Examining differences in diet quality between Canadian Indigenous and non-Indigenous adults: results from the 2004 and 2015 Canadian Community Health Survey Nutrition Surveys

Abstract

Objectives

The Truth and Reconciliation Commission includes a call to action to close gaps in health outcomes, including type 2 diabetes, of which diet quality must be considered an important mediator. The objectives of this study were to compare diet quality between off-reserve Indigenous and non-Indigenous adults in 2004 and 2015, and examine food security as a predictor of diet quality.

Methods

We employed a repeated cross-sectional design using the 2004 and 2015 Canadian Community Health Surveys-Nutrition. Both surveys include a representative sample of the Canadian population in the 10 provinces, excluding the northern territories and people living on-reserve. Healthy Eating Index (HEI) values were estimated, using 24-hour dietary recalls, for the Indigenous and non-Indigenous population in each time period. After matching, a generalized linear model was applied to test for differences in HEI between groups across time period, adjusting for household food security.

Results

Overall, HEI scores were not significantly different for Indigenous men and women in 2015 as compared with 2004, but continued to be lower compared with those of the non-Indigenous population. Indigenous adults reported significantly lower diet quality independent of food security status and other factors. Scores pertaining to percent energy from ‘other’ foods improved in 2015 compared with 2004.

Conclusion

Dietary disparities persist between Indigenous and non-Indigenous populations. While addressing household food insecurity among Indigenous populations is necessary to improve diet quality, it is not sufficient. Results suggest that factors other than food insecurity and socio-economic status are impacting disparities in diet quality among Indigenous adults.

Introduction

First Nations, Métis, and Inuit health inequities are a result of Canada’s colonial history and present (Truth and Reconciliation Commission, 2015). Type 2 diabetes is one of these health inequities, though diabetes does not impact all Indigenous1 peoples equally. Prevalence of diabetes varies considerably among First Nations (Green et al., 2019; Ruth et al., 2020), Métis (Martens et al., 2010; Randall et al., 2019; Shah et al., 2010), and Inuit populations (Public Health Agency of Canada (PHAC), 2018). The Public Health Agency of Canada estimates prevalence of self-reported diabetes between 2008 and 2013 at 19.0% for First Nations adults living on-reserve and in the north, 12.7% for First Nations living off-reserve, 9.9% for Métis adults, and 4.7% for Inuit adults, as compared with 6.9% for non-Indigenous adults (PHAC 2018). Furthermore, prevalence of physician-diagnosed diabetes among First Nations and Métis populations varies between provinces (see as examples: Shah et al., 2010; Martens et al., 2010; Randall et al., 2019), and among First Nations populations evidence shows that prevalence varies further among communities within a province (see Halseth, 2019 for a review of diabetes among First Nations; Green et al., 2019; Ruth et al., 2020). Prevalence of diabetes is also higher among First Nations women than among First Nations men, which is in contrast to the non-First Nations population (Green et al., 2019; Ruth et al., 2020). Data from the 2012 Aboriginal Peoples Survey also indicate considerable variation in diabetes prevalence between Inuit regions, with the highest prevalence estimated at 7% among Inuit living outside Inuit Nunangat (Wallace, 2014), though the author notes prevalence should be ‘used with caution’ due to small sample size. Variation in diabetes prevalence among Indigenous populations is due to a variety of complex factors related to colonialism; among these factors are diet and food security.

Diet is important not only in the prevention and management of diabetes (Diabetes Canada Clinical Practice Guidelines Expert Committee, 2018), but also for health, generally. Diet quality, commonly measured using the Healthy Eating Index (HEI), is a composite indicator of the variety, adequacy, moderation, and balance of an individual’s diet with respect to key food groups and nutrients, which has been adapted to the 2007 Canada’s Food Guide (Garriguet, 2009). Disparities in diet quality between off-reserve Indigenous and non-Indigenous people 12 years of age and older have been described using the 2004 Canadian Community Health Survey (CCHS) Nutrition survey (Garriguet, 2009). Notably, this difference in diet quality was no longer significant after adjusting for other factors, though food security was not investigated as a co-variate. The examination of food groups in a separate study revealed significant disparities, particularly between Indigenous and non-Indigenous women (Garriguet, 2008); intakes of grain products and fruits and vegetables were significantly lower among Indigenous women as compared with non-Indigenous women.

