point and their interaction term. We interpreted the interaction term, the effect of primary interest, as a comparison of the in-person effect in control versus intervention store shoppers. We adjusted between-group effect models for age, gender, and education. We used restricted maximum likelihood estimation, which is the preferred approach when few clusters are available.²⁶ Inferential tests were on the basis of the t distribution with degrees of freedom appropriate for the number of clusters (stores) included in the analysis.²⁷

Analyses for participant response to changes implemented at the intervention stores were restricted to participants from intervention stores who shopped at that store at least once during the previous month. We report the overall frequency of participants who reported each behavior and according to store visit frequency. Inferential models used Poisson regression with robust SE estimates and included fixed effects for age, gender, education, and store.

We estimated changes in store-level nutrition environments as the median difference between baseline and follow-up in the NEMS scores, including food category and domain subscores. We compared intervention and control stores’ NEMS score changes using the Wilcoxon Rank Sum test because not all were normally distributed. We stratified all individual-level analyses in Nation and conducted analyses in Stata version 14.2 (StataCorp LLC, College Station TX). We conducted store-level analyses in SAS version 9.4 (SAS Institute, Cary, NC).

RESULTS

The cohort of shoppers included 1637 individuals, with 813 from Nation A and 824 from Nation B. We reached 1204 (73.5%) members of this cohort for follow-up. The follow-up response rate differed by Nation and store, with 65.4% overall for Nation A with a range of 56.9% to 71.4% by store and 81.6% overall for Nation B with a range of 74.4% to 85.2% by store. Participant characteristics, stratified by Nation, are shown in Table 1.

TABLE 1—

Tribal Health and Resilience in Vulnerable Environments (THRIVE) Study Participant Characteristics According to Intervention Group, by Oklahoma Native American Nation: April 2016–June 2017

	Nation A			Nation B
Characteristic	Control (n = 403), No. or Mean ±SD	Intervention (n = 410), No. or Mean ±SD	Pa	Control (n = 409), No. or Mean ±SD	Intervention (n = 415), No. or Mean ±SD	Pa
Age, y	42.4 ±15.5	40.1 ±14.9	.03	42.6 ±15.0	42.2 ±14.6	.74
Female	45	62	< .01	70	70	.91
Marital status, %
Married/in a relationship	64	59	.17	67	63	.29
Widowed/divorced	21	26		23	23
Never married	15	16		10	14
Persons aged < 18 y living in household	1.4 ±1.6	1.8 ±1.7	< .01	1.3 ±1.4	1.2 ±1.3	.88
Education, %			< .01			.05
< high school/GED	17	28		10	8
High school diploma	31	26		18	23
Some college or technical school	30	31		32	28
Associate’s degree or technical college degree	9	4		17	13
≥ 4-year college degree	12	11		22	28
Employed ≥ part-time, %	77	71	.06	73	81	.01
Body mass index, kg/m2	29.9 ±7.0	31.1 ±6.6	.02	31.2 ±6.6	32.0 ±7.4	.16

Note. GED = general equivalency diploma.

^at test or χ² test comparing control and intervention store participants in Native American Nation.

Individual-Level Outcomes

Fruit, vegetable, and other food intake frequency.

Reported fruit and vegetable intake before the intervention was low, with mean fruit intake of approximately one half times daily and mean vegetable intake of just less than once daily (Table 2). Daily fruit and vegetable intake after the intervention remained low in both control and intervention participants. Between-group effects comparing control and intervention participants were close to zero and not statistically significant for fruit or vegetables in either Nation. Sensitivity analyses examining fruit and vegetable intake according to reported shopping frequency showed similar results across all shopping frequencies (data not shown). Reported consumption frequency of some foods on the screener, including meats, fried potatoes, and chips, significantly decreased in the intervention and, in some cases, the control group.

TABLE 2—

Change in Fruit and Vegetable Intake Between Before and After Intervention in Control and Intervention Groups, by Oklahoma Native American Nation: April 2016–June 2017

