Socioeconomic Inequalities in Intakes of Ultraprocessed and Minimally Processed Foods in Nationally Representative Samples of Adults in Canada: An Analysis of Trends between 2004 and 2015

Abstract

Background

Individuals with a lower socioeconomic position (SEP) often have higher intakes of ultraprocessed food (UPF) and lower intakes of minimally processed food (MPF); however, studies have not examined trends in absolute and relative gaps and gradients in UPF and MPF intake using multiple indicators of SEP.

Objectives

We examined within-year absolute and relative gaps and gradients in UPF and MPF intake and trends between 2004 and 2015 according to 6 indicators of SEP among nationally representative samples of adults in Canada.

Methods

Adults (≥18 y) in the Canadian Community Health Survey-Nutrition 2004 (n = 20,880) or 2015 (n = 13,970) reported SEP (individual and household education, household income adequacy, household food insecurity, neighborhood material and social deprivation) and completed a 24-h dietary recall. Multivariable linear regression assessed within-year absolute and relative gaps and gradients in the proportion of energy from UPF and MPF and trends between 2004 and 2015.

Results

The largest and most consistent within-year inequities in UPF and MPF intake were for individual and household educational attainment. Overall and among males, higher SEP groups had more favorable intakes over time based on trends in absolute and relative gaps and gradients in UPF and MPF intake by household food insecurity [for example, the absolute gap in UPF intake declined from −1.2% (95% confidence interval: −5.3%, 2.9%) to −7.9% of energy (95% confidence interval: −11.2%, −4.5%) in the overall population]. Overall and among males, lower SEP groups had more favorable intakes over time based on trends in absolute and relative gaps in UPF and MPF intake by neighborhood material deprivation.

Conclusions

Socioeconomic inequalities in UPF and MPF intake were most pronounced for individual and household education. Between 2004 and 2015, several inequalities in UPF and MPF intake emerged according to household food insecurity (favoring higher SEP groups) and neighborhood material deprivation (favoring lower SEP groups).

Introduction

In many high-income countries such as Canada, ultraprocessed foods and drinks (UPF) account for the largest proportion of daily energy intake [[1], [2], [3]]. UPF are industrially manufactured formulations created through extensive processing. These foods contain little or no whole foods and are made with additives to enhance taste, palatability, and shelf life (for example, prepackaged snacks, soft drinks). Unprocessed or minimally processed foods (MPF) typically represent the second largest proportion of daily energy intake and are whole foods that have undergone minimal or no processing (for example, milk and fresh or frozen vegetables) [4]. Higher intake of UPF has been linked with poorer diet quality [3,5] and a higher risk of nutrition-related chronic diseases [[6], [7], [8], [9], [10], [11]] and mortality [12]. Conversely, higher MPF intake has been associated with higher diet quality [5,13] and lower risk of obesity [14], metabolic syndrome [15], and mortality [16,17].

Evidence indicates that individuals with a lower position in the social hierarchy, termed socioeconomic position (SEP), tend to consume more UPF and fewer MPF than individuals with a higher SEP [1,3,18,19]. However, these patterns may not apply uniformly across different indicators of SEP and also appear to differ between countries [3,[20], [21], [22]]. These varied associations underscore the importance of examining how a variety of indicators of SEP at multiple levels, including individual and household educational attainment, household income adequacy, household food insecurity status, and neighborhood-level deprivation, are associated with UPF and MPF intake in multiple countries.

Individual-level educational attainment is an indicator of SEP that captures access to material resources through its influence on occupation and income, and also denotes prestige and knowledge-related and cultural assets [[23], [24], [25]]. Individuals with higher educational attainment may have greater access to nutrition and health-related information [23], and thus consume fewer UPF and more MPF than those with lower levels of education. Household-level educational attainment may similarly influence UPF and MPF intake, as the benefits of one household member’s educational attainment may be experienced by other household members. Household income is another commonly used indicator of SEP that most directly denotes access to material resources such as food and shelter [23]. Individuals living in households with lower incomes may consume more UPF as they can be more affordable than MPF [26]; however, associations between UPF intake and household income are mixed and in some cases individuals with higher incomes actually consume more UPF than their lower income counterparts [20]. Household food insecurity status can also be used as an indicator of SEP as it reflects access to material resources. However, it is more specific to food as it indicates that resources have become so constrained that food intake must be compromised to make ends meet [19,27]. Individuals living in food insecure households tend to consume more UPF and less MPF, as UPF are often more affordable, readily available, and require fewer resources to prepare, aligning with the extreme lack of resources of these households [19].

In Canada, more materially deprived neighborhoods do not typically lack access to stores selling nutritious, whole foods, but there is some evidence that they may have higher access to outlets offering an abundance of less nutritious, shelf-stable options [28]. Consequently, residents of such neighborhoods may purchase and consume more UPF than MPF [29,30]. Neighborhoods with high levels of social deprivation may be disorderly and lacking in social cohesion [29,31]. As a result, residents of these neighborhoods may experience higher levels of psychosocial stress and consume foods that provide emotional comfort as a coping mechanism—many of these being nutrient-poor and energy-dense, such as UPF [29,31,32].

Previous studies have shown that several socioeconomic inequalities in UPF and MPF intake have remained stable or increased over the last 2 decades in the United States and the United Kingdom on an absolute basis [1,18,33]. However, to provide a more comprehensive perspective of dietary inequalities, studies should examine trends in absolute and relative dietary gaps and gradients for multiple indicators of SEP [[34], [35], [36], [37]]. Dietary gaps reflect differences in UPF and MPF intake between groups at the extreme ends of the socioeconomic spectrum, whereas gradients also take into account the size of each group and the slope of the full distribution [34,35]. Inequalities in UPF and MPF intake can also be quantified on absolute and relative scales. Absolute measures (that is, differences between groups) reflect a perspective that emphasizes the importance of optimizing UPF and MPF intake at a population level, whereas relative measures (that is, the ratio of one group’s intake to another) reflect an egalitarian perspective that emphasizes the importance of equalizing UPF and MPF intake across the socioeconomic spectrum [36,38]. It is also important to examine whether dietary inequalities differ by sex and gender as females tend to consume fewer UPF and more MPF than males [1,2]. These differences could relate to factors such as biological differences in energy and nutrient requirements (sex) or differences in socially constructed norms and values related to food (gender). Therefore, the primary objective of this study was to examine within-year absolute and relative gaps and gradients in UPF and MPF intake, and trends in these inequalities between 2004 and 2015 according to 6 indicators of SEP among nationally representative samples of adults in Canada, both overall and by sex. As a secondary objective, we also replicated these analyses to examine inequalities in intake of processed culinary ingredients and processed foods.

Methods

Study design and participants

This study utilized data from the most recent cycles of the Canadian Community Health Survey (CCHS)-Nutrition, conducted in 2004 and 2015 by Statistics Canada. These cross-sectional, nationally representative surveys collected dietary intake data and sociodemographic and health-related information from individuals living in Canada [[39], [40], [41]]. The surveys were designed to be comparable to enable examination of trends over time [39]. The CCHS-Nutrition surveys cover 98% of Canada’s population in all 10 provinces, excluding individuals living in Canada’s 3 Northern territories, on Indigenous reserves, in institutions, and those in the Canadian Forces. Both surveys utilized multistage, stratified, clustered, probabilistic sampling procedures with response rates of 77% in 2004 and 62% in 2015 [39]. This study included adults aged ≥18 y. Individuals whose food intake was null or otherwise deemed unreliable by Statistics Canada were excluded from the analyses. Pregnant and breastfeeding females (n = 260 in 2004 and n = 305 in 2015) were also excluded. The final sample sizes were n = 20,880 in 2004 and n = 13,970 in 2015. This study used Statistics Canada data and was therefore deemed exempt from ethical review by the Conjoint Health Research Ethics Board at the University of Calgary.

Data collection

Trained interviewers conducted in-home interviews during all days of the week and seasons of the year. Respondents self-reported their age, sex (gender identity was not queried), Indigenous status, and race/ethnicity, as well as the exposures and outcomes described below.

Indicators of SEP

Educational attainment

Respondents reported their highest educational attainment and the highest educational attainment within the household. Educational attainment was categorized as less than high school diploma, high school diploma, some postsecondary below bachelor’s degree, higher education equivalent to bachelor’s degree, and university degree above bachelor’s degree.

