“What Can I Eat?” Healthy Choices for American Indians and Alaska Natives with Type 2 Diabetes: Outcomes from a Randomized Waitlist-Controlled Trial of a Diabetes Nutrition Education Program

Abstract

Objective:

To evaluate the “What Can I Eat? (WCIE) Healthy Choices for American Indians and Alaska Natives (AI/ANs) with Type 2 Diabetes (T2D)” intervention.

Design:

Pilot randomized waitlist-controlled trial. Recruitment through diabetes registries and randomized to either the immediate intervention (n=35) or waitlist control group (n=25). Immediate arm started classes immediately upon randomization; waitlist arm started classes 3 months after randomization.

Setting:

Classes were taught synchronously online by registered dietitian nutritionists at 5 reservation-based or urban inter-tribal clinical sites nationwide in 2021.

Participants:

AI/AN adults with T2D (N=60)

Intervention:

Topics in AI/AN WCIE classes include: the diabetes plate, sugar-sweetened beverages, decreasing sodium, increasing consumption of healthful traditional Native foods. Class activities included didactic sessions, hands-on interactive learning, physical activity, mindful eating, and goal setting.

Main outcome measure:

Diabetes nutrition self-efficacy, behavior, and clinical measures.

Analysis:

Linear mixed models examined change in outcomes from baseline to 1 month and 3 months by randomization group. By 3 months, immediate intervention participants had completed the classes; the waitlist control group had not yet begun the intervention.

Results:

After 3 months, Confidence in Using the Diabetes Plate (β=0.80; 95% CI: 0.56-1.03, p<0.001) and Healthy Nutrition Behavior (β=0.88; 95% CI: 0.57-1.19, p=0.004) improved significantly in the immediate intervention group but not in the waitlist control group; Confidence in Making Healthy Nutrition Choices (β=0.65; 95% CI: 0.43-0.88, p=0.019) improved significantly more in the immediate intervention group than in the waitlist control group. No significant changes were identified in clinical outcomes.

Conclusions:

AI/AN WCIE enhanced self-efficacy and healthful nutrition choices among adults with T2D.

INTRODUCTION

Nutrition education, an essential component of a comprehensive diabetes education program, is associated with positive outcomes for people living with type 2 diabetes (T2D).¹ People living with T2D who take part in nutrition education classes experience improved dietary self-efficacy (i.e., confidence that one can follow a healthy diet), increased intake of healthy foods, and improved glycemia.^2,3 Nutrition education interventions are particularly effective when designed to address the needs of specific communities.⁴ To be effective, nutrition education must address culturally and regionally relevant foods, food acquisition habits, and typical food preparation practices.⁵ Specific to diabetes nutrition education, it is important to contextualize concepts such as “healthy carbohydrate foods,” “lean proteins,” and “healthy fats” using examples that are familiar and accessible to the learner.⁶ Further, it is critical that nutrition education resources be designed to directly address social drivers of health such as food or housing insecurity experienced by the priority community, which could impair engagement in recommended diabetes nutrition behaviors.⁷

Nutrition education programs designed for communities that are disproportionately impacted by poverty can improve intake of healthy foods through incorporation of a food resource management focus (i.e., meal planning, grocery shopping on a budget, home-based meal preparation) as a means to mitigate the negative health effects of food insecurity.^{8-12 8}Program adaptation and targeting of nutrition education and other health promotion interventions for cultural responsiveness and relevancy for members of specific racial and ethnic groups is particularly important to motivate community participation and achieve positive program outcomes.^13,14 Although the scientific literature supports the value of such interventions, no large-scale, diabetes-specific, nutrition education programs prioritizing the needs of American Indian and Alaska Native (AI/AN) adults have been developed in recent years. Given the high prevalence of T2D and related complications among AI/AN populations,¹⁵ as well as the considerable role food plays in traditional cultural practices in AI/AN communities,^16,17 the need for a diabetes nutrition education intervention that is culturally relevant for AI/AN people is compelling.^18,19

Seeking to address the paucity of updated, culturally appropriate nutrition education resources for AI/AN adults with T2D, the American Diabetes Association (ADA), in collaboration with the Shakopee Mdewakanton Sioux Community, provided grant funding to support the cultural adaption of the ADA’s intensive, diabetes nutrition program entitled, “What Can I Eat: Healthy Choices for People with Type 2 Diabetes and their Families” (WCIE) for AI/AN communities. In 2019-2020, our research team adapted the original WCIE curriculum to be appropriate for AI/AN individuals. The adaptation process involved a thorough review of the literature on educational programs for AI/AN people, focus groups with AI/AN adults living with T2D and their family members, key informant interviews with tribal leaders as well as clinical and education experts, and an evaluation of the health literacy burden of the original WCIE curriculum.^20,21 Throughout the adaptation process, the research team continued to obtain feedback from AI/AN community members, clinicians, and educators and iteratively evaluated health literacy-related and culturally adapted features of the WCIE curriculum for AI/AN populations.