Food insecurity refers to “inadequate or uncertain access to food because of financial constraints” (Health Canada, 2017, p 13). Indigenous people living off-reserve continue to have higher odds of household food insecurity, independent of household factors (McIntyre et al., 2014; Tarasuk et al., 2019). Food insecurity is associated with incident type 2 diabetes (Tait et al., 2018) and greater risk for premature mortality compared with food-secure Canadians (Men et al., 2020). There are multiple mechanisms, or pathways, by which food insecurity negatively influences health and well-being of Indigenous people (for examples see Cidro et al., 2015 and Richmond et al., 2021) and nutritional status among them. Using the 2004 CCHS, Kirkpatrick and Tarasuk (2008) described multiple nutrient and food group inadequacies among food-insecure Canadians compared with those who were food secure, particularly for intakes of fruits and vegetables, and milk products. Among Indigenous adults, fruit and vegetable intake was significantly lower among food-insecure adults compared with those reporting food security (Willows et al., 2011). However, the relationship between food security and overall diet quality, as measured using the HEI, among Indigenous people has not been investigated, nor the extent to which food security accounts for disparities in diet quality between Indigenous and non-Indigenous populations.

The current federal government has committed to addressing all 94 of the Truth and Reconciliation Commission (TRC) Calls to Action (2015). Number 19 includes a call to action to close gaps in health outcomes, including type 2 diabetes. While changes in the prevalence of type 2 diabetes will undoubtedly take time given their chronic nature, changes in diet quality are more acute. In this way, improvements in diet quality among Indigenous people may result in improved health outcomes, potentially contributing to reductions in the disparity of type 2 diabetes, as called for in the TRC. Given previous associations between food security and nutritional inadequacies, examining the relationships of food security with diet quality is critical. Therefore, the main purpose of this study was to compare diet quality between Indigenous and non-Indigenous adults during two periods, 2004 and 2015. A secondary objective was to examine food security as a predictor of diet quality.

Methods

Design

We utilized a repeated cross-sectional design using both the 2004 and 2015 CCHS Nutrition surveys. Both surveys include a general health survey and a single 24-hour dietary recall, with a subset completing a second 24-hour dietary recall. The surveys include basic demographic and socio-economic variables as well as the 18-item Household Food Security Survey Module (HFSSM). The surveys have been described in greater detail elsewhere (Health Canada, 2006a, 2017). This analysis was approved by the Social Sciences and Humanities Research Council, which allowed data access to the Research Data Centre and precluded institutional Research Ethics Board approval.

This study was conducted in partnership with the National Indigenous Diabetes Association (JL), who co-led the application for funding, developed and refined the study objectives, informed the analysis throughout, interpreted the study findings, and provided ongoing leadership in the dissemination of the findings. This occurred through regular face-to-face meetings with the study team. The partnership was grounded in the principles of self-determination, equity, reciprocity, and capacity building, as outlined in our shared research agreement.

Sample

The 2004 and 2015 CCHS Nutrition surveys include a representative sample covering approximately 98% of the Canadian population (≥1 year old) in the 10 provinces; northern territories were excluded (Yukon, Northwest Territories, and Nunavut). All those living on-reserve or in institutions, full-time members of the armed forces, and those living in remote areas were excluded. In 2006, a census year, Indigenous people accounted for 3.8% of the Canadian population; by 2016, they accounted for 4.9% of the population, representing a 42.5% increase. Despite a substantial number of Indigenous people excluded in the survey sampling frames, a sizeable population of Indigenous people remained eligible for the survey (Statistics Canada, 2017b). Métis were least likely to be excluded in the survey sampling frames given that approximately 99% reside in the provinces. In 2016, there were 587,545 Métis people in Canada, and the population grew considerably between 2006 and 2016 due to both natural growth and increased self-identification (Statistics Canada, 2017b). In 2016, an estimated 55.8% of the 744,855 registered First Nations people in Canada lived off-reserve, with population growth being greater off-reserve compared with on-reserve between 2006 and 2016 (Statistics Canada, 2017b). Furthermore, there were an additional 232,375 First Nations people without treaty status in 2016. The majority of Inuit (72.8% of 65,025) live in Inuit Nunangat and therefore were not included in the CCHS; however, the number of Inuit living outside Inuit Nunangat is increasing at a greater rate (Statistics Canada, 2017b).

Generally, the sampling methods were similar between the two surveys and included a multistage stratified cluster design (Health Canada, 2006a; Statistics Canada, 2017a). However, Indigenous people aged 19–50 years old were purposely oversampled in the 2004 CCHS (Health Canada, 2006a), and some provinces purchased additional samples. This contributed to a smaller sample in 2015 as compared with 2004. The sample size for the 2004 CCHS is 35,107, with a 76.5% response rate for the survey, including 1528 who self-identified as Indigenous and 10,786 who completed a second dietary recall. The sample size for the 2015 CCHS is 20,487, with a 61.6% response rate (Health Canada, 2017a), including 950 who self-identified as Indigenous and 7608 who completed a second dietary recall. An Indigenous-specific response rate is not available. All non-pregnant, non-breastfeeding adults aged ≥18 years old in each survey year were included in this study analysis (n=31,957).