Household income adequacy

Household income adequacy was determined by adjusting the total annual household income after transfers and before taxes in the past 12 mo for inflation-adjusted low-income cutoffs. Low-income cutoffs are income thresholds derived by Statistics Canada to denote income levels where some households spend at least 20% more on essentials like food and shelter than average households [42]. They include 35 thresholds (7 family sizes, considering economies of scale, and 5 geographic areas to adjust for cost of living in different locations) [42]. Missing income values were imputed for 11.4% of records in the CCHS-Nutrition 2004 following Statistics Canada procedures, and Statistics Canada imputed total household income for 24.1% of adults in the CCHS-Nutrition 2015 [43]. Household income adequacy was categorized into quintiles.

Household food insecurity status

Household food insecurity status was assessed using Health Canada’s validated 18-item Household Food Security Survey Module [44]. On the basis of the number of affirmative responses, household food insecurity status was classified as food secure (no affirmative responses), marginally food insecure (1 affirmative response), moderately food insecure (adult scale: 2–5 affirmative responses; child scale: 2–4 affirmative responses), and severely food insecure (adult scale: ≥6 affirmative responses; child scale ≥5 affirmative responses) [27].

Indigenous status and race/ethnicity

Respondents self-reported their Indigenous status and race/ethnicity. We use the term race/ethnicity because the CCHS asked participants to report whether they belonged to one or more racial or cultural groups without distinguishing between the 2. Indigenous status and race/ethnicity were categorized as Indigenous, White, East/Southeast Asian, South Asian, Black, Latin American, Middle Eastern and Other (that is, non-Indigenous participants who self-identified as “Other” or who selected >1 response option).

Neighborhood deprivation

The Pampalon index of deprivation was used as a measure of neighborhood deprivation at the Dissemination Area level (that is, the smallest area for which Census data are available, typically encompassing 400–700 residents) [45]. Using participants’ residential postal codes, the Pampalon index was assigned by employing the Postal Code Conversion File Plus to map these codes to their corresponding Dissemination Area and associated index scores. Indicators of material and social deprivation were included as 2 separate variables. Material deprivation reflects the proportion of individuals residing in a Dissemination Area with low education or income and the employment-to-population ratio [46]. Social deprivation captures the proportion of individuals living in a Dissemination Area who are living alone, separated, widowed, or divorced [46]. Material and social deprivation indices were derived from the 2006 and 2016 Canada Census data [[69], [70]] for the analyses of the CCHS-Nutrition 2004 and 2015, respectively, given they were the nearest years with available Census data. National-level quintiles of the Pampalon index were used, with the first and fifth quintiles indicating the most and least deprived Dissemination Areas, respectively.

Outcomes

Proportion of energy from ultraprocessed foods, unprocessed and minimally processed foods, processed culinary ingredients, and processed foods

Participants completed an in-person 24-h dietary recall, reporting all foods and beverages consumed from midnight to midnight the previous day. The Automated Multiple Pass Method was employed to enhance the precision of recalls [47]. Booklets with food models were provided for portion size estimation. A random subset of participants completed a second dietary recall 3–10 d later; however, these data were not included in our analyses because group means from a single dietary recall offer unbiased population-level estimates [48].

All reported foods and beverages were categorized according to the Nova classification [4]. Although there are several other schemes that classify foods based on their level of processing, Nova is the most commonly used and has been associated with morbidity and mortality in multiple systematic reviews [10,11,49]. Nova categorizes foods based on their degree of industrial processing, including MPF, processed culinary ingredients, processed foods, and UPF [4]. Using food item descriptions from each survey’s respective Nutrition Survey System database [40,50], foods reported in the CCHS-Nutrition 2015 were categorized into Nova food groups and corresponding items from the CCHS-Nutrition 2004 were placed in the same groups [2,4]. For meals and mixed dishes, individual ingredients were classified separately unless consumed at a fast food restaurant, in which case they were automatically considered UPF. Two researchers (JCM and JP) independently classified all foods and beverages and resolved any discrepancies through collaborative discussion [2]. Lastly, the energy content of each food was extracted from each Nutrition Survey System database, and the percentage of total daily energy intake (excluding alcohol) from each Nova food group was calculated.

Data analysis

Indices of inequalities

Absolute and relative gaps and gradients were calculated to examine inequalities in the proportion of energy from each Nova food group [34]. Absolute gaps were quantified by subtracting the mean proportion of energy consumed from UPF and MPF of the lowest from that of the highest SEP group. The Slope Index of Inequality (SII) was calculated as a measure of absolute gradients in the proportion of energy from each Nova food group [34,35]. The SII reflects the absolute difference in intake of each Nova food group between the highest and lowest SEP groups while considering the size of each group and the slope of the entire distribution. Relative gaps were quantified as the ratio of the proportion of energy from each Nova food category in the highest to the lowest SEP group. The Relative Index of Inequality was calculated similarly to the SII but using relative values to quantify relative gradients [34,35].

Interpretations of absolute and relative gaps and gradients for UPF and MPF intake are described in Table 1. Note that a higher proportion of energy from MPF is desirable, whereas a higher intake of UPF is not, and thus the interpretation of gaps and gradients differ for these 2 outcomes.

TABLE 1.

Description and interpretation of absolute and relative gaps and gradients in the proportion of energy from ultraprocessed foods and unprocessed or minimally processed foods

	Calculation	Interpretation of within-year values	Interpretation of trends	Example
Ultraprocessed foods
Absolute gaps and gradients	Absolute gaps: the difference in the proportion of energy from UPF between the highest and lowest SEP groups (highest SEP UPF% − lowest SEP UPF%) Absolute gradients: first, individuals are ranked based on their SEP from the most disadvantaged (rank 0) to the most advantaged (rank 1). Weights are then assigned to each SEP group, reflecting their representation in the overall population. For each group, the midpoint of their range in the cumulative population distribution is calculated. Subsequently, the proportion of energy from UPF is regressed against the midpoint value of each SEP group. The values of the proportion of energy from UPF for the 2 extremes of the SEP spectrum are then predicted. The absolute gradient is calculated as the difference between these predicted values.	0 = complete equality Positive values: indicate that higher SEP groups derive a higher proportion of their energy from UPF relative to lower SEP groups Negative values: indicate that higher SEP groups derive a lower proportion of their energy from UPF relative to lower SEP groups	Socioeconomic inequalities in UPF intake are increasing if absolute gaps and/or gradients become more negative over time. This indicates that higher SEP groups had more favorable intakes over time.	Absolute gaps in UPF intake according to household food insecurity status decreased from −1.2% in 2004 to −7.9% in 2015
Relative gaps and gradients	Relative gaps: the ratio of the proportion of energy from UPF between the highest and lowest SEP groups (highest SEP UPF%/lowest SEP UPF%) Relative gradients: first, individuals are ranked based on their SEP from the most disadvantaged (rank 0) to the most advantaged (rank 1). Weights are then assigned to each SEP group, reflecting their representation in the overall population. For each group, the midpoint of their range in the cumulative population distribution is calculated. Subsequently, the proportion of energy from UPF is regressed against the midpoint value of each SEP group. The values of the proportion of energy from UPF for the 2 extremes of the SEP spectrum are then predicted. The relative gradient is then determined as the ratio of these predicted values	1.0 = complete equality >1.0 = indicates that higher SEP groups derive a higher proportion of their energy from UPF relative to lower SEP groups <1.0 = indicates that higher SEP groups derive a lower proportion of their energy from UPF relative to lower SEP groups	Socioeconomic inequalities in UPF intake are increasing if relative gaps and/or gradients decrease over time. This indicates that higher SEP groups had more favorable intakes over time	Relative gaps in UPF intake according to household food insecurity status decreased from 1.0 in 2004 to 0.9 in 2015
Unprocessed or minimally processed foods (MPF)
Absolute gaps and gradients	Absolute gaps: the difference in the proportion of energy from MPF between the highest and lowest SEP groups (highest SEP MPF% − lowest SEP MPF%) Absolute gradients: first, individuals are ranked based on their SEP from the most disadvantaged (rank 0) to the most advantaged (rank 1). Weights are then assigned to each SEP group, reflecting their representation in the overall population. For each group, the midpoint of their range in the cumulative population distribution is calculated. Subsequently, the proportion of energy from MPF is regressed against the midpoint value of each SEP group. The values of the proportion of energy from MPF for the 2 extremes of the SEP spectrum are then predicted. The absolute gradient is calculated as the difference between these predicted values.	0 = complete equality Positive values: indicate that higher SEP groups derive a higher proportion of their energy from MPF relative to lower SEP groups Negative values: indicate that higher SEP groups derive a lower proportion of their energy from MPF relative to lower SEP groups	Socioeconomic inequalities in MPF intake are increasing if absolute gaps and/or gradients become more positive over time. This indicates that higher SEP groups had more favorable intakes over time.	Absolute gaps in MPF intake according to household food insecurity status increased from −0.5% in 2004 to 4.2% in 2015
Relative gaps and gradients	Relative gaps: the ratio of the proportion of energy from MPF between the highest and lowest SEP groups (highest SEP MPF %/lowest SEP MPF %) Relative gradients: first, individuals are ranked based on their SEP from the most disadvantaged (rank 0) to the most advantaged (rank 1). Weights are then assigned to each SEP group, reflecting their representation in the overall population. For each group, the midpoint of their range in the cumulative population distribution is calculated. Subsequently, the proportion of energy from MPF is regressed against the midpoint value of each SEP group. The values of the proportion of energy from MPF for the 2 extremes of the SEP spectrum are then predicted. The relative gradient is then determined as the ratio of these predicted values	1.0 = complete equality >1.0 = indicates that higher SEP groups derive a higher proportion of their energy from MPF relative to lower SEP groups <1.0 = indicates that higher SEP groups derive a lower proportion of their energy from MPF relative to lower SEP groups	Socioeconomic inequalities in MPF intake are increasing if relative gaps and/or gradients increase over time. This indicates that higher SEP groups had more favorable intakes over time	Relative gradients in MPF intake according to household food insecurity status increased from 1.0 in 2004 to 1.4 in 2015 among males