The resultant “What Can I Eat? Healthy Choices for American Indians and Alaska Natives with Type 2 Diabetes” (AI/AN WCIE) program is a 5-session, classroom-based curriculum that addresses the learning preferences of many of AI/AN adults (e.g., visual and kinesthetic learning opportunities). Details on the AI/AN WCIE curriculum can be found in Table 1. The ADA has made the AI/AN WCIE curriculum available at no charge for AI/AN-serving organizations.²² Each class includes opportunities for individual or family-based goal setting.²³ Drawing on cultural strengths of AI/AN communities and guided by the findings of our qualitative focus groups and interviews,^24-26 the curriculum focuses on traditional foods, arts-based activities, and storytelling, and contains original photography of AI/AN community members engaging in healthful behaviors promoted in the curriculum throughout. In addition to designing the program to be culturally appropriate for AI/AN people from a variety of backgrounds (e.g., urban, reservation-based), the research team also sought to address key environmental factors that may make following a healthy diabetes diet and accessing diabetes nutrition education difficult for Native people. Specifically, the curriculum includes strategies explicitly designed to support communities that have high rates of household food insecurity and limited income,²⁷ that lack registered dietitian nutritionists (RDNs) who are trained to work with AI/AN and clients from other communities ,²⁸ and that experience transportation challenges that might interfere with class attendance. Because the AI/AN population is extremely heterogenous, we sought to make it easy for communities to tailor the contents to their local Native population and culture. For example, each lesson includes “placeholders” where class facilitators add examples of local, traditional foods. In addition, within several of the classes, there are “alternative” activities that can be interchanged based upon the preference of community participants, and there are ample peer-to-peer sharing opportunities for class participants to discuss local foods, resources, and cultural and community norms related to nutrition and food practices.

Table 1.

Details on “What Can I Eat?” Healthy Choices for American Indians and Alaska Natives with T2D Curriculum

Class*	Class Topic	Nutrition Lesson and Activity	Physical Activity	Mindful Nutrition Activity
1	Introduction to Carb Foods and the Diabetes Plate	• Carb identification game • Diabetes plate drawing activity • Focus on traditional carb foods	Group balloon toss (cardiovascular)	Mindful breathing before making nutrition choices
2	Sweet Foods and a Healthy Diabetes Diet	• Sweet foods and the diabetes plate • Sugar-sweetened beverage / food label activity with sugar packets • Traditional unsweetened beverage tasting activity	Resistance band strength training (strength)	Mindful visualization before making nutrition choices
3	Planning Healthy Diabetes Meals	• Healthy protein identification • Focus on traditional protein foods • Meal planning activity (grocery list, pantry inventory, calendar) • Facilitator-selected optional lessons: 1) Healthy eating on a budget; 2) Healthy eating with commodity foods; 3) Meal planning for small households; 4) Meal planning with kids	Walking (cardiovascular)	Mindful guided imagery to visualize healthy eating
4	Fat, Salt, and Healthy Eating Away from Home	• Sodium/salt label-reading and comparing common foods activity • Healthy fats • Eating healthy on-the-go: interactive class skits	Body weight resistance and stretching (strength, flexibility, balance)	Mindful walking to reduce autopilot nutrition decisions
5	Review, Reunion, and Celebration	• Diabetes plate review activity • Class review with BINGO • Long-term goal setting / vision board • Certificate of class completion	Dance train activity (cardiovascular)	Mindful guided chocolate kiss / jellybean eating activity

^*Classes 1-4 are offered weekly. Class 5 is offered 3 months after class 1.

The purpose of this paper is to report the outcomes of a pilot randomized waitlist-controlled trial of the AI/AN WCIE curriculum. We hypothesized that participation in the AI/AN WCIE program would result in improved self-efficacy related to eating healthy for people with diabetes, greater adherence to recommended nutrition and physical activity behaviors, and improved clinical outcomes.

METHODS

Study Design

Due to the COVID-19 pandemic, the adapted AI/AN WCIE curriculum pilot evaluation was offered online (instead of in person) as part of a waitlist-controlled trial at 5 reservation-based or urban inter-tribal clinical sites nationwide between January 2021 – December 2021. In this paper, we summarize the effect of participation in the program on AI/AN adults with T2D. Specifically, we report the effect of participation on diabetes-related self-efficacy, nutrition and physical activity behavior, and clinical outcomes (i.e., body mass index [BMI], hemoglobin A1c, blood pressure).