	Control Store Participants			Intervention Store Participants
Food	Preintervention, Mean ±SD	Postintervention, Mean ±SD	In-Person Effect,a (95% CI)	Preintervention, Mean ±SD	Postintervention, Mean ±SD	In-Person Effect,a (95% CI)	Between-Group Effect,b (95% CI)
Nation A
Fruit, times/d	0.48 ±0.55	0.51 ±0.53	0.02 (–0.06, 0.09)	0.47 ±0.53	0.52 ±0.52	0.03 (–0.03, 0.10)	0.02 (–0.20, 0.25)
Green salad and other vegetables, times/d	0.82 ±0.78	0.83 ±0.77	0.01 (–0.11, 0.12)	0.76 ±0.77	0.74 ±0.70	−0.04 (–0.13, 0.05)	−0.04 (–0.36, 0.29)
Fried potatoes, times/wk	1.90 ±2.25	1.67 ±2.27	−0.20 (–0.53, 0.14)	2.14 ±2.73	1.72 ±2.16	–0.43 (–0.76, −0.10)	−0.16 (–1.18, 0.86)
High-fat meats,c times/wk	8.39 ±9.95	7.62 ±9.62	−0.61 (–2.08, 0.87)	9.77 ±10.97	7.92 ±7.88	–1.67 (–2.90, −0.43)	−1.21 (-5.43, 3.01)
Leaner meats,d times/wk	3.52 ±3.93	3.25 ±4.41	−0.26 (–0.91, 0.39)	3.83 ±4.20	2.86 ±3.12	–0.92 (–1.40, −0.44)	−0.63 (–2.39, 1.14)
Nonbaked chips, times/wk	2.00 ±2.50	1.48 ±2.46	–0.50 (–0.88, −0.12)	2.30 ±2.64	1.58 ±2.21	–0.69 (–1.02, −0.37)	−0.15 (–1.24, 0.95)
Baked chips, times/wk	1.04 ±1.89	0.79 ±1.79	−0.22 (–0.50, 0.06)	1.38 ±2.45	0.74 ±1.37	–0.63 (–0.91, −0.35)	−0.37 (–1.24, 0.50)
Nation B
Fruit, times/d	0.52 ±0.58	0.48 ±0.48	−0.04 (–0.11, 0.02)	0.47 ±0.50	0.48 ±0.51	0.02 (–0.04, 0.07)	0.04 (–0.15, 0.24)
Green salad and other vegetables, times/d	0.87 ±0.85	0.82 ±0.70	−0.05 (–0.15, 0.04)	0.78 ±0.68	0.73 ±0.61	−0.06 (–0.13, 0.01)	0.01 (–0.24, 0.26)
Fried potatoes, times/wk	2.02 ±2.33	1.56 ±1.76	–0.46 (–0.73, −0.19)	1.93 ±2.18	1.74 ±2.04	−0.17 (–0.40, 0.07)	0.33 (–0.48, 1.14)
High-fat meats,c times/wk	7.44 ±9.26	5.88 ±6.50	–1.32 (–2.32, −0.33)	7.88 ±9.06	7.26 ±7.72	−0.49 (–1.46, 0.47)	0.93 (–2.26, 4.13)
Leaner meats,d times/wk	2.91 ±3.55	2.40 ±2.79	–0.47 (–0.88, −0.06)	2.70 ±3.26	2.31 ±2.50	–0.40 (–0.78, −0.01)	0.12 (–1.16, 1.41)
Nonbaked chips, times/wk	2.22 ±2.59	1.26 ±1.86	–0.96 (–1.26, −0.65)	1.91 ±2.37	1.33 ±1.92	–0.56 (–0.81, −0.31)	0.49 (–0.40, 1.39)
Baked chips, times/wk	0.93 ±1.69	0.90 ±1.75	−0.02 (–0.25, 0.21)	0.77 ±1.38	0.76 ±1.31	−0.02 (–0.20, 0.15)	0.08 (–0.57, 0.73)

Note. CI = confidence interval.

^a(Postintervention consumption – preintervention consumption), unadjusted.

^b(In-person effect for intervention store shoppers – in-person effect for control store shoppers); adjusted for age, gender, education.

^cHot dogs, hamburgers, bacon/sausage, fried meat/cheese, regular lunch meats (pastrami, salami, bologna, liverwurst, regular ham), canned luncheon meats (regular Spam, Vienna sausages, potted meat).

^dLean lunch meats (turkey, chicken, lean ham), canned chicken, tuna, or sardines.

Perceived nutrition environment.

Compared with before the intervention, intervention store participants in Nation A perceived an increase in postintervention placement and promotion that encouraged purchase of healthy items (mean in-person effect = 0.21; 95% confidence interval [CI] = 0.08, 0.34) and availability of healthy options at store grills (mean in-person effect = 0.19; 95% CI = 0.03, 0.34; Table 3). In Nation B, both control and intervention store participants’ perceived changes in placement and promotion that either moved unhealthy items away from common purchase areas (end of aisles, near cash register) or discouraged their purchase. Control and intervention store participants from Nation B also perceived an increase in placement and promotions that encouraged purchase of healthy items and availability and promotion of healthy grill options. In addition, both control and intervention store participants from Nation B perceived a decrease in the cost of healthy grill options (control: mean in-person effect = −0.33; 95% CI = −0.53, −0.14; intervention: mean in-person effect = −0.39; 95% CI = −0.54, −0.24). Although the perceived positive changes in nutrition environment were larger for intervention compared with control store participants in both Nations, the between-group effects were not statistically significant.