Abbreviations: MPF, unprocessed or minimally processed food; SEP, socioeconomic position; UPF, ultraprocessed food.

Statistical analyses

Descriptive statistics were calculated to describe the samples in 2004 and 2015. Analyses were conducted for the total population and among males and females. Multivariable linear regression models were used to estimate the mean proportion of energy from each Nova food group for each of the 6 indicators of SEP in 2004 and 2015 separately. Models were adjusted for age, sex, and dietary recall day (weekday compared with weekend, including Fridays). Standard errors were calculated using the delta method [51].

Analyses for the CCHS-Nutrition 2004 and 2015 were conducted separately to estimate within-year intake from each Nova food group, along with absolute and relative gaps and gradients. Using each year’s estimates, z-scores were calculated to examine trends between 2004 and 2015 in intake of all Nova food groups, along with absolute and relative gaps and gradients [52].

After imputing household income, missing data were minimal (<1.0% for all variables, except neighborhood deprivation which had 5.2% missing), and therefore list-wise deletion was used. To generate nationally representative estimates, person-specific and bootstrap weights supplied by Statistics Canada were used to accommodate the complex sampling design, to adjust for nonresponse, and to more precisely estimate variance components. Analyses were performed using RStudio [53], with P < 0.05 deemed statistically significant. No adjustments were made for multiple testing to reduce risk of type 2 errors and because all analyses were prespecified [54].

Sensitivity analyses

Two sensitivity analyses were conducted. The first adjusted for Indigenous status and race/ethnicity to examine if changes in the population composition between 2004 and 2015 affected findings. The second adjusted for the ratio of total energy intake to total energy expenditure as an indicator of dietary misreporting. Total energy expenditure was estimated using equations developed by the Institute of Medicine [55] based on measured height and weight and assuming a low physical activity level.

Results

Descriptive results

Although the proportion of energy from UPF decreased slightly from 43.3% ± 20.0% to 41.9% ± 20.3% of energy (P < 0.001) in the total population, the proportion of energy from MPF remained stable overall, accounting for 39.8% ± 19.0% of energy in 2004 and 39.6% ± 19.0% in 2015 (Table 2). There was evidence of a socioeconomic patterning in UPF and MPF intake in 2004 and/or 2015 for several indicators of SEP. The largest and most consistent within-year inequalities were found according to individual and household educational attainment (FIGURE 1, FIGURE 2), with absolute gaps and gradients ranging from 8.3% to 12.9% of energy from UPF and from 3.7% to 9.2% for MPF. Between 2004 and 2015, UPF intake declined and MPF intake increased among most SEP groups in the overall population (Supplemental Tables 1 and 4). Results for processed culinary ingredients and processed foods are provided in Supplemental Tables 2 and 3.

TABLE 2.

Weighted characteristics of adults who participated in the Canadian Community Health Survey-Nutrition 2004 (n = 20,880) or 2015 (n = 13,970)¹

Variables		2004	2015
Age (mean ± SD)		46.0 ± 17.3	48.9 ± 17.5
Nova food group2 (% of energy ± SD)	Unprocessed or minimally processed foods	39.8 ± 19.0	39.6 ± 19.0
	Processed culinary ingredients	7.3 ± 7.4	7.2 ± 7.3
	Processed foods	5.8 ± 8.3	7.3 ± 10.5
	Ultraprocessed foods	43.3 ± 20.0	41.9 ± 20.3
Sex, N (%)	Female	11,846,000 (50.0)	13,778,000 (50.0)
	Male	11,836,000 (50.0)	13,788,000 (50.0)
Household food insecurity status, N (%)	Food insecure, severe	737,000 (3.1)	895,000 (3.2)
	Food insecure, moderate	1,241,000 (5.2)	1,331,000 (4.8)
	Food insecure, marginal	917,000 (3.9)	943,000 (3.4)
	Food secure	20,788,000 (87.8)	24,397,000 (88.5)
Individual educational attainment, N (%)	Less than high school	4,645,000 (19.8)	3,254,000 (11.8)
	High school diploma	4,256,000 (18.1)	7,337,000 (26.7)
	Some postsecondary	10,008,000 (42.6)	9,259,000 (33.8)
	Bachelor’s degree	3,139,000 (13.3)	5,071,000 (18.5)
	Higher than bachelor’s degree	1,402,000 (5.9)	2,468,000 (9.0)
Household educational attainment, N (%)	Less than high school	2,431,264 (10.3)	1,920,063 (7.0)
	High school diploma	2,602,686 (11.0)	4,701,275 (17.1)
	Some postsecondary	11,126,245 (47.0)	10,274,309 (37.3)
	Bachelor’s degree	4,462,374 (18.8)	6,863,020 (24.9)
	Higher than bachelor’s degree	2,551,777 (10.8)	3,748,045 (13.6)
Household income adequacy3,4, N (%)	Quintile 1 (lowest income)	4,009,000 (16.9)	5,095,000 (18.5)
	Quintile 2	4,231,000 (17.9)	5,334,000 (19.4)
	Quintile 3	4,889,000 (20.6)	5,031,000 (18.3)
	Quintile 4	5,276,000 (22.3)	6,018,000 (21.8)
	Quintile 5 (highest income)	5,008,000 (21.1)	6,085,000 (22.1)
Neighborhood material deprivation, N (%)	Quintile 1 (most deprived)	4,616,000 (19.5)	4,888,000 (17.7)
	Quintile 2	4,698,000 (19.8)	5,708,000 (20.7)
	Quintile 3	4,302,000 (18.2)	5,513,000 (20)
	Quintile 4	4,772,000 (20.2)	5,213,000 (18.9)
	Quintile 5 (least deprived)	4,626,000 (19.5)	4,594,000 (16.7)
Neighborhood social deprivation, N (%)	Quintile 1 (most deprived)	4,989,000 (21.1)	5,377,000 (19.5)
	Quintile 2	4,806,000 (20.3)	5,669,000 (20.6)
	Quintile 3	4,704,000 (19.9)	5,118,000 (18.6)
	Quintile 4	4,546,000 (19.2)	4,611,000 (16.7)
	Quintile 5 (least deprived)	3,969,000 (16.8)	5,142,000 (18.7)
Day of dietary recall, N (%)	Weekday	13,811,000 (58.3)	15,976,000 (58.0)
	Weekend (includes Fridays)	9,870,000 (41.7)	11,590,000 (42.0)

¹Data are weighted to be nationally representative of adults aged ≥18 y. Weighted N for the total population in 2004 and 2015 are 23,682,000 and 27,566,000, respectively.

²Alcohol was excluded from total daily energy intake.

³Household income adequacy was calculated by adjusting total household income for the low-income cutoffs that correspond to the household and community size of each respondent.