The study took place at 3 rural, reservation-based, tribally run, clinical sites in 3 states on the West Coast and East Coast of the United States and at 2 Urban Indian Health Organizations (UIHOs) in 2 states in the Midwest and Southern Plains of the US. Urban Indian Health Organizations are authorized through Title V of PL 94-437 of the Indian Health Care Improvement Act to serve the health care needs of urban AI/AN populations. Figure 1 depicts the location of each study site. Of note, this pilot randomized waitlist-controlled trial was scheduled to launch in January 2020. With the advent of the COVID-19 pandemic, all data collection and intervention delivery procedures had to be modified to be exclusively remote (e.g., home-based) per clinic “safer at home” and Institutional Review Board (IRB) requirements. This study was approved by the University of Colorado Multiple IRB (#19-2269), the National IRB at Indian Health Service (#N19-N-09), the Eastern Band of Cherokee Indians Medical IRB (MIRB #124), and reviewed by the Oklahoma City Area IHS IRB (who ceded approval to National IRB at Indian Health Service) prior to human subjects research commencing through expedited review. All participants completed informed consent with site-based co investigators via teleconference technology and mailed their signed consent form to their affiliated clinical site.

Figure 1.

Location of 5 sites involved in pilot evaluation of the AI/AN WCIE curriculum.

Participants and Recruitment

Study coordinators at each clinical partner site recruited participants with T2D using their local diabetes registries. The study team provided recruitment scripts for various methods of recruitment. For example, there was a short telephone script, a script for email query, and a script for mailed letters. Each site coordinator was able to select the recruitment method that was most congruent to the norm of how they typically communicate with their patients. All scripts contained information about the study including the benefits and risks should the potential participant be interested. Eligibility criteria included: 1) self-identification as AI/AN, 2) age ≥ 18 years, 3) ability to speak and read in English, 4) a current diagnosis of T2D, and 5) owning or having access to a smartphone or computer with internet access. The following exclusion criteria were applied: 1) diagnosis of a major medical issue that could interfere with participation (e.g., prior cardiovascular or cerebrovascular disease including a recent myocardial infarction or major stroke; cancer not in remission; dialysis; active alcohol and/or substance abuse), and 2) a planned move away from the area during the data collection period, which could interfere with completion of study visits. Participants were screened for eligibility by phone and participated in an online informed consent visit. Informed consent documents were mailed to the participants along with a stamped return envelope addressed to the local site coordinator or emailed to the participant if they hadaccess to a home printer. Participants mailed a paper signed consent form back to the site coordinator at their respective collaborating clinical site.

As shown in Figure 2, 83 people were screened for eligibility. Because study coordinators at the participating clinical sites only sought to enroll individuals they knew to be eligible and interested in participation, there were no screen failures and all participants who were screened consented to participate. Forty participants were randomized to the immediate intervention group, while 43 participants were randomized to the waitlist control group. Twenty-three participants were lost to follow-up for unknown reasons. Ultimately, 60 participants contributed data to this analysis, 35 from the immediate intervention group and 25 in the waitlist control group.

Figure 2.

CONSORT Diagram of Participant Enrollment, Allocation, Follow-Up, and Analysis.

Data Collection and Randomization

After providing informed consent and mailing in their signed consent form, participants received a “class kit,” which was mailed to their home addresses and included tools to collect evaluation data. These class kits included: a body weight scale (Weight Watchers by Conair up to 400 lbs.), a point of care A1C glucometer (Pts Diagnostics A1CNOW^® Self Check), a home blood pressure monitor (OMRON Bronze Blood Pressure Monitor, upper cuff Digital Blood Pressure Machine Model BP5100), and paper surveys, as well as all supplies to engage in the AI/AN WCIE classes (e.g., paper plates, markers, sugar packets, balloons).

During baseline data collection, participants were trained to use the tools in the class kit to measure their clinical outcomes during a one-on-one virtual visit through teleconference with the study coordinator or other clinic staff. Participants reported these data directly to the clinical site coordinator via phone or teleconference. To ease the data collection burden on participants, if body weight, blood pressures, and/or A1c were found in their medical records in the 30 days prior to the baseline data collection visit, those data were used in lieu of self-collection at home. Additionally, the most recent height measurement in the past 12 months was extracted from the medical record (for BMI calculation). If there was no recorded height in the medical record, self-reported height was used. Finally, each participant completed paper surveys, which they mailed back to their clinical site.