TABLE 3—

Change in Perceived Nutrition Environment (NEMS-P) Between Before and After Intervention in Control and Intervention Groups, by Oklahoma Native American Nation: April 2016–June 2017

	Control Store Participants			Intervention Store Participants
NEMS-P domain	Preintervention, No. or Mean ±SD	Postintervention, No. or Mean ±SD	In-Person Effect,a (95% CI)	Preintervention, No. or Mean ±SD	Postintervention, No. or Mean ±SD	In-Person Effect,a (95% CI)	Between-Group Effect,b (95% CI)
Nation A
All stores	401	128		410	255
Placement/promotion of unhealthy items	3.13 ±0.74	3.07 ±0.72	−0.08 (–0.21, 0.05)	3.17 ±0.76	3.08 ±0.74	−0.08 (–0.18, 0.02)	−0.01 (–0.38, 0.36)
Placement/promotion of healthy items	3.36 ±0.97	3.55 ±1.02	0.16 (–0.03, 0.35)	3.42 ±0.97	3.64 ±0.90	0.21 (0.08, 0.34)	0.07 (–0.42, 0.56)
Stores with a grill	0	0		205	147
Availability of healthy options at the grill	. . .	. . .	. . .	3.37 ±0.85	3.57 ±0.77	0.19 (0.03, 0.34)	. . .
Grill promotes healthy options or nutrition information	. . .	. . .	. . .	3.04 ±0.64	3.18 ±0.64	0.10 (–0.02, 0.22)	. . .
It costs more to buy the healthy options at the grill	. . .	. . .	. . .	3.56 ±1.22	3.46 ±1.14	−0.07 (–0.31, 0.17)	. . .
Nation B
All stores	400	322		403	349
Placement/promotion of unhealthy items	3.30 ±0.85	3.15 ±0.66	0.15 (–0.24, −0.06)	3.24 ±0.79	3.11 ±0.68	–0.12 (–0.21, −0.03)	0.07 (–0.22, 0.35)
Placement/promotion of healthy items	3.22 ±1.05	3.39 ±0.83	0.17 (0.04, 0.29)	3.22 ±1.04	3.44 ±0.91	0.22 (0.10, 0.34)	0.06 (–0.33, 0.44)
Stores with a grill	202	172		398	347
Availability of healthy options at the grill	2.81 ±0.92	3.24 ±0.76	0.42 (0.29, 0.55)	2.70 ±0.89	3.34 ±0.80	0.64 (0.53, 0.75)	0.20 (–0.22, 0.62)
Grill promotes healthy options or nutrition information	2.75 ±0.88	3.05 ±0.58	0.27 (0.15, 0.39)	2.78 ±0.76	3.06 ±0.57	0.28 (0.19, 0.37)	−0.02 (–0.37, 0.32)
It costs more to buy the healthy options at the grill	3.82 ±1.28	3.51 ±1.05	–0.33 (–0.53, −0.14)	3.83 ±1.23	3.44 ±1.12	–0.39 (–0.54, −0.24)	−0.002 (–0.58, 0.57)

Note. CI = confidence interval.

^a(Postintervention score – preintervention score), unadjusted.

^b(In-person effect for intervention store shoppers – in-person effect for control store shoppers); adjusted for age, gender, education; domain scores range from 1 to 5, with higher scores indicating agreement.

Promotion recall and subsequent purchase of intervention items.

Compared with participants who shopped less frequently, intervention participants who shopped at their intervention store 3 or more times weekly were more likely to report noticing 1 or more intervention signs in Nation A (P = .05) and Nation B (P = .02; Table A, available as a supplement to the online version of this article at http://www.ajph.org). In Nation A, intervention participants who reported more frequent shopping noticed signs on endcaps (P = .03), reach-in coolers (P = .03), and grocery shelves (P = .01) more than did those shopping less frequently. Also, in Nation A, shopping frequency was related to purchases of healthy items from the reach-in cooler and grocery aisle foods marked with promotional signs. In contrast to Nation A’s frequent shoppers noticing the promotional signs on intervention food displays, Nation B’s frequent shoppers were more likely to notice the intervention food displays themselves than were less frequent shoppers. This was true for the endcap displays (P = .03), reach-in coolers (P = .02), and grocery aisles (P = .02). Nation B intervention store frequent shoppers were also more likely to report noticing promotional signs on reach-in coolers (P < .01). In Nation B, shopping frequency was related to the purchases of healthy food from reach-in coolers and items marked with promotional signs on grocery aisle shelves (P = .03).