⁴In 2004, the income adequacy quintiles were 0%–96.5%, 96.6%–153%, 154%–224%, 225%–322%, and >322% of the low-income cutoff. In 2015, the quintiles were 0%–106%, 107%–169%, 170%–249%, 250%–376%, and >376% of the low-income cutoff.

FIGURE 1.

The proportion of energy from ultraprocessed foods by socioeconomic position among adults who participated in the Canadian Community Health Survey-Nutrition 2004 (n = 20,880) or 2015 (n = 13,970)¹ Analyses were conducted using multivariable linear regression adjusted for age, dietary recall day, and sex. All P values for within-year inequalities were <0.001. For time trends, all P values were <0.05 overall, except for the “severe food insecurity” category. Figures are rounded in accordance with Statistics Canada’s confidentiality policies. Household income adequacy was calculated by adjusting total household income for the low-income cutoffs that correspond to the household and community size of each respondent. In 2004, the income adequacy quintiles were 0%–96.5%, 96.6%–153%, 154%–224%, 225%–322%, and >322% of the low-income cutoff. In 2015, the quintiles were 0%–106%, 107%–169%, 170%–249%, 250%–376%, and >376% of the low-income cutoff. ¹Data are weighted to be nationally representative of adults aged ≥18 y. Weighted N for the total population in 2004 and 2015 are 23,682,000 and 27,566,000, respectively. Weighted N for males in 2004 and 2015 are 11,836,000 and 13,788,000, respectively. Weighted N for females in 2004 and 2015 are 11,846,000 and 13,778,000, respectively.

FIGURE 2.

The proportion of energy from unprocessed or minimally processed foods by socioeconomic position among adults who participated in the Canadian Community Health Survey-Nutrition 2004 (n = 20,880) or 2015 (n = 13,970)¹ Analyses were conducted using multivariable linear regression adjusted for age, dietary recall day, and sex. All P values for within-year inequalities were <0.001. For time trends, all P values were <0.05 overall, except for the “severe food insecurity” category. Figures are rounded in accordance with Statistics Canada’s confidentiality policies. Household income adequacy was calculated by adjusting total household income for the low-income cutoffs that correspond to the household and community size of each respondent. In 2004, the income adequacy quintiles were 0%–96.5%, 96.6%–153%, 154%–224%, 225%–322%, and >322% of the low-income cutoff. In 2015, the quintiles were 0%–106%, 107%–169%, 170%–249%, 250%–376%, and >376% of the low-income cutoff. ¹Data are weighted to be nationally representative of adults aged ≥18 y. Weighted N for the total population in 2004 and 2015 are 23,682,000 and 27,566,000, respectively. Weighted N for males in 2004 and 2015 are 11,836,000 and 13,788,000, respectively. Weighted N for females in 2004 and 2015 are 11,846,000 and 13,778,000, respectively.

For parsimony, in the sections below we summarize within-year gaps and gradients that were significant in both 2004 and 2015, and those for which there were no inequalities in either year (TABLE 3, TABLE 4). For all other SEP indicators, within-year gaps and gradients were significant in either 2004 or 2015, but not both. We also describe trends in gaps and gradients that were significant over time. A visual summary of trends over time can also be found in Table 5.

TABLE 3.

Trends in absolute and relative gaps and gradients in the proportion of energy from ultraprocessed foods by socioeconomic position among adults who participated in the Canadian Community Health Survey-Nutrition 2004 (n = 20,880) or 2015 (n = 13,970)¹

Measure of inequality	Socioeconomic position	Total population			Males			Females
		2004 Mean (95% CI)	2015 Mean (95% CI)	P-trend	2004 Mean (95% CI)	2015 Mean (95% CI)	P-trend	2004 Mean (95% CI)	2015 Mean (95% CI)	P-trend
Absolute gaps	Household food insecurity status	−1.2 (−5.3, 2.9)	−7.9 (−11.2, −4.5)∗∗	0.015	2.4 (−4.8, 9.5)	−6.5 (−12.8, −0.1)∗	0.069	−4.9 (−8.8, −1.0)∗	−8.8 (−12.7, −5.0)∗∗	0.163
	Individual educational attainment	−8.8 (−11.2, −6.4)∗∗	−8.9 (−12.1, −5.7)∗∗	0.961	−8.3 (−11.1, −5.6)∗∗	−8.8 (−13.5, −4.1)∗∗	0.863	−9.3 (−13.4, −5.3)∗∗	−8.9 (−13.4, −4.3)∗∗	0.878
	Household educational attainment	−10.3 (−12.4, −8.2)∗∗	−10.3 (−13.3, −7.4)∗∗	0.980	−8.9 (−12.0, −5.7)∗∗	−9.1 (−13.8, −4.4)∗∗	0.941	−11.6 (−14.5, −8.8)∗∗	−11.5 (−15.5, −7.5)∗∗	0.963
	Household income adequacy2,3	−1.0 (−2.8, 0.8)	0.9 (−1.4, 3.2)	0.215	0.4 (−2.3, 3.1)	1.3 (−2.1, 4.7)	0.682	−2.2 (−4.5, 0.2)	0.8 (−2.3, 3.9)	0.141
	Neighborhood material deprivation	−2.5 (−4.2, −0.9)∗∗	0.4 (−2.1, 2.9)	0.052	−2.2 (−4.6, 0.1)	4.5 (1.1, 8.0)∗∗	0.001	−2.8 (−5.0, −0.6)∗	−3.9 (−7.3, −0.6)∗	0.580
	Neighborhood social deprivation	−2.2 (−4.1, −0.4)∗	−3.5 (−5.8, −1.2)∗∗	0.393	−2.6 (−5.3, 0.0)	−4.1 (−7.7, −0.6)∗	0.518	−1.6 (−4.0, 0.7)	−2.8 (−5.8, 0.3)	0.566
Relative gaps	Household food insecurity status	1.0 (0.9, 1.1)	0.8 (0.8, 0.9)∗∗	0.005	1.0 (0.9, 1.2)	0.9 (0.8, 1.0)∗	0.060	0.9 (0.8, 1.0)∗∗	0.8 (0.7, 0.9)∗∗	0.051
	Individual educational attainment	0.8 (0.8, 0.9)∗∗	0.8 (0.8, 0.9)∗∗	0.471	0.9 (0.8, 0.9)∗∗	0.8 (0.7, 0.9)∗∗	0.517	0.8 (0.8, 0.9)∗∗	0.8 (0.7, 0.9)∗∗	0.775
	Household educational attainment	0.8 (0.8, 0.9)∗∗	0.8 (0.7, 0.8)∗∗	0.347	0.8 (0.8, 0.9)∗∗	0.8 (0.7, 0.9)∗∗	0.504	0.8 (0.7, 0.8)∗∗	0.8 (0.7, 0.8)∗∗	0.594
	Household income adequacy	1.0 (0.9, 1.0)	1.0 (1.0, 1.1)	0.235	1.0 (1.0, 1.1)	1.0 (0.9, 1.1)	0.646	1.0 (0.9, 1.0)	1.0 (0.9, 1.1)	0.182
	Neighborhood material deprivation	0.93 (0.9, 1.0)∗∗	1.0 (0.9, 1.1)	0.089	1.0 (0.9, 1.0)	1.1 (1.0, 1.2)∗	0.003	0.9 (0.9, 1.0)∗	0.9 (0.8, 1.0)∗	0.384
	Neighborhood social deprivation	0.98 (0.9, 1.0)∗	0.9 (0.9, 1.0)∗∗	0.195	1.0 (0.9, 1.0)∗	0.9 (0.8, 1.0)∗	0.339	1.0 (0.9, 1.0)	0.9 (0.9, 1.0)	0.394
Slope Index of Inequality (absolute gradients)	Household food insecurity status	−6.4 (−11.2, −1.6)∗∗	−12.4 (−18.7, −6.2)∗∗	0.134	−3.4 (−11.2, 4.3)	−13.5 (−22.8, −4.2)∗∗	0.105	−9.2 (−14.9, −3.4)∗∗	−11.6 (−19.3, −3.9)∗∗	0.618
	Individual educational attainment	−9.5 (−12.4, −6.5)∗∗	−10.8 (−14.8, −6.6)∗∗	0.620	−9.3 (−13.4, −5.2)∗∗	−10.9 (−16.7, −5.1)∗∗	0.660	−9.4 (−13.6, −5.4)∗∗	−10.4 (−15.7, −5.1)∗∗	0.773
	Household educational attainment	−11.0 (−13.8, −8.2)∗∗	−11.3 (−15.3, −7.3)∗∗	0.918	−9.1 (−13.1, −5.0)∗∗	−11.4 (−17.0, −5.9)∗∗	0.499	−12.9 (−16.5, −9.2)∗∗	−11.1 (−16.5, −5.7)∗∗	0.597
	Household income adequacy	−1.7 (−4.6, 1.3)	0.3 (−4.0, 4.6)	0.456	0.2 (−4.0, 4.4)	1.0 (−5.2, 7.2)	0.836	−3.4 (−7.3, 0.6)	−0.3 (−6.0, 5.4)	0.375
	Neighborhood material deprivation	−2.5 (−5.6, 0.5)	0.6 (−3.8, 5.0)	0.247	−1.7 (−5.9, 2.6)	4.6 (−1.6, 10.9)	0.102	−3.4 (−7.5, 0.6)	−3.4 (−9.1, 2.3)	0.989
	Neighborhood social deprivation	−2.9 (−5.9, 0.0)	−3.5 (−7.8, 0.9)	0.836	−3.4 (−7.6, 0.8)	−4.2 (−10.5, 2.1)	0.833	−2.3 (−6.1, 1.5)	−2.5 (−8.2, 3.1)	0.950
Relative Index of Inequality (relative gradients)	Household food insecurity status	0.9 (0.8, 1.0)∗∗	0.7 (0.6, 0.9)∗∗	0.055	0.9 (0.8, 1.1)	0.7 (0.6, 0.9)∗∗	0.068	0.8 (0.7, 0.9)∗∗	0.7 (0.6, 0.9)∗∗	0.355
	Individual educational attainment	0.8 (0.8, 0.9)∗∗	0.8 (0.7, 0.9)∗∗	0.203	0.8 (0.8, 0.9)∗∗	0.8 (0.7, 0.9)∗∗	0.344	0.8 (0.8, 0.9)∗∗	0.8 (0.7, 0.9)∗∗	0.370
	Household educational attainment	0.8 (0.8, 0.9)∗∗	0.8 (0.7, 0.8)∗∗	0.328	0.8 (0.8, 0.9)∗∗	0.8 (0.7, 0.9)∗∗	0.231	0.8 (0.7, 0.8)∗∗	0.8 (0.7, 0.9)∗∗	0.825
	Household income adequacy	1.0 (0.9, 1.0)	1.0 (0.9, 1.1)	0.514	1.0 (0.9, 1.1)	1.0 (0.9, 1.2)	0.821	0.9 (0.9, 1.0)	1.0 (0.9, 1.1)	0.475
	Neighborhood material deprivation	1.0 (0.9, 1.0)	1.0 (0.9, 1.1)	0.308	1.0 (0.9, 1.0)	1.1 (1.0, 1.3)	0.117	0.9 (0.9, 1.0)	0.9 (0.8, 1.1)	0.852
	Neighborhood social deprivation	0.9 (0.9, 1.0)	0.9 (0.8, 1.0)	0.663	0.9 (0.9, 1.0)	0.9 (0.8, 1.0)	0.705	1.0 (0.9, 1.0)	0.9 (0.8, 1.1)	0.836