After collecting baseline data, participants at each site were randomized to either the immediate intervention group or the waitlist control group. The immediate intervention group started classes immediately upon randomization; the waitlist control group started classes 3 months after randomization. Classes #1-4 were taught weekly for 4 consecutive weeks; class #5 was taught 3 months after class #1. Each class lasted approximately 1 hour, though some went as long as 90 minutes. This depended on how much participants engaged and whether or not they had to leave the teleconference room at any particular time. Data on class participation were not collected. Fidelity of the delivery of the intervention was supported by 4 days of training for all of the RDNs who taught the classes (led by our research team of RDNs), provision of a script for teaching, and the research team’s observation of classes taught by each site-based RDN. Surveys and clinical data were collected at 3 timepoints for the immediate intervention group: baseline (before randomization), 1 month after beginning the intervention (after class #4), and 3 months after beginning the intervention (after class #5). Surveys and clinical data were collected at 5 timepoints for the waitlist control group: baseline (before randomization), 1 month after randomization, 3 months after randomization, 4 months after randomization (1 month after beginning the intervention; after class 4), and 6 months after randomization (3 months after beginning the intervention; after class 5). Classes were taught synchronously via teleconference by RDNs using a scripted facilitator guide. Figure 3 provides a graphic representation of the study design.

Figure 3.

Study design for pilot randomized controlled trial evaluation of AI/AN WCIE.

Paper surveys and all clinical outcome measures were collected at each timepoint, with the exception that A1c, which was not collected at the 1-month timepoint for either randomization arm or at the 4-month timepoint for the waitlist control group. To ease the burden of data collection on the participant, body weight and blood pressures could be obtained from the medical record if they were recorded within 7 days before or after the relevant data collection timepoint. For the 3-month and 6-month class sessions, an A1c from the medical record could be obtained if collected before or after 15 days of these classes.

Participants mailed paper surveys in pre-stamped, self-addressed envelopes back to each clinical site. The site coordinator then scanned these documents and shared the PDF with the research team, who entered the data into a REDCap (Research Electronic Data Capture) database.²⁹ These data collection methods were devised to meet COVID-19 safety protocols and to reduce the burden on research coordinators and other clinical staff at our partner sites, who experienced a high level of demands during the COVID-19 pandemic.

Measures

The AI/AN WCIE Impact Survey included questions addressing sociodemographic characteristics, diabetes nutrition self-efficacy, and nutrition and physical activity behaviors.³⁰

Sociodemographic Characteristics.

The AI/AN WCIE Impact Survey solicited information to assess the following sociodemographic characteristics: age, gender (e.g., male, female, other), highest grade completed, total household income before taxes in the prior year, household size, and employment status. These measures were used to describe the participant sample.

Diabetes Nutrition Self-efficacy.

The AI/AN WCIE Impact Survey included 12 items that assessed the degree to which participants felt confident that they could engage in the nutrition behaviors recommended in the AI/AN WCIE curriculum. Participants were asked: “How confident are you that you can…” (e.g., follow the diabetes plate). Items used a 4-point scale, ranging from not at all confident (0) to very confident (3). An earlier study provided strong support for reliability and validity of the diabetes nutrition self-efficacy items.³⁰ The item set produced two distinct factors reflecting (1) Confidence in Making Healthy Nutrition Choices and (2) Confidence in Using the Diabetes Plate. Both factors showed strong internal consistency reliability (standardized Cronbach alphas and McDonald omegas from 0.85 to 0.89) and – in convergent validity analyses – were significantly and positively associated with healthy nutrition behaviors. Confidence in Using the Diabetes Plate was also significantly and inversely associated with A1c.³⁰ In the current study, we examined change in both factor scores as a result of participation in the AI/AN WCIE intervention.

Diabetes Nutrition Behavior.

The AI/AN WCIE Impact Survey included 24 items evaluating the frequency with which participants engaged in the healthy and unhealthy nutrition and physical activity behaviors addressed by the AI/AN WCIE program. Participants were asked: “During a typical week in the last month, how often did you…” (e.g., fill half of your plate with vegetables). Items used a 6-point scale, ranging from never (1) to more than once a day (6). In earlier validation work, survey items addressing healthy nutrition behaviors showed strong internal consistency reliability (standardized Cronbach alpha = 0.89 and McDonald omega = 0.92) and a composite measure capturing Healthy Nutrition Behavior was significantly and inversely associated with A1c.³⁰ In this analysis, we examined change over time in the Healthy Nutrition Behavior score. Consistent with our prior work this score was computed as the average frequency with which participants engaged in 18 healthy nutrition behaviors identified in the Impact Survey.³⁰ Water intake was assessed using a self-reported measure asking participants how often they drank water or unsweetened beverages (more than once a day, once a day, 5-6 days a week, 3-4 days a week, 1-2 days a week, or never). Physical activity was assessed using a self-reported measure asking participants how many days a week they participated in 30 minutes or more of physical activity (0, 1, 2 ,3 4, 5, 6, 7 days).

Clinical Outcomes.

