Socio-demographic correlates of ultra-processed food consumption in Canada

Data for this study came from the 2015 Canadian Community Health Survey–Nutrition ⁽²³⁾. The 2015 Canadian Community Health Survey–Nutrition has a sample size of 20 487 respondents, in which the target population were Canadian household residents aged one year or older living in the ten Canadian provinces. Full-time members of the Canadian Forces and individuals who lived on reserves or in other Indigenous settlements, in some remote areas, or in institutions were excluded from this study. The overall survey response rate was 61·6 %. The two components of the survey completed by all respondents were (1) a general demographic questionnaire and health questions and (2) a dietary recall of all foods and drinks consumed, including descriptions and amounts, in the 24 h prior to the interview day. About 30 % of participants completed a second 24-hour dietary recall 3–10 d after the initial interview; however, these data were not used in this study because group means from a single dietary recall represent unbiased estimates at the population level⁽²⁴⁾.

Data were mainly collected in person by trained interviewers for the first recall and via telephone for the second recall. The Automated Multiple Pass Method adapted for Canada was used to help respondents maximise their dietary recall. For children under age 6, a parent or guardian provided information; for children 6–11, the interview was conducted with the child, with help from a parent; respondents aged 12 years and older provided their own information.

The present study sample was composed of Canadians aged 2 years or older who responded to the 2015 Canadian Community Health Survey–Nutrition. After excluding respondents younger than age 2 (n 372) and those who did not consume any energies on the previous day (n 12), the final analytic sample was 20 103 persons.

All food and drinks (excluding alcoholic drinks, which are not readily classifiable by NOVA) were categorised according to the NOVA classification system into four mutually exclusive groups: unprocessed or minimally processed foods; processed culinary ingredients; processed foods or UPF.⁽⁷⁾ The UPF group was further broken down into 13 subgroups. Classification of food items according to NOVA proceeded in two phases, following a previously described protocol⁽²²⁾. Briefly, in the first phase, all ingredients, basic foods (i.e. foods that cannot be broken down into underlying ingredients like an apple or milk) and recipes without nutritional information available (e.g. some granola bars) were classified into one of four NOVA groups based on food item description. Energy values were based on the reported food amount converted into gram weight and were derived from the Canadian Nutrition File version 2015. In phase two, mixed dishes were searched to flag frozen meals, lunch kits and common ultra-processed dishes (e.g. burger, pizza and donut). If the flagged dish was consumed in a quick-service setting (e.g. pizza or fast-food restaurant), then all of its underlying ingredients were re-classified as UPF (subgroup ‘fast-food and frozen dishes’). For example, if a hamburger was consumed in a fast-food restaurant, then all of its underlying ingredients (e.g. bun, meat patty, tomato, lettuce, condiments) were re-classified as UPF. However, if the same hamburger was consumed at home, then phase one classification was maintained (i.e. bun and condiments categorised as UPF and meat and vegetables as processed or minimally processed).

Consumption of foods and drinks according to each of the four NOVA groups was defined as their relative energy contribution, i.e. the percentage of total daily energy intake (kcal) from each NOVA group or UPF subgroup.

The choice of socio-demographic variables included in this study was informed by previous literature and data available on the survey. The socio-demographic variables included were sex (male or female), age (grouped as 2–5, 6–12, 13–18, 19–30, 31–54 or 55+ years) and the highest level of education attained in the household (less than high school, high school, certificate or diploma, or university degree or above). Household income adequacy