Abbreviation: CI, confidence interval.

Analyses were conducted using multivariable linear regression adjusted for age, dietary recall day, and sex (excluding analyses for males and females). z-scores were used to examine trends between 2004 and 2015.

For within-year inequalities, ∗ and ∗∗ show P < 0.05 and P < 0.001, respectively. Bold highlights statistically significant differences.

¹Data are weighted to be nationally representative of adults aged ≥18 y. Weighted N for the total population in 2004 and 2015 are 23,682,000 and 27,566,000, respectively. Weighted N for males in 2004 and 2015 are 11,836,000 and 13,788,000, respectively. Weighted N for females in 2004 and 2015 are 11,846,000 and 13,778,000, respectively. Figures are rounded in accordance with Statistics Canada’s confidentiality policies.

²Household income adequacy was calculated by adjusting total household income for the low-income cutoffs that correspond to the household and community size of each respondent.

³In 2004, the income adequacy quintiles were 0%–96.5%, 96.6%–153%, 154%–224%, 225%–322%, and >322% of the low-income cutoff. In 2015, the quintiles were 0%–106%, 107%–169%, 170%–249%, 250%–376%, and >376% of the low-income cutoff.

TABLE 4.

Trends in absolute and relative gaps and gradients in the proportion of energy from unprocessed or minimally processed foods by socioeconomic position among adults who participated in the Canadian Community Health Survey-Nutrition 2004 (n = 20,880) or 2015 (n = 13,970)¹