Using data that were collected and reported by participants, we examined BMI, systolic and diastolic blood pressure, and A1c. We calculated BMI as weight in kilograms divided by height in meters squared and analyzed systolic and diastolic blood pressure separately. We did not collect A1c more frequently than 3 month intervals because an A1C test measures the average of blood glucose level over the past 3 months and hence the ADA Standards of Care only recommends assessing glycemic status by A1c every 3 to 6 months.^31,32

Data Analysis

This study followed an intent-to-treat approach using randomization assignment as the independent variable. First, descriptive statistics were calculated for all study variables at baseline to describe the sample. T-tests and chi-squared tests of independence were used for continuous and categorical measures, respectively, to evaluate balance between randomization arms at enrollment. Linear mixed models with fixed effects for randomization assignment and time and random effects for participant were fitted to estimate the marginal effect of the intervention on changes in outcome measures from baseline to 1 month and 3 months for immediate intervention compared to waitlist control participants. Regression models adjusted for observed differences between randomization groups at baseline, which are described in the Results section. Last, a sensitivity analysis was conducted to examine the effect of the intervention on participants in the waitlist control group, comparing change in all outcomes of interest to baseline using paired t-tests. All analyses were conducted using Stata 18 (StataCorp. 2023. Stata Statistical Software: Release 18. College Station, TX: StataCorp LLC).

RESULTS

A total of 60 participants were randomized. Seventy-two percent of study participants were female, and the mean age for participants was 59 years (Table 2). Nearly half of participants (48%) were married or co-habiting; over half (56%) had completed at least some college, vocational or technical school; and 19% were college graduates or more. One third (32%) were employed, and 26% were retired. The average household size was 2.8 individuals. Forty percent reported an annual household income of at least $40,000. The two randomization arms were balanced on all demographic characteristics, with the exception of age; participants randomized to the immediate intervention group were significantly older than those randomized to the waitlist control group (61 years vs. 54 years, p<0.05). Diabetes nutrition self-efficacy, behavior, and clinical characteristics at baseline by randomization assignment are shown in Table 3. There were no significant differences between groups at baseline on any of these variables.

Table 2.

Sociodemographic Characteristics of Study Participants by Randomization Group at Baseline

Characteristic	Total	Immediate Intervention	Waitlist Control	p-value
Sex	% (n)	% (n)	% (n)
Male	28.3 (17)	25.7 (9)	32.0 (8)	0.594
Female	71.7 (43)	74.3 (26)	68.0 (17)
Other	0	0	0
Age, years
Mean (SD)	58.6 (11.1)	61.3 (9.0)	54.1 (13.0)	0.0202
Marital Status
Single	29.6 (16)	35.3 (12)	20.0 (4)	0.505
Married or living with someone in a marriage-like relationship	48.2 (26)	44.1 (15)	55.0 (11)
Widowed/Separated/Divorced	22.2 (12)	20.6 (7)	25.0 (5)
Educational Attainment
High school/GED or less	25.9 (14)	29.4 (10)	20.0 (4)	0.744
Vocational/technical school, some college, or associate degree	55.6 (30)	52.9 (18)	60.0 (12)
College graduate or beyond	18.5 (10)	17.7 (6)	20.0 (5)
Employment Status
Employed	31.5 (17)	29.4 (10)	35.0 (7)	0.911
Unemployed	42.6 (23)	44.1 (15)	40.0 (8)
Retired	25.9 (14)	26.5 (9)	25.0 (5)
Student	0	0	0
Household Size
Number of people living in your household, Mean (SD)	2.8 (2.3)	3.0 (2.7)	2.6 (1.4)	0.5662
Number under the age of 18, Mean (SD)	0.9 (1.6)	0.9 (1.9)	0.8 (1.1)	0.8797
Number over the age of 65, Mean (SD)	0.8 (1.0)	0.9 (1.0)	0.6 (1.1)	0.2658
Household Income
less than $15,000	25.6 (11)	33.3 (8)	15.8 (3)	0.141
$15,000-$39,999	34.9 (15)	20.8 (5)	52.6 (10)
$40,000-$74,999	25.6 (11)	33.3 (8)	15.8 (3)
$75,000 or more	14.0 (6)	12.5 (3)	15.8 (3)
Recruitment Site
1	28.3 (17)	28.6 (10)	28.0 (7)	0.784
2	16.7 (10)	14.3 (5)	20.0 (5)
3	26.7 (16)	31.4 (11)	20.0 (5)
4	26.7 (16)	22.9 (8)	32.0 (8)
5	1.7 (1)	1	0

P-values for categorical variables (i.e., sex, marital status, educational attainment, employment status, household income, site) are derived from Chi-squared and Fisher’s exact tests; p-values for continuous variables (e.g. age, household size and composition) are derived from t-tests; alpha=0.05.

Table 3.