Store-Level Outcomes

At baseline, there were no differences between intervention and control stores’ median NEMS scores, including food category and domain subscores, overall or by Tribe. Intervention store NEMS scores increased from baseline to follow-up in 6 of the food categories, all domains, and overall scores. Although not statistically significant, the median change in intervention store NEMS overall score was higher than was the median change for control stores (19.5 vs 4.0; P = .11). However, median changes in NEMS subscores were significantly higher in intervention stores than were control stores for fruits (5.3 vs 1.0; P = .03), canned lean meats (1.5 vs 0.0; P = .04), and the availability domain (9.0 vs −2.0; P = .01).

Ancillary Analyses

Process evaluations were largely collected according to protocol. Overall intervention fidelity was high across strategies in the intervention stores in both Nations. Promotional signage was present (95.5% of assessments). Fruit was usually present in fruit baskets (apples, 99.5%; oranges, 97.8%; and bananas, 83.0%). Salads (79.0%), apple packs (88.5%), fruit and yogurt parfaits (86.0%), berry cups (93.0%), and juices (90.5%) were usually available in coolers.

We assessed contamination of control participants shopping more than once a week at intervention stores. Contamination in the same Nation was higher for Nation A (11%–18% by control store) than for Nation B (< 5% for both control stores). Contamination of control participants shopping at the other Nation’s stores was low, with 5% or less reporting shopping at intervention stores in the other Nation.

DISCUSSION

We report on the, to our knowledge, first healthy retail intervention to be implemented in Tribally owned convenience stores. Results indicate that THRIVE increased both the availability and the purchasing of fruit, vegetables, and other healthy foods in both Nations. However, similar to other healthy retail studies,^13,14,31,32 fruit and vegetable intake did not increase from before to after the intervention in either Nation. The low fruit and vegetable consumption in our study population mirrors consumption in Oklahoma and the United States.³³ In addition, Tribal convenience stores are only part of the community food environment; hence, overall dietary intake may be influenced by the severely limited access to fruits and vegetables in both Nations.²² Future research should examine additional environmental influences and individual and community preferences and norms regarding dietary intake.

Limitations and Strengths

This study has a few noteworthy limitations. Beyond the Tribal stores, no other aspects of the community food environments were modified. Dietary screeners are less sensitive in detecting dietary intake than 24-hour dietary recalls,³⁴ likely reducing our ability to detect significant changes in fruit and vegetable intake. Additionally, because of vendor availability changes, some of the final intervention foods were not captured by the dietary screener, preventing us from assessing intake of these foods. Finally, during the course of the intervention, individuals from control and nonstudy communities in both Nations requested that the healthy foods be stocked in their stores, and several packaged snack items were introduced to nonstudy stores in both Nations.

We did not assess body mass index in this study. Previous studies reported that Native American shoppers were uncomfortable being weighed in store settings,^15,16 which THRIVE participants also expressed. The THRIVE study was the first time, to our knowledge, that health, commerce, and government leaders worked together in either Nation, and we were requesting significant access to the stores. We sought to balance the research needs with our priority of developing a long-term relationship with commerce and government leaders built on trust and reciprocity. The strong relationships built, coupled with the fact that the stores are all Tribally owned, eliminated the need to recruit individual stores and allowed us to address the absence of convenience store participation, which was cited as a limitation in previous research.¹⁶ The choice of data collected—health measures, shopping preferences, and perceptions of the stores—represent health, commerce, and government interests, respectively.

One of the most significant study outcomes is the increased demand for and access to healthy foods in both Nations. The process of securing healthy foods for the study shed light on the limited availability of these foods in the stores. To provide the requested study foods, the food distributors had to make supply changes, and they did so to avoid losing Tribal contracts. The expanded list of healthier options is a policy change that has increased the healthy options available not only to the Tribal convenience stores but also to all Tribal programs.

The cross-sector relationships developed as part of this study continue, fostering greater alignment between economic and health goals in both Nations. Future research should focus on multicomponent, multilevel interventions that can create sustainable Tribal food systems and incorporate health-promoting traditional foods into Tribally owned stores.

Public Health Implications

This study adds to the scare literature on healthy retail interventions in Tribal Nations. Study processes and findings may inform future interventions in the nearly 300 Tribal convenience stores across the United States and yield evidence for environmental and policy interventions to reduce obesity and improve Native American health.