Measure of inequality	Socioeconomic position	Total population			Males			Females
		2004 Mean (95% CI)	2015 Mean (95% CI)	P-trend	2004 Mean (95% CI)	2015 Mean (95% CI)	P-trend	2004 Mean (95% CI)	2015 Mean (95% CI)	P-trend
Absolute gaps	Household food insecurity status	−0.5 (−4.2, 3.2)	4.2 (1.4, 7.0)∗∗	0.046	−3.6 (−10.2, 3.0)	4.0 (−1.2, 9.2)	0.075	2.6 (−0.9, 6.1)	4.3 (1.1, 7.4)∗∗	0.486
	Individual educational attainment	4.8 (2.7, 7.0)∗∗	6.5 (3.7, 9.3)∗∗	0.368	5.9 (3.1, 8.7)∗∗	7.3 (3.6, 11.1)∗∗	0.554	3.7 (0.2, 7.2)∗	5.6 (1.3, 9.9)∗	0.515
	Household educational attainment	6.9 (4.8, 9.0)∗∗	5.9 (3.2, 8.7)∗∗	0.579	7.2 (3.8, 10.5)∗∗	5.3 (1.1, 9.4)∗	0.482	6.8 (4.1, 9.5)∗∗	6.7 (3.0, 10.4)∗∗	0.951
	Household income adequacy2,3	−2.1 (−3.7, −0.5)∗∗	−3.5 (−5.8, −1.2)∗∗	0.338	−1.2 (−3.6, 1.3)	−4.0 (−7.5, −0.6)∗	0.182	−2.9 (−4.9, −0.8)∗∗	−3.2 (−6.2, −0.1)∗	0.877
	Neighborhood material deprivation	0.6 (−1.0, 2.2)	−2.6 (−5.0, −0.1)∗	0.032	0.6 (−1.9, 3.2)	−4.3 (−7.5, −1.1)∗∗	0.018	0.6 (−1.4, 2.6)	−0.7 (−4.1, 2.7)	0.516
	Neighborhood social deprivation	2.3 (0.6, 4.0)∗∗	4.8 (2.4, 7.2)∗∗	0.100	3.5 (0.9, 6.2)∗∗	5.6 (2.1, 9.2)∗∗	0.355	1.1 (−1.0, 3.1)	4.0 (0.9, 7.1)∗	0.124
Relative gaps	Household food insecurity status	1.0 (0.9, 1.1)	1.1 (1.0, 1.2)∗∗	0.063	0.9 (0.7, 1.1)	1.1 (1.0, 1.3)	0.072	1.1 (1.0, 1.2)	1.1 (1.0, 1.2)∗	0.668
	Individual educational attainment	1.2 (1.1, 1.2)∗∗	1.2 (1.1, 1.3)∗∗	0.688	1.2 (1.1, 1.3)∗∗	1.2 (1.1, 1.3)∗∗	0.994	1.1 (1.0, 1.2)∗	1.2 (1.0, 1.3)∗	0.652
	Household educational attainment	1.2 (1.1, 1.3)∗∗	1.2 (1.1, 1.2)∗∗	0.232	1.3 (1.1, 1.4)∗∗	1.1 (1.0, 1.3)∗	0.211	1.2 (1.1, 1.3)∗∗	1.2 (1.1, 1.3)∗∗	0.675
	Household income adequacy	0.9 (0.9, 1.0)∗∗	0.9 (0.9, 1.0)∗∗	0.512	1.0 (0.9, 1.0)	0.9 (0.8, 1.0)∗	0.250	0.9 (0.9, 1.0)∗∗	0.9 (0.9, 1.0)∗	0.911
	Neighborhood material deprivation	1.0 (1.0, 1.1)	0.9 (0.9, 1.0)∗	0.034	1.0 (0.9, 1.1)	0.9 (0.8, 1.0)∗∗	0.024	1.0 (1.0, 1.1)	1.0 (0.9, 1.1)	0.498
	Neighborhood social deprivation	1.1 (1.0, 1.1)∗∗	1.1 (1.1, 1.2)∗∗	0.254	1.1 (1.0, 1.2)∗	1.1 (1.0, 1.2)∗∗	0.660	1.0 (1.0, 1.1)	1.1 (1.0, 1.2)∗	0.201
Slope Index of Inequality (absolute gradients)	Household food insecurity status	2.3 (−2.2, 6.8)	7.7 (2.5, 13.0)∗∗	0.125	0.5 (−6.6, 7.6)	12.4 (4.7, 20.1)∗∗	0.026	3.9 (−1.4, 9.3)	3.7 (−2.9, 10.4)	0.961
	Individual educational attainment	5.8 (3.0, 8.6)∗∗	7.5 (3.8, 11.1)∗∗	0.483	6.5 (2.6, 10.5)∗∗	8.1 (2.9, 13.3)∗∗	0.630	5.0 (1.2, 8.8)∗	6.7 (1.9, 11.6)∗∗	0.587
	Household educational attainment	7.2 (4.5, 10.0)∗∗	8.2 (4.5, 11.8)∗∗	0.691	6.4 (2.5, 10.4)∗∗	9.2 (4.1, 14.3)∗∗	0.401	8.0 (4.5, 11.4)∗∗	7.1 (2.3, 11.8)∗∗	0.764
	Household income adequacy	−2.1 (−4.8, 0.4)	−3.8 (−7.6, 0.0)	0.487	−1.8 (−5.7, 2.1)	−4.7 (−10.3, 1.0)	0.416	−2.5 (−5.9, 0.9)	−2.9 (−7.9, 2.1)	0.904
	Neighborhood material deprivation	0.3 (−2.5, 3.1)	−3.0 (−7.0, 0.9)	0.180	0.2 (−3.9, 4.3)	−3.7 (−9.2, 1.8)	0.268	0.4 (−3.1, 4.0)	−2.4 (−7.4, 2.7)	0.378
	Neighborhood social deprivation	3.2 (0.5, 6.0)∗	4.8 (0.8, 8.7)∗	0.527	4.8 (0.7, 8.8)∗	5.5 (−0.1, 11.2)	0.835	1.6 (−1.8, 5.0)	4.0 (−1.1, 9.1)	0.456
Relative Index of Inequality (relative gradients)	Household food insecurity status	1.1 (0.9, 1.2)	1.2 (1.1, 1.4)∗∗	0.192	1.0 (0.8, 1.3)	1.4 (1.1, 1.6)∗∗	0.041	1.1 (1.0, 1.3)	1.1 (0.9, 1.3)	0.825
	Individual educational attainment	1.2 (1.1, 1.3)∗∗	1.2 (1.1, 1.3)∗∗	0.857	1.2 (1.1, 1.4)∗∗	1.2 (1.1, 1.4)∗∗	0.982	1.2 (1.0, 1.3)∗	1.2 (1.0, 1.3)∗	0.830
	Household educational attainment	1.2 (1.1, 1.4)∗∗	1.2 (1.1, 1.3)∗∗	0.811	1.2 (1.1, 1.4)∗∗	1.3 (1.1, 1.5)∗∗	0.725	1.3 (1.1, 1.4)∗∗	1.2 (1.1, 1.3)∗∗	0.433
	Household income adequacy	0.9 (0.9, 1.0)	0.9 (0.8, 1.0)∗	0.637	0.9 (0.8, 1.1)	0.9 (0.8, 1.0)	0.516	0.9 (0.8, 1.0)	0.9 (0.8, 1.1)	0.961
	Neighborhood material deprivation	1.0 (0.9, 1.1)	0.9 (0.8, 1.0)	0.202	1.0 (0.9, 1.1)	0.9 (0.8, 1.0)	0.307	1.0 (0.9, 1.1)	0.9 (0.8, 1.1)	0.394
	Neighborhood social deprivation	1.1 (1.0, 1.2)∗	1.1 (1.0, 1.2)∗	0.740	1.2 (1.0, 1.3)∗	1.1 (1.0, 1.3)	0.920	1.0 (0.9, 1.2)	1.1 (1.0, 1.2)	0.552

Abbreviation: CI, confidence interval.

Analyses were conducted using multivariable linear regression adjusted for age, dietary recall day, and sex (excluding analyses for males and females). z-scores were used to examine trends between 2004 and 2015.

For within-year inequalities, ∗ and ∗∗ show P < 0.05 and P < 0.001, respectively. Bold highlights statistically significant differences.

¹Data are weighted to be nationally representative of adults aged ≥18 y. Weighted N for the total population in 2004 and 2015 are 23,682,000 and 27,566,000, respectively. Weighted N for males in 2004 and 2015 are 11,836,000 and 13,788,000, respectively. Weighted N for females in 2004 and 2015 are 11,846,000 and 13,778,000, respectively. Figures are rounded in accordance with Statistics Canada’s confidentiality policies.

²Household income adequacy was calculated by adjusting total household income for the low-income cutoffs that correspond to the household and community size of each respondent.

³In 2004, the income adequacy quintiles were 0%–96.5%, 96.6%–153%, 154%–224%, 225%–322%, and >322% of the low-income cutoff. In 2015, the quintiles were 0%–106%, 107%–169%, 170%–249%, 250%–376%, and >376% of the low-income cutoff.

TABLE 5.

Summary of trends in absolute and relative gaps and gradients in the proportion of energy from ultraprocessed foods and unprocessed or minimally processed foods between 2004 and 2015 by socioeconomic position among adults who participated in the Canadian Community Health Survey-Nutrition 2004 (n = 20,880) or 2015 (n = 13,970)¹

	Ultraprocessed foods				Unprocessed or minimally processed foods
	Absolute gaps	Relative gaps	Absolute gradients	Relative gradients	Absolute gaps	Relative gaps	Absolute gradients	Relative gradients
Total population
Household food insecurity status	$\uparrow$	$\uparrow$	$\leftrightarrow$	$\leftrightarrow$	$\uparrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$
Individual educational attainment	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$
Household educational attainment	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$
Household income adequacy	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$
Neighborhood material deprivation	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\downarrow$	$\downarrow$	$\leftrightarrow$	$\leftrightarrow$
Neighborhood social deprivation	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$
Males
Household food insecurity status	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\uparrow$	$\uparrow$
Individual educational attainment	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$
Household educational attainment	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$
Household income adequacy	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$
Neighborhood material deprivation	$\downarrow$	$\downarrow$	$\leftrightarrow$	$\leftrightarrow$	$\downarrow$	$\downarrow$	$\leftrightarrow$	$\leftrightarrow$
Neighborhood social deprivation	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$
Females
Household food insecurity status	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$
Individual educational attainment	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$
Household educational attainment	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$
Household income adequacy	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$
Neighborhood material deprivation	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$
Neighborhood social deprivation	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$	$\leftrightarrow$

Z-scores were used to examine trends between 2004 and 2015.

$\leftrightarrow$ = stable trends; $\uparrow$= inequalities emerged whereby higher SEP groups had more favorable intakes over time, $\downarrow$ = inequalities emerged whereby lower SEP groups had more favorable intakes over time.

¹Data are weighted to be nationally representative of adults aged ≥18 y. Weighted N for the total population in 2004 and 2015 are 23,682,000 and 27,566,000, respectively. Weighted N for males in 2004 and 2015 are 11,836,000 and 13,788,000, respectively.

Absolute gaps and gradients in the proportion of energy from ultraprocessed foods

Within-year

In both 2004 and 2015, there were significant negative absolute gaps in UPF intake according to individual and household educational attainment overall and among males and females (Table 3). There were also significant negative absolute gaps according to neighborhood social deprivation overall, and according to household food insecurity status and neighborhood material deprivation among females. There were no absolute gaps according to household income adequacy overall or among males or females in either year, nor according to neighborhood social deprivation among females.

In both 2004 and 2015, there were significant negative absolute gradients in UPF intake according to individual and household educational attainment overall and among males and females (Table 3). There were also significant negative absolute gradients in UPF intake according to household food insecurity status overall and among females. There were no absolute gradients in UPF intake according to household income adequacy or neighborhood social and material deprivation in either year.