Baseline Diabetes-Nutrition self-efficacy, behaviors, and clinical characteristics by Randomization Group

		Immediate Intervention	Waitlist Control
Characteristic	Total n	n=35 M (95% CI)	n=25 M (95% CI)	p-value
Confidence in using the Diabetes Plate	54	1.6 (1.3-1.9)	1.0 (1.6-2.2)	0.205
Confidence in Making Healthy Nutrition Choices	54	1.9 (1.6-2.1)	1.8 (1.5-2.1)	0.775
Healthy nutrition behaviors	54	2.8 (2.4-3.1)	3.2 (2.8-3.6)	0.114
Water intake	46	1.6 (1.2-2.0)	1.3 (0.8-1.7)	0.347
Days per week with at least 30 min. of physical activity	53	3.3 (2.5-4.1)	3.1 (2.0-4.2)	0.792
A1C	54	7.8 (7.1-8.6)	7.4 (6.7-8.0)	0.383
Systolic blood pressure	55	129.6 (123.6-135.6)	130.6 (121.8-139.4)	0.843
Diastolic blood pressure	55	77.9 (74.7-81.0)	79.7 (74.7-84.7)	0.503
Body Mass Index (BMI)	55	33.3 (31.3-35.2)	34.9 (30.7-39.2)	0.406

Possible range of responses: Confidence in using the Diabetes Plate (0-3); Confidence in making healthy nutrition choices (0-3); Healthy nutrition behaviors (1-6); Water intake (1-5) ; Days per week with at least 30 min. of physical activity (0-7); p-values are derived from t-tests using alpha=0.05.

Results from linear mixed models estimating the effect of the AI/AN WCIE intervention on diabetes nutrition self-efficacy and behavior as well as clinical characteristics adjusting for age at baseline are shown in Table 4. Estimates in Table 4 display marginal effects from post-estimation while p-values are derived from the interaction terms (e.g., time x randomization assignment) in the corresponding linear mixed models. There were no statistically significant changes over time in the waitlist control group for Confidence in Using the Diabetes Plate, Healthy Nutrition Behaviors, water intake, physical activity, or any of the clinical characteristics. Among those randomized to the immediate intervention, however, there were statistically significant increases over time in Confidence in Using the Diabetes Plate at both Month 1 (β=0.70; 95% Confidence Interval (CI): 0.47-0.92, p<0.001) and Month 3 (β=0.80; 95% CI: 0.56-1.03, p<0.001). In other words, the participants in the immediate intervention group had an improvement of 0.7 points in their Confidence in Using the Diabetes Plate from baseline to 1 month post-baseline. Meanwhile, those randomized to the waitlist control group only had an increase of 0.02 points in their Confidence in Using the Diabetes Plate, which is not statistically significant (β=0.02; 95% CI: −0.22-0.26, p>0.05). Comparing the two groups, the immediate intervention group had significantly more improvements in their Confidence in Using the Diabetes Plate than the waitlist control group at both Month 1 and Month 3 (P<0.001). Those randomized to immediate intervention also demonstrated a significant change over time in Confidence in Making Healthy Nutrition Choices at Month 1 (β=0.40; 95% CI: 0.19-0.62, p=0.008) and Month 3 (β=0.65; 95% CI: 0.43-0.88, p=0.019), and this was greater than the improvement observed among the waitlist control group at Month 3 (β=0.27; 95% CI: 0.04-0.50, p=0.024). Immediate intervention participants also had significantly increased Healthy Nutrition Behavior scores at both timepoints (Month 1: β=0.54; 95% CI: 0.24-0.83, p=0.033).; Month 3: β=0.88; 95% CI: 0.57-1.19, p=0.004). Among those randomized to the immediate intervention, there was a significant change over time in physical activity at Month 1 (β=0.92; 95% CI: 0.10-1.74, p=0.033), but this did not remain significant at the p<0.05 at Month 3. There were no significant changes over time for either group in A1c, water intake, systolic or diastolic blood pressure, or BMI. Of note, as indicated in Supplementary Table 1, among waitlist controls, A1c dropped from 7.58 at baseline to 6.94 at 3 months, and this was significant (p=0.007). However, we believe this could be due to placebo effect or simply due to the small sample size.

Table 4.

Change over time on Diabetes Nutrition Self-efficacy, Behaviors, and Clinical Characteristics by Randomization Group

Characteristic by Timepoint	Immediate Intervention	Waitlist Control	p-value
Confidence in using the Diabetes Plate
Month 1	0.70 (0.47-0.92)	0.02 (−0.22-0.26)	<0.001
Month 3	0.80 (0.56-1.03)	0.01 (−0.24-0.25)	<0.001
Confidence in making healthy nutrition choices
Month 1	0.40 (0.19-0.62)	−0.03 (−0.27-0.20)	0.008
Month 3	0.65 (0.43-0.88)	0.27 (0.04-0.50)	0.019
Healthy nutrition behavior
Month 1	0.54 (0.24-0.83)	0.08 (−0.22-0.38)	0.033
Month 3	0.88 (0.57-1.19)	0.25 (−0.05-0.55)	0.004
Water intake
Month 1	0.02 (−0.36-0.40)	−0.14 (−0.58-0.31)	0.606
Month 3	0.07 (−0.32-0.46)	0.10 (−0.34-0.53)	0.914
Days per week with at least 30 min. of physical activity
Month 1	0.92 (0.10-1.74)	−0.36 (−1.21-0.49)	0.033
Month 3	1.09 (0.23-1.94)	−0.05 (−0.91-0.80)	0.064
A1C
Month 3	−0.10 (−0.71-0.51)	−0.47 (−1.09-0.15)	0.408
Systolic blood pressure
Month 1	2.53 (−4.40-9.47)	1.13 (−6.30-8.56)	0.787
Month 3	−1.88 (−9.11-5.36)	0.51 (−7.07-8.10)	0.655
Diastolic blood pressure
Month 1	0.14 (−3.52-3.80)	1.97 (−1.94-5.88)	0.504
Month 3	−0.99 (−4.81-2.82)	1.28 (−2.72-5.27)	0.421
Body Mass Index (BMI)
Month 1	−0.25 (−0.68-0.20)	−0.15 (−0.62-0.32)	0.767
Month 3	−0.31 (−0.78-0.15)	0.16 (−0.32-0.64)	0.162

P-values are derived from the interaction terms (e.g., time x randomization assignment) in the corresponding linear mixed models with alpha=0.05.

Results from the sensitivity analysis examining change in clinical outcomes at all 5 timepoints among participants in the waitlist control group are shown in Supplemental Table 1. There was no change over time in Confidence in Using the Diabetes Plate or Confidence in Making Healthy Nutrition Choices prior to the beginning of the intervention (Months 1 and 3), but these scores significantly increased at Month 4, after the intervention began, and remained significantly higher than baseline at Month 6 (after the conclusion of the intervention). There were no other significant differences over time observed in any behavioral or clinical outcomes, with the exception of significantly lower A1c levels observed at Month 3 compared to baseline, which were not observed at Month 6. We also analyzed those who completed follow-up compared to those who dropped out of the study. There were no statistically significant differences between those who dropped out and those who remained in the study in terms of age, marital status, gender, employment status, educational status, income or household size or composition.

DISCUSSION

Findings from this pilot randomized waitlist-controlled trial of the AI/AN WCIE curriculum suggest that the intervention was successful in improving diabetes nutrition self-efficacy related to use of the diabetes plate and making healthful food choices. Self-efficacy is a key predictor of healthful nutrition behavior among people living with diabetes.^33-35 The AI/AN WCIE curriculum is informed by the Expanded Health Belief Model, which includes self-efficacy as a predictor of behavior change,³⁶ and appears to be effective in improving confidence that one can engage in healthful diabetes nutrition behaviors. The intervention also was successful in increasing the frequency with which participants engaged in healthy nutrition behaviors recommended for people living with T2D. As supported by the literature, one of these key nutrition behaviors was use of the diabetes plate for meal planning.^37-39 The ADA’s 2019 consensus report on nutrition therapy⁴⁰ and the ADA Standards of Care⁴¹ states that medical nutrition therapy and nutrition education is fundamental to overall diabetes management plans and that cultural preferences should always be taken into account. The Indian Health Service Standards of Care for Diabetes, directly adapted from ADA Standards of Care, also supports integration of culturally relevant diabetes education resources.⁴² A culturally tailored intervention, such as AI/AN WCIE, that can enhance diabetes nutrition self-efficacy and behavior may provide a valuable resource for AI/AN-serving organizations across the country.

Though no significant changes were observed in clinical outcomes (e.g., A1c, blood pressure), it is important to contextualize these null findings. This pilot randomized waitlist-controlled trial was scheduled to begin just as the COVID-19 pandemic began. Because AI/AN communities were disproportionately impacted by COVID-19 morbidity and mortality,^43,44 and participants were all people living with T2D who have a higher risk for COVID-19 complications ^45,46 and mortality,^47,48 the research team was exceedingly careful to refine study procedures in ways that would protect our participants and partner communities. Drastic changes were made to both the curriculum delivery (e.g., changing from in-person to Zoom-based) and evaluation procedures (e.g., home-based and self-reported measures). It is possible that home-based collection of clinical data was less accurate than data collected in a clinical setting, perhaps resulting in failure to see changes in clinical outcomes over time, or it could be that there was simply no difference. Further, it may be that 3 months is too short of a period of time to see clinical changes. Literature suggests home-based A1c testing kits are accurate,⁴⁹ and especially during COVID-19, there was much consideration about telemedicine and reliability of home-based monitoring.⁵⁰ Additionally home-based blood pressure monitoring can be reliable,⁵¹ and have been determined to be reliable and valid in making new diagnoses of hypertension.⁵² However, there can large variations in accuracy between patients and more work needs to be done to establish reliability of clinical outcomes through telehealth.⁵³ Another study suggested home-based A1c testing underestimates A1c as compared to reference standard, however these home-based tests have the potential to expand research and clinical care, especially for underserved communities.⁵⁴ With regard to our outcomes, it is also possible that the small sample size reduced our ability to see changes in these indicators. Indeed, based on our previous power analysis, we will need a total sample size of 150 to have 80% power for detecting a different of 0.4% in A1C between the two groups. Since the current pilot study only have a sample size of 54 for A1C, this study is under-powered to detect a clinically significant difference in A1C. Future trial with larger sample sizes are needed to formally test the effects of our interventions on clinical outcomes.