Indigenous includes all participants who self-reported sole or mixed Indigenous ancestry and all other participants were considered non-Indigenous. Indigenous is further categorized as self-identified Métis, Inuit, or First Nations, regardless of status or registration under the Indian Act. Individuals who identified as both Métis and First Nations were categorized as First Nations. Due to limited sample size and disclosure avoidance, results pertaining to number of Inuit participants could not be reported, but were included in the Indigenous sample. Where possible, we have completed analysis specific to First Nations and Métis populations, though opportunity was limited to produce reliable estimates given the small sample sizes when disaggregated.

Measures

Variables included are sex, age group, time period, highest level of household education, income adequacy, and household food security status. Sex is dichotomized as male and female. Age was included as a continuous variable. Time period is dichotomized as 2004 and 2015, according to the survey years. Highest level of education is divided into three groups: less than secondary school graduation, secondary school graduation, and post-secondary education. Income adequacy, as defined by Statistics Canada using total household income and number of individuals in the household, is classified into four categories (Health Canada, 2006a), and then dichotomized as was previously applied to the 2004 survey (Garriguet, 2008). We did not adjust for inflation.

Food security status is operationalized using the HFSSM, an 18-question module that asked respondents about their food security situation in the household over the previous 12 months, with households categorized as food secure, moderate-food insecure, and severe-food insecure (Health Canada, 2008). Moderate and severe food insecurity were then combined given the comparatively small cell sizes when grouped separately. The CCHS did not provide household survey weights; as such, we are unable to provide estimates of household food insecurity, rather we are reporting food insecurity using individual weighting, as described in the analyses section below.

HEI is considered a robust estimate of diet quality, useful for the estimation of population diet quality, comparison of population groups, and assessment of time period trends (Garriguet, 2009). HEI was calculated based on the method previously described by Garriguet (2009), which is an individual score out of 100 based on age- and sex-specific dietary recommendations, adapted to the 2007 Canadian dietary recommendations. Briefly, component scores were generated for each of the following dietary components (maximum points allotted) and summed: total fruits and vegetables (10 pts), whole fruit (5 pts), dark green and orange vegetables (5 pts), total grain products (5 pts), whole grain products (5 pts), milk and alternatives (10 pts), meat and alternatives (10 pts), unsaturated fat (10 pts), saturated fats (10 pts), sodium (10 pts), and percent energy from ‘other’ food (20 pts). Saturated fat, sodium, and percent energy from ‘other’ foods are reverse-scored.

Statistical analysis

First, we sought to characterize the Indigenous sample according to First Nation and Métis identifiers in each time period. Statistics Canada had indicated the number of Canadians identifying as Métis had increased over time (Statistics Canada, 2017b). Therefore, describing these proportions is critical to understanding any changes in diet quality over time for the Indigenous population. Second, we conducted a descriptive analysis of the socio-economic variables stratified by (non-)Indigenous identifiers and time period, including proportion and standard error. HEI is reported as mean and standard error given its normal distribution. Usual HEI and component scores were calculated with the Software for Intake Distribution Estimation (SIDE) program for the 2004 CCHS (Health Canada, 2006a) and using the National Cancer Institute method for the 2015 CCHS (National Cancer Institute, 2020). Both methods are used to estimate the distribution of usual intake for a population or subpopulation using the subsample with two dietary recalls. Importantly, the SIDE and NCI adjustments were applied to the overall sample rather than to the Indigenous and non-Indigenous samples separately, as the Indigenous subsample with two recalls was too small. This analysis assumes within-person variation is similar between both populations.

We applied coarsened exact matching (CEM) to match Indigenous to non-Indigenous participants according to age groups used for Dietary Reference Intakes (DRI) (Health Canada, 2006b), sex, education, income adequacy, and time period. Given the limitations of working with observational data, matching was used to create a more comparable comparison group, limiting the influence of unobservable confounders. CEM matches observations based on strata generated by the categorical variables used in the matching procedure (Jeon, 2014). Treated observations are then matched to control observations within the same strata. Observations in strata containing only ‘treatment’ or only ‘control’ variables are pruned. Finally, remaining control observations are reweighted using weights described by Iacus et al. (2012). Pruning unmatched observations and reweighting the remaining observations balances the empirical distribution of the variables used in the matching procedure, leading to a better comparison between the groups. The CEM weights were then multiplied by the survey weights, and bootstrapping was applied. The matched samples were described according to (non-)Indigenous identifiers for each of the demographic and socio-economic variables matched on to ensure successful matching. We have previously applied CEM to dietary data (Mudryj et al., 2021).

Using a General Linear Model (GLM) with HEI as the outcome variable, we tested for Indigenous*time period interaction effects, further adjusting for the main effects, energy intake, and household food security. Energy intake was included given the documented increase in under-reporting of energy intake in 2015 as compared with 2004 (Garriguet, 2018). For comparison, we completed the same analysis with the unmatched sample to examine the implications of matching, adjusting for age, sex (where relevant), income adequacy, highest level of household education, energy intake, and food insecurity. Further models were completed with subgroups (e.g., Indigenous people, men only) to tease out the reasons for any significant interaction effects. Using the matched sample, we also conducted a descriptive analysis of the component scores of HEI by Indigenous identifier and time period. T-tests were used to test for differences between Indigenous and non-Indigenous adults within each time period.

We were unable to conduct First Nations and Métis-specific analysis by time period due to small sample sizes. However, we did pool the 2004 and 2015 survey data and applied a GLM to test for differences in HEI according to non-Indigenous, First Nations, and Métis identifiers, adjusting for age, sex, education, income, time period, and food security status. This unmatched analysis was completed to inform our interpretation of the results according to time period given the different proportions of Métis and First Nations people in each survey. We cannot assume diet quality is similar or different between First Nations living off-reserve and Métis; we have limited nutritional data on which to base any assumption. However, First Nations people living off-reserve have higher prevalence of diabetes compared with Métis, which could suggest that there may be differences in diet quality. Since the proportions of First Nations and Métis people within the Indigenous population changed between 2004 and 2015, as well as in the surveys, this should be considered in the interpretation of the results. Our interest is not to test for differences between First Nations and Métis, but rather both groups as compared with non-Indigenous.

STATA statistical software package version 14 (Stata Corp., College Station, TX, USA) was used and significance was set at α=0.05. A sex- and gender-based analysis was employed such that analyses were stratified by sex. Standard errors and confidence intervals were estimated using the bootstrapping procedure to provide estimates that best reflect the population.

Results

The proportion of the sample who identified as Indigenous was higher in 2015 as compared with 2004. There was a greater proportion of Métis participants as compared with First Nations in 2015 relative to that in 2004 (Table 1). In 2004, 35.1% of the Indigenous sample were Métis and in 2015 that proportion rose to 57.5%. Importantly, these findings generally reflect population proportions eligible for the survey (Statistics Canada, 2017b).

Table 1.

Proportion of adult samples (≥ 18 years old) who self-identify as Indigenous and non-Indigenous

	Indigenousa		Non-Indigenous
	Métis	First Nations
2004 (n=20,885)	35.1 (5.0)	62.5 (5.3)	98.2(0.1)
2015 (n=11,072)	57.5 (4.2)	42.5 (4.4)	96.0 (0.1)

^aResults are presented as a proportion (SE) of the Indigenous sample; Inuit participants are included within the Indigenous sample of the survey but could not be reported on due to small sample sizes and concerns related to disclosure avoidance

A total of 1384 adult participants identified as Indigenous in both surveys; after matching, 60 Indigenous participants were pruned compared with 2277 non-Indigenous participants (Table 2). In unmatched descriptive analysis, HEI was not significantly different in 2015 than in 2004 for Indigenous men, women, and all Indigenous adults combined (Table 2). On the other hand, HEI scores were significantly lower in 2015 than in 2004 for non-Indigenous men, women, and all non-Indigenous adults combined.

Table 2.

Socio-economic status, household food security, and diet quality among Canadian adults in the 2004 and 2015 CCHS Nutrition Surveys

Variables	Indigenousa			Non-Indigenousa			Pooled by year and matchedb
	2004	2015	p-value	2004	2015	p-value	Indigenous	Non-Indigenous
All adults (n)	826	558		20,059	10,513		1324	28,295
Highest level of household education
< Secondary school	20.0 (2.2)	14.0 (3.0)	0.270	10.0 (0.4)	7.7 (0.6)	0.365	16.7 (2.1)	15.7 (6.7)
Secondary graduation	16.8 (4.0)	25.4 (4.3)		11.1 (0.3)	18.0 (0.7)		23.2 (2.4)	18.8 (4.6)
Post-secondary education	62.1 (3.3)	59.8 (3.9)		78.0 (1.0)	74.2 (1.1)		60.0 (2.8)	65.4 (11.2)
Income adequacy
Low	22.5 (4.0)	11.0 (3.0)	<0.05	10.5 (0.5)	10.4 (0.4)	0.836	11.1 (1.5)	16.1 (9.9)
Middle/high class	77.4 (1.2)	87.0 (3.5)		89.2 (0.8)	89.0 (0.7)		88.9 (1.5)	83.9 (9.9)
Household food insecurity	23.0 (2.9)	25.0 (4.9)	0.396	6.0 (0.6)	7.0 (0.5)	0.070	23.3 (2.3)	11.8 (5.2)
HEI (mean (SE))	52.1 (1.3)	52.8 (1.2)	0.078	58.3 (0.4)	56.2 (0.3)	<0.001	52.4 (0.8)	56.3 (1.0)
Men (n)	325	260		9121	4925
Household food insecurity	22.0 (4.3)	18.6 (3.7)	0.310	5.6 (0.5)	6.5 (0.7)	0.031
HEI (mean (SE))	51.6 (1.5)	50.9 (1.3)	0.070	56.4 (0.3)	54.6 (0.4)	<0.01
Women (n)	501	298		10,938	5588
Household food insecurity	24.2 (3.4)	29.9 (5.3)	0.077	6.5 (0.5)	7.8 (0.8)	<0.01
HEI (mean (SE))	52.6 (1.4)	54.2 (1.6)	0.184	60.2 (0.3)	57.8 (0.3)	<0.001

Results presented as percent (SE) unless otherwise noted; HEI, Healthy Eating Index

^aUnmatched, weighted analysis

^bIndigenous and non-Indigenous participants were matched, using coarsened exact matching according to age, sex, income adequacy, household education, and survey year, and weighted using CEM and survey weights

In a matched analysis using GLM, and adjusting for energy intake and food security status, there was no significant interaction effect for time period*Indigenous identifier on HEI in sex pooled analysis (Table 3, model 4). This indicates that while HEI was significantly different according to time period and between Indigenous and non-Indigenous populations, the relationship between each variable and HEI was not dependent on the other. In a sex-stratified analysis, both men and women did not exhibit a significant time period*Indigenous identifier interaction effect (Table 3, models 5 and 6). However, the unmatched GLM, adjusting for confounders, indicated a significant (p<0.05) interaction effect for time period*Indigenous identifier on HEI (Table 3, model 2). This suggests the importance of matching for obtaining a more accurate estimate given the known differences between the Indigenous and non-Indigenous populations. In a pooled analysis of both surveys and unmatched, both First Nations (−3.24 [−5.20, −1.17]; p<0.01) and Métis (−2.45 [−4.59, −0.31]; p<0.05) adults had significantly lower HEI scores compared with non-Indigenous adults, independent of age, sex, energy intake, time period, highest level of education, income adequacy, and food insecurity (data not shown).

Table 3.

Linear regression coefficient (95% CI) of Healthy Eating Index (HEI) for Canadian adults ≥18 years old

	Unmatcheda		Matchedb
	Model 1a	Model 2a	Model 3b	Model 4b	Model 5: Men onlyb	Model 6: Women onlyb
Time periodc	−1.64*** (−2.24, −1.03)	−1.64*** (−2.24, −1.03)	−1.82*** (−2.44, −1.20)	−1.97*** (−2.69, −1.24)	−1.59** (−2.57, −0.61)	−1.37* (−2.49, −0.25)
Indigenousd	−5.35*** (−7.36, −3.35)	−4.86*** (−6.90, −2.82)	−4.78*** (−7.40, −2.15)	−4.17** (−6.62, −1.72)	−2.99 (−6.72, 0.74)	−4.80** (−8.15, −1.45)
Time period*Indigenous	2.23* (0.54, 5.01)	2.35* (0.43, 5.11)	1.74 (−1.07, 4.55)	1.87 (−0.93, 4.68)	1.18 (−2.65, 5.02)	3.12 (−0.90,7.14)
Household food insecurity		−4.09*** (−5.22, −2.96)		−5.89*** (−8.53, −3.24)	−5.08 (−7.10, −3.07)	−6.59*** (−10.18, −3.00)

* p<0.05; ** p<0.01; *** p<0.001

^aUnmatched and adjusted for age, sex, energy intake, highest level of household education, and income adequacy

^bMatched and adjusted for energy intake

^cResults are for 2015, with 2004 as the reference time period

^dResults are for Indigenous adults, with non-Indigenous as the reference group

Further models stratified by subgroups (Table 4) confirm that the significant Indigenous identifier*time period interaction effect presented in Table 3 (model 2) is due to a significantly lower HEI score among non-Indigenous women in 2015 as compared with 2004, and not due to significant improvements in HEI among Indigenous women in 2015 as compared with 2004. Furthermore, while food security was negatively and independently associated with HEI among non-Indigenous men and women (Table 4, models 2, 4, and 6), food security was not significantly associated with HEI among Indigenous men or women, or combined (Table 4, models 1, 3, and 5).

Table 4.

Linear regression coefficient (95% CI) of Healthy Eating Index for subgroups of Canadian adults ≥18 years old

Modelsa	β (95% CI)
Model 1: Indigenous only
Time period, 2015	1.38 (−1.54, 4.31)
Food insecurity	−0.80 (−5.09, 3.48)
Model 2: Non-Indigenous only
Time period, 2015	−1.58 (−2.19, −0.97) ***
Food insecurity	−4.33 (−5.53, −3.14) ***
Men only
Model 3: Indigenous only
Time period, 2015	0.52 (−3.51, 4.56)
Food insecurity	2.51 (−2.80, 7.82)
Model 4: Non-Indigenous only
Time period, 2015	−1.24 (−2.16, −0.31) **
Food insecurity	−4.35 (−6.36, −2.32) ***
Women only
Model 5: Indigenous only
Time period, 2015	2.02 (−2.04, 6.08)
Food insecurity	−3.29 (−8.77, 2.19)
Model 6: Non-Indigenous only
Time period, 2015	−1.85 (−2.62, −1.10) ***
Food insecurity	−4.22 (−5.65, −2.79) ***

* p<0.05; ** p<0.01; *** p<0.001

^aAdjusted for energy intake, age, sex (where relevant), household education and household income adequacy

In examining individual component scores of HEI according to year and Indigenous identifier, few significant differences were noted in 2015 compared with 2004. In 2004, five of eleven HEI component scores were significantly lower for Indigenous adults compared with non-Indigenous adults (Table 5). However, in 2015, only one component score, whole grain products, was significantly lower for Indigenous adults compared with non-Indigenous adults. Notably, the percent energy from ‘other’ foods was not significantly different between Indigenous and non-Indigenous adults in 2015, but the component score improved from 8.91 in 2004 to 10.15 in 2015 among the Indigenous population.

Table 5.

Mean Healthy Eating Index (HEI) component scores (SE) among Canadians ≥18 years old using a matched and weighted analysis

Component (maximum score possible)	2004		2015
	Non-Indigenous	Indigenous	Non-Indigenous	Indigenous
Fruits and vegetables (10)	5.52 (0.34)*	4.55 (0.25)	5.03 (0.26)	4.76 (0.27)
Dark green or orange vegetables (5)	1.43 (0.11)*	1.06 (0.15)	1.25 (0.09)	1.04 (0.14)
Whole fruits (5)	2.17 (0.03)*	1.54 (0.15)	2.19 (0.17)	1.93 (0.16)
Grain products (5)	3.35 (0.05)	3.18 (0.11)	3.15 (0.03)	3.08 (0.14)
Whole grain products (5)	1.47 (0.03)	1.22 (0.12)	0.95 (0.13)*	0.81 (0.11)
Milk products (10)	5.25 (0.10)*	4.50 (0.30)	5.00 (0.11)	4.37 (0.32)
Meat and alternatives (10)	7.30 (0.04)	7.36 (0.24)	6.41 (0.07)	5.98 (0.26)
Unsaturated fats (10)	8.19 (0.04)	8.33 (0.18)	8.32 (0.05)	7.93 (0.22)
Saturated fats (10)	6.59 (0.06)	6.54 (0.29)	6.20 (0.07)	5.98 (0.29)
Sodium (10)	5.79 (0.17)	5.51 (0.26)	6.01 (0.14)	5.81 (0.34)
Percentage of energy from ‘other’ foods (20)	10.39 (0.20)**	8.91 (0.53)	10.33 (0.25)	10.15 (0.56)

*p<0.05; **p<0.01; *** p<0.001 according to t-test and as compared with Indigenous sample within the same year

Discussion

Our results show that diet quality, as measured using the adapted Canadian HEI, was not significantly different among Indigenous adults in 2015 as compared with 2004. While Indigenous adults continue to have significantly lower diet quality compared with non-Indigenous adults, this disparity was not significantly different in 2015 compared with 2004. It should be mentioned that unmatched, sex-stratified analysis indicated that the disparity in diet quality between Indigenous and non-Indigenous women was significantly reduced; yet, this change was attributed to lower diet quality among non-Indigenous women in 2015 compared with 2004, not any significant improvement in diet quality among Indigenous women. This may be attributable to changes in the non-Indigenous population between 2004 and 2015, such as increased immigration (Statistics Canada, 2016); however, this requires further examination. These results are in contrast to our other findings among children and youth in the same survey, which shows significantly improved HEI scores among Indigenous children and youth in 2015 compared with 2004 (Riediger et al., 2021).

A closer examination of HEI component scores among adults reported here does suggest some improvement in aspects of diet quality however; scores pertaining to percent energy from ‘other’ foods increased from 8.91 in 2004 to 10.15 in 2015 among Indigenous adults, unlike among non-Indigenous adults, for whom scores remained static. Score for percent energy from ‘other’ foods are reverse-scored, meaning higher intake (and percent energy) results in a lower score. These findings are also similar to our findings among Indigenous children and youth, where the greatest increase in component score from 2004 to 2015 was for ‘other’ foods (Riediger et al., 2021). This may indicate success in attempts to consume a more nutritious diet, in which reducing intake of less nutritious ‘other’ foods may be more amenable in the context of food insecurity. This is in contrast to fruits and vegetables, for which scores remained relatively stable, and intake of which would be more difficult to increase in the context of food insecurity given their costs.

Our pooled 2004 and 2015 analyses indicate similar disparities in HEI between both First Nations and Métis populations, and the non-Indigenous population. This result suggests that the lack of difference in HEI among the Indigenous population in 2015 compared with 2004 is likely not attributed to changes in the proportions of First Nations versus Métis populations over time. This finding also further supports the importance of examining and incorporating both First Nations- and Métis-specific perspectives in food policies.

Our results also appear to mirror patterns in diabetes prevalence reported in our introduction. The mean, or average, HEI for the off-reserve Indigenous population reported here in 2015, at 52.8, is higher than that reported for First Nations adults on-reserve according to the First Nations Food, Nutrition, and Environment Study, which ranged from a mean of 47.0 among First Nations in Manitoba to 50.9 among First Nations in Ontario (Willows et al., 2019). Generally, this suggests that Indigenous people living off-reserve have better diet quality as compared with First Nations living on-reserve; however, Willow and colleagues’ (Willows et al., 2019) results also suggest considerable variation in on-reserve First Nation diet quality between provinces, which complicates direct comparison. Again, this mirrors patterns in diabetes prevalence. Unfortunately, the CCHS surveys do not include a sufficient sample size to investigate geographical or provincial differences in HEI among the off-reserve Indigenous population reported here.

Indigenous adults reported significantly lower diet quality, independent from economic circumstances and food security, suggesting other unmeasured factors may be contributing to this disparity. One potential factor may be trauma or intergenerational trauma, which has been reviewed by Aguiar and Halseth (2015) among Indigenous people. Though the authors do not discuss a link between trauma and eating among Indigenous people, there is considerable research among general populations that supports negative implications of trauma for eating (Echeverri-Alvarado et al., 2020; Mason et al., 2017; Springer et al., 2003). Our results may also suggest that, while economic food security is necessary to improve diet quality among Indigenous people, it is not sufficient. This aligns with the work of several Indigenous scholars and communities calling for Indigenous food sovereignty, including consideration for cultural aspects of food, incorporating Indigenous knowledge, self-determination, and connections to land in food security policy (Cidro et al., 2015; Morrison, 2011; Richmond et al., 2021; Working Group on Indigenous Food Sovereignty, 2011). Importantly, as discussed by Aguiar and Halseth (2015), re-establishing individual and collective identity through culture will be critical for healing from trauma.

Food security was not significantly associated with diet quality among Indigenous men or women, while this was not the case for non-Indigenous adults. Interestingly, similar results have been reported for the Black and Hispanic populations in the United States, while food insecurity was significantly and inversely associated with diet quality among the white American population (Leung & Tester, 2019). The reasons for racial differences in this relationship are unclear but may point to HEI as a racially biased measure of diet quality. Food insecurity measured as economic access may be a less relevant construct among Indigenous populations with respect to diet quality, as discussed in the preceding paragraph, or more relevant to only certain food groups. The coefficient for Indigenous women (β: −3.29), though not statistically significant, does suggest that food insecurity may be a more relevant factor for HEI for them than for Indigenous men, whose coefficient was above 0 (β: 2.51). The lack of statistical significance for this relationship among women may more likely be due to the small sample size and limited power to detect a relationship.