Trends

Overall, higher SEP groups had more favorable intakes over time based on trends in absolute gaps in UPF intake by household food insecurity status {all trends hereafter are presented as mean in 2004 [95% confidence interval (CI)] to mean in 2015 [95% CI]; −1.2% [95% CI: −5.3%, 2.9%] to −7.9% [95% CI: −11.2%, −4.5%]; Table 3}. Among females, absolute gaps in UPF intake remained stable for all indicators of SEP. Among males, lower SEP groups had more favorable intakes over time based on trends in absolute gaps in UPF intake by neighborhood material deprivation [−2.2% (95% CI: −4.6%, 0.1%) to 4.5% (95% CI: 1.1%, 8.0%)]. Absolute gradients in UPF intake remained stable for all SEP indicators overall and among males and females.

Relative gaps and gradients in the proportion of energy from ultraprocessed foods

Within-year

In both 2004 and 2015, there were significant relative gaps (<1.0) in UPF intake according to individual and household educational attainment overall and among males and females (Table 3). There were significant relative gaps (<1.0) in UPF intake according to neighborhood social deprivation overall and among males. There were significant relative gaps (<1.0) in UPF intake according to household food insecurity status and neighborhood material deprivation among females. There were no relative gaps in UPF intake according to household income adequacy overall or among males and females in either year, nor according to neighborhood social deprivation among females.

In both 2004 and 2015, there were significant relative gradients (<1.0) in UPF intake according to individual and household educational attainment overall and among males and females (Table 3). There were also significant relative gradients (<1.0) in UPF intake according to household food insecurity status overall and among females. There were no relative gradients in UPF intake according to household income adequacy or neighborhood social and material deprivation in either year.

Trends

Overall, higher SEP groups had more favorable intakes over time based on trends in relative gaps in UPF intake by household food insecurity status [1.0 (95% CI: 0.9, 1.1) to 0.8 (95% CI: 0.8, 0.9); Table 3]. Among females, relative gaps in UPF intake remained stable. Among males, lower SEP groups had more favorable intakes over time based on trends in relative gaps in UPF intake by neighborhood material deprivation [1.0 (95% CI: 0.9, 1.0) to 1.1 (95% CI: 1.0, 1.2)]. Relative gradients in UPF intake remained stable across all SEP indicators overall and among males and females.

Absolute gaps and gradients in the proportion of energy from unprocessed or minimally processed foods

Within-year

In both 2004 and 2015, there were significant positive absolute gaps in MPF intake according to individual and household educational attainment overall and among males and females (Table 4). There were also significant positive absolute gaps in MPF intake according to neighborhood social deprivation overall and among males. However, there were significant negative absolute gaps in MPF intake according to household income adequacy overall and among females. There were no absolute gaps in MPF intake according to household food insecurity status among males or according to neighborhood material deprivation among females in either year.

In both 2004 and 2015, there were significant positive absolute gradients in MPF intake according to individual and household educational attainment overall and among males and females (Table 4). There were also significant positive absolute gradients in MPF intake according to neighborhood social deprivation overall. There were no absolute gradients in MPF intake according to household income adequacy or neighborhood material deprivation overall or among males and females in either year. Among females, there were no absolute gradients in MPF intake according to household food insecurity status or neighborhood social deprivation in either year.

Trends

Overall, higher SEP groups had more favorable intakes over time based on trends in absolute gaps in MPF intake by household food insecurity status [−0.5% (95% CI: −4.2%, 3.2%) to 4.2% (95% CI: 1.4%, 7.0%)], whereas lower SEP groups had more favorable intakes over time based on trends in absolute gaps by neighborhood material deprivation [0.6% (95% CI: −1.0%, 2.2%) to −2.6% (95% CI: −5.0%, −0.1%); Table 4]. Among females, absolute gaps in MPF intake remained stable for all indicators of SEP. Among males, lower SEP groups had more favorable intakes over time based on trends in absolute gaps in MPF intake by neighborhood material deprivation [0.6% (95% CI: −1.9%, 3.2%) to −4.3% (95% CI: −7.5%, −1.1%)].

Absolute gradients in MPF intake remained stable across all SEP indicators overall and among females (Table 4). Among males, higher SEP groups had more favorable intakes over time based on trends in absolute gradients in MPF intake by household food insecurity status [0.5% (95% CI: −6.6%, 7.6%) to 12.4% (95% CI: 4.7%, 20.1%)].

Relative gaps and gradients in the proportion of energy from unprocessed or minimally processed foods

Within-year

In both 2004 and 2015, there were significant relative gaps (>1.0) in MPF intake according to individual and household educational attainment overall and among males and females (Table 4). There were also significant relative gaps (>1.0) in MPF intake according to neighborhood social deprivation overall and among males. Additionally, there were significant relative gaps (<1.0) in MPF intake according to household income adequacy overall and among females. There were no relative gaps in MPF intake according to household food insecurity status among males or according to neighborhood material deprivation among females in either year.

In both 2004 and 2015, there were significant relative gradients (>1.0) in MPF intake according to individual and household educational attainment overall and among males and females (Table 4). There were also significant relative gradients (>1.0) in MPF intake according to neighborhood social deprivation overall. There were no relative gradients according to neighborhood material deprivation overall or among males and females in either year, nor according to household income adequacy in males or females. There were also no relative gradients according to household food insecurity status or neighborhood social deprivation among females in either year.

Trends

Overall, lower SEP groups had more favorable intakes over time based on trends in relative gaps in MPF intake by neighborhood material deprivation [1.0 (95% CI: 1.0, 1.1) to 0.9 (95% CI: 0.9, 1.0)] as well as among males [1.0 (95% CI: 0.9, 1.1) to 0.9 (95% CI: 0.8, 1.0); Table 4]. Among females, relative gaps in MPF intake remained stable for all indicators of SEP.

Overall and among females, relative gradients in MPF intake remained stable for all SEP indicators (Table 4). Among males, higher SEP groups had more favorable intakes over time based on trends in relative gradients in MPF intake by household food insecurity status [1.0 (95% CI: 0.8, 1.3) to 1.4 (95% CI: 1.1, 1.6)].

Absolute and relative gaps and gradients in the proportion of energy from processed culinary ingredients and processed foods are presented in Supplemental Tables 5 and 6.

Sensitivity analyses

With very few exceptions, trends for most indicators of SEP remained identical to the main analyses (Supplemental Tables 7 and 10). After adjusting for Indigenous status and race/ethnicity, trends in absolute and relative gradients in MPF intake according to household food insecurity status became statistically nonsignificant among males. After adjusting for dietary misreporting, several trends in absolute and relative gaps and gradients in UPF and MPF intake according to household food insecurity status became significant overall and among females, indicating that higher SEP groups had more favorable intakes over time. The sensitivity analyses for trends in absolute and relative gaps and gradients in the proportion of energy from processed culinary ingredients and processed foods are presented in .Supplemental Tables 8 and 9

Discussion

This study provided the most comprehensive analyses of trends in inequalities in UPF and MPF intake internationally according to 6 indicators of SEP at the individual, household, and area levels. We also examined absolute and relative gaps and gradients in UPF and MPF intake—none of which have ever been quantified—and used nationally representative data. Four key findings emerged from this research. First, the largest and most consistent within-year absolute and relative gaps and gradients in UPF and MPF intake were found according to individual and household educational attainment. Conversely, there were no inequalities in UPF intake according to household income adequacy, because the lowest income groups actually consumed more MPF than those with higher incomes. Second, between 2004 and 2015, several inequalities in UPF and MPF intake emerged according to household food insecurity status in the overall population and among males, whereby higher SEP groups had more favorable intakes over time. Third, several inequalities in UPF and MPF intake emerged according to neighborhood material deprivation overall and among males whereby lower SEP groups had more favorable intakes over time. Fourth, all trends in absolute and relative gaps and gradients by SEP remained stable among females.

Our first key finding was that absolute and relative gaps and gradients in UPF and MPF intake were the largest and most consistent according to individual and household educational attainment in both years. The size of absolute gaps and gradients by educational attainment was considerable and likely to be clinically meaningful [13,14]. These results align with prior nationally representative studies in the United States [1,18] and Canada [20,22], including our prior findings that educational attainment was a more important determinant of inequalities in diet quality than household income and neighborhood deprivation among adults in Canada [56]. Individuals with a higher educational attainment may adopt healthier dietary patterns because of their greater cultural capital, which encompasses factors such as health-related values, norms, knowledge, and skills [23,25,57]. Given that highly processed foods are often regarded as unhealthy, inexpensive, mass-produced, and of lower quality than less processed foods, more educated groups may also consume fewer UPF as a form of social distinction to set themselves apart from less educated groups [24,58,59].