The COVID-19 pandemic made diabetes self-management even more complicated than it had been in pre-pandemic times. Literature on diabetes and the COVID-19 pandemic suggests the need for modified diabetes self-management and support protocols^55-57 and the importance of social support for people living with diabetes during COVID-19.⁵⁸ Qualitative findings from this AI/AN WCIE pilot intervention suggest that participants had high satisfaction with the program and specifically appreciated the peer-to-peer interaction opportunities.⁵⁹ Given that remote, online, group-based health education was rare prior to COVID-19, it may be an unintended positive finding that this method of class delivery can mitigate some of the traditional challenges to attending in-person group-based education. Challenges with transportation, childcare, eldercare, and the lack of RDNs in Native communities all complicate AI/AN people’s ability to attend group-based classes.⁶⁰ Another limitation includes the decreased sample size as a result of study dollars being allocated to accommodate home-based data collection (e.g., mailed class kits that participants received at home for data collection) and costs of shipping. Because of these unanticipated costs related to COVID-19, the sample size was much smaller than intended (e.g., original goal sample was 300 patients across 5 collaborating sites). We did not conduct post experimental power analyses to estimate observed power given this practice is contrary to recommendations,⁶¹ and is already a function of the observed p-values. In addition, we did not track which clinical data (e.g., weight, height, blood pressure, A1c) were collected via home-based self-report or via chart abstraction from an electronic health record. Therefore, we cannot assess the impact that data collection method may have had on data quality or study results. Though each participant was invited to complete a “post class” satisfaction survey after every class,⁶² this was optional, and we do not have reliable data on rates of participations within the classes so are unable to report accurately on this metric. Finally, although intervention participants showed significant improvement in Healthy Nutrition Behavior, the healthy behavior of drinking water or unsweetened beverages did not show similar improvement. It is possible, however, that the scale used to measure intake of healthy beverages was insensitive in change. Because the scale captures consumption of water and unsweetened beverages only up to the frequency of “more than once a day,” we were not able to identify higher levels of consumption nor improvements in intake at these higher levels.

As aligned with ADA and Indian Health Service recommendations for standards in diabetes care, ^42,63 it is important to emphasize the cultural relevance of the AI/AN WCIE curriculum. The diabetes plate³⁹ is a visual tool that can easily be culturally tailored by picturing and offering culturally relevant foods for each of the plate sections, which also allows class facilitators to accommodate local and regional differences across AI/AN communites.^38,39 Further cultural adaptation of the AI/AN WCIE curriculum as supported by traditional ecological systems/knowledge ^64-66 includes robust peer-to-peer interaction and sharing opportunities, prioritizing traditional foods and traditional foodways, family- and community-based healthy eating support, and teaching strategies such as arts-based drawing and storytelling. Additional details of the cultural adaptations in this curriculum are forthcoming.⁶⁷ As supported by the literature,^25,68-70 the inclusion of these strengths-based, culturally relevant concepts very likely contributed to participants’ high satisfaction with the AI/AN WCIE.⁵⁹

IMPLICATIONS FOR RESEARCH AND PRACTICE

The ADA has generously made the AI/AN WCIE curriculum available free of charge to all tribal-serving organizations.²² The unique “content tailoring” features (e.g., placeholders for facilitators to add local food examples) allow AI/AN WCIE to honor the heterogenous nature of AI/AN communities. The impact survey has been validated to help evaluators understand the outcomes of the AI/AN WCIE curriculum for their community members living with T2D.³⁰ Future plans with the AI/AN WCIE program are to include expanding evaluation to peer-educator delivery models, adding more lessons by shortening several of the 5 lessons and splitting them into two, and expanding to other AI/AN individuals such as those with prediabetes and/or gestational diabetes.

Supplementary Material

1

Supplemental Table 1. Mean Scores at Each Timepoint for Waitlist Control Participants with Paired t-test, Comparing Each Timepoint to Baseline