There are a number of limitations of this study, particularly related to data quality specific to Indigenous peoples (see Smylie & Anderson, 2006 for a comprehensive review of data issues for Indigenous health). First, we are limited by a pan-Indigenous approach when examining changes between time periods; that is First Nations, Métis, and Inuit people are not uniquely identified because of limited sample size. Despite the inability to disaggregate the data by Indigenous groups, the results of the study remain valuable given the similarities between First Nations and Métis living in urban areas, particularly with respect to policy implications. This survey is not sufficiently powered to describe diet quality among Inuit living in the 10 provinces and there were few Inuit participants in these surveys. There is limited research regarding the health or nutritional status of Inuit people living outside Inuit Nunangat. Given the changes in proportions between First Nations and Métis populations off-reserve between 2004 and 2015, the lack of change in HEI may also reflect differences in the self-identified Indigenous population. We also did not take into account how the non-Indigenous population has changed over time, aside from socio-economic variables, which may be further biasing comparisons between time periods.

Second, the utilization of the HEI based on previous dietary guidelines may be considered a limitation. A new Canadian Food Guide was recently released (Government of Canada, 2020) and updated research measures have not been developed to reflect the updated guidelines. However, the index does reflect the guidelines at the time of data collection in 2015. Third, this analysis presumes a measure such as HEI represents an Indigenous perspective of diet quality, however, this is not the case. The CCHS Nutrition surveys do not include questions developed by and for Indigenous people, such as those relating to traditional food intake, relationships with land and food, and eating communally (i.e., feasting), as examples. The CCHS surveys do not include additional questions pertaining to sex or gender beyond the dichotomous variable included here. Furthermore, there were changes in survey methods between the two surveys and response rate was less in 2015 as compared with 2004, potentially increasing non-response bias. Last, dietary data are limited by the use of a single dietary recall, which is subject to recall bias and under-reporting, among other limitations of dietary recall data (Health Canada, 2017).

Conclusion

Diet quality remains a health concern for off-reserve Indigenous adults. Between 2004 and 2015, some progress was made in closing gaps or disparities in diet quality between Indigenous and non-Indigenous populations, though this is attributed to lower diet quality among non-Indigenous women in 2015. However, reductions in the percent energy from ‘other’ foods among Indigenous adults is a positive sign. Overall, current policies and systems have done little to improve opportunities for Indigenous adults to access and utilize food to meet dietary guidelines. Furthermore, improvements in diet quality for Indigenous people will not be achieved through economic food security alone. Considerable efforts are required to enhance Indigenous food sovereignty, which, through self-determination, will both address food security and enhance relationships with food.

Contributions to knowledge

What does this study add to existing knowledge?

• Disparities in diet quality between off-reserve Indigenous and non-Indigenous adults persisted from 2004 to 2015.

• This disparity in diet quality was independent of food (in)security.

• The percent of energy from ‘other’ foods (e.g., sugar-sweetened beverages, potato chips) among Indigenous adults was lower in 2015 compared with 2004 (i.e., an improvement in component of diet quality).

What are the key implications for public health interventions, practice or policy?

• Current policies were not sufficient in addressing dietary disparities among off-reserve Indigenous adults.

• To close gaps in nutrition-related health inequities among Indigenous populations, such as type 2 diabetes, as called for in the Truth and Reconciliation Commission Calls to Action, addressing disparities in diet quality will be critical.

• Addressing food insecurity, which continues to disproportionately impact off-reserve Indigenous populations, is necessary, but likely not sufficient to eliminate disparities in diet quality.

Author contributions

NR was involved in formulating the research question, designing the study, carrying it out, and writing the article. JL was involved in formulating the research question, designing the study, contributing to interpretation, and reviewing the manuscript for intellectual content. AM was involved in designing the study, carrying it out, analyzing the data, contributing to interpretation, and reviewing the manuscript for intellectual content. LC was involved in designing the study, carrying it out, analyzing the data, contributing to interpretation, and reviewing the manuscript for intellectual content.

Funding

This work was funded by the Canadian Institutes of Health Research (CIHR), Canadian Community Health Survey Nutrition Analysis (Grant #151546). NDR is the recipient of a CIHR Early Career Investigator Award (2018-2022; grant #155435). Research at the Manitoba Research Data Centre is supported by funds to the Canadian Research Data Centre Network (CRDCN) from the Social Sciences and Humanities Research Council (SSHRC), CIHR, the Canadian Foundation for Innovation (CFI), and Statistics Canada. Although the research and analysis are based on data from Statistics Canada, the opinions expressed do not represent the views of Statistics Canada.

Declarations

Conflict of interest

The authors declare no competing interests.

Ethics approval

This article does not contain any studies with human participants performed by any of the authors. All procedures performed in studies involving human participants were in accordance with the ethical standards of Statistics Canada (Microdata research contract 17-SSH-WIN-5184) and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Consent to participate

Informed consent was obtained from all individual participants included in the study.