We did not find any inequalities in UPF intake according to household income adequacy. These results are similar to previous studies in Canada [3,22] and our prior study that found small absolute and relative gaps and gradients in diet quality by household income adequacy [56]. Although dietary inequalities are often attributed to the higher cost of healthier foods [26,60], findings regarding the importance of educational attainment compared with income-based indicators of SEP suggest that mechanisms underlying inequalities in UPF intake extend beyond the cost of food. It is also plausible that the lack of significant inequalities in UPF intake by income adequacy may be attributed to qualitative differences in the types of UPF consumed. Higher income individuals might opt for UPF with added functional properties—such as fiber or probiotics—that are usually more costly [61].

Nevertheless, we did find a small number of absolute and relative gaps and gradients in MPF intake according to household income adequacy. However, they were in the reverse direction of what was expected, such that individuals in the lowest household income adequacy quintile actually consumed more MPF than those in the highest household income adequacy quintile. This aligns with previous research that also found higher intake of MPF among lower income groups compared with higher income groups [1]. These findings could be attributed to the relatively low costs of several minimally processed staple foods such as rice, oats, potatoes, corn, and dried beans and lentils [62] that lower income groups may be more reliant on. Absolute differences in MPF intake between the highest and lowest income quintiles were, however, relatively modest, ranging from 1.2% to 4.0% of energy intake, and an absolute gradient was not found across the entire socioeconomic spectrum. Future studies should examine the specific types of MPF consumed by different income groups to enhance understanding of these findings.

Contrary to our findings for household income adequacy, absolute and relative gaps and gradients in UPF and MPF intake in relation to household food insecurity status tended to be larger and in the expected direction. These findings are consistent with a study finding strong and graded associations between household food insecurity and UPF intake in Canada [19]. Although the patterning of UPF and MPF intake according to household income adequacy and food insecurity might appear inconsistent, it is important to note the distinctions between these 2 measures. Household income adequacy is an objective indicator of household income over the past year in relation to a low-income threshold that does not account for factors such as household debt or wealth. Household food insecurity status, by contrast, is a subjective indicator of whether a household’s available resources were sufficient to ensure sustained and secure access to food over the past year [19,27]. Thus, this measure inherently accounts for all financial constraints a household may have encountered in relation to procuring adequate food—such as repayment of debt, unexpected expenses, and limited assets—and any fluctuations in these factors. In this way, household food insecurity status may provide a more comprehensive perspective of a household’s financial situation. For these reasons, it is not entirely unexpected that patterns for these indicators differed.

In 2015, females living in materially deprived neighborhoods had higher UPF and similar MPF intake as those living in less deprived areas. By contrast, males living in the most materially deprived neighborhoods consumed less UPF and more MPF than those living in the least materially deprived areas. These findings may suggest a differential sensitivity to neighborhood material conditions. It is possible that the diets of females are more constrained by their neighborhood environments because of time (for example, managing work, household, and caregiving responsibilities) and cost-related barriers (for example, females tend to have lower incomes and may therefore have fewer transportation options) that limit their opportunities to travel outside of their residential neighborhoods to procure less processed food. Notably, however, these sex-based differences were confined to the extreme ends of the socioeconomic continuum because there were no gradients in UPF or MPF intake among males or females. They also appear to relate exclusively to food processing, because absolute and relative gaps and gradients in diet quality according to neighborhood deprivation were similar and in the expected direction among males and females in our prior study, although in that study we did not decompose neighborhood deprivation into its material and social dimensions [56].

Our second key finding was that between 2004 and 2015 several inequalities in UPF and MPF intake emerged according to household food insecurity status in the overall population and among males because of larger declines in the intake of UPF and larger increases in the intake of MPF among food secure relative to food insecure groups. Several of these trends also became statistically significant among females when we additionally adjusted for dietary misreporting. Our findings may be attributable to the considerable increase in food costs (30%–50%) and other expenses that outpaced average income growth (13.4%) in Canada from 2004 to 2015 [63,64]. This may have led to a higher reliance on more affordable and accessible foods such as UPF among food insecure households [65].

Our third key finding was that several inequalities in UPF and MPF intake emerged according to neighborhood material deprivation overall and among males because of larger decreases in UPF and larger increases in MPF intake among those living in the most materially deprived neighborhoods compared with the least materially deprived neighborhoods. As such, these inequalities were in the opposite direction of what was expected as they favored lower SEP groups. These patterns mirror our findings on the within-year patterning of these inequalities whereby the diets of males appeared to be less affected by neighborhood material deprivation. However, these trends were not evident across the entire socioeconomic spectrum and thus the explanations for these trends remain unclear. Moreover, we have previously shown that absolute and relative gaps and gradients in diet quality according to neighborhood material deprivation remained unchanged during this time frame [56] and thus these trends were limited to the level of food processing.

Finally, in our main analyses all trends in gaps and gradients by SEP remained stable among females. These stable trends contrast with the aforementioned trends among males. However, as mentioned, when we adjusted for dietary misreporting, trends in absolute and relative gaps and gradients in UPF and MPF intake among females according to household food insecurity status emerged in favor of higher SEP groups. Thus, the lack of significant trends among females in our main analyses may reflect the greater tendency of females to misreport their dietary intake [66] and any differential trends by sex may be minimal. It is important to point out that although we were limited to analyzing biological sex, <0.3% of adults in Canada identify as transgender [67] and thus differential trends by sex likely reflect differences in gender roles and identities.

Several limitations of our analyses should be considered. First, potential misreporting bias from 24-h dietary recalls may have influenced the estimates of UPF and MPF intake. However, after adjusting for potential dietary misreporting, dietary inequalities persisted in most instances. Second, although a single recall is recommended to describe mean intake at a population level, one recall cannot fully capture usual intake, particularly of episodically consumed foods [48]. Third, although the CCHS captures 98% of the Canadian population, it does not include individuals living on Indigenous reserves or in Canada’s 3 territories, which have high rates of household food insecurity [39]. Fourth, the classification of foods as UPF and MPF within the CCHS data involves inherent challenges because of the evolving and sometimes ambiguous Nova definitions of level of food processing [68]. Finally, we conducted multiple statistical tests, potentially increasing risk of type I error. However, our results appear unlikely to be due to chance given that we found consistent findings across multiple analyses for each indicator of SEP.

In conclusion, in Canada, the largest and most consistent absolute and relative gaps and gradients in UPF and MPF intake were for individual and household educational attainment. Between 2004 and 2015, several inequalities in UPF and MPF intake emerged according to household food insecurity status in the overall population and among males that favored higher SEP groups. After adjusting for dietary misreporting, similar trends also emerged among females. Inequalities in UPF and MPF intake also emerged according to neighborhood material deprivation overall and among males whereby lower SEP groups had more favorable intakes over time. These findings indicate that educational attainment and household food insecurity status may be key contributors to socioeconomic inequalities in intake of UPF and MPF among adults in Canada.

Author contributions

The authors’ responsibilities were as follows—SHP: conceptualization, methodology, formal analysis, data interpretation, writing—original draft, writing—review and editing; MLA: data interpretation, writing—original draft, writing—review and editing; JCM, JP: methodology, data interpretation, writing—review and editing; LV: data interpretation, writing—review and editing, funding acquisition; DLO: conceptualization, methodology, data interpretation, writing—original draft, writing—review and editing, supervision, funding acquisition. SHP, DLO: responsible for the final content of the manuscript; and all authors: read and approved the final manuscript.

Conflict of interest

The authors report no conflicts of interest.

Funding

This work was supported by the Canadian Institutes of Health Research (FRN 155916; PJT-173367). The study funders had no role in study design, in collection, analysis, or interpretation of the data, in writing the manuscript or in the decision to submit it for publication.

Data availability

The data that support the findings of this study are available from Statistics Canada; however, restrictions apply to the availability of these data. Statistics Canada permits approved researchers to access its data sets within designated Research Data Centres (https://www.statcan.gc.ca/eng/microdata/data-centres).

Appendix A. Supplementary data

The following is the Supplementary data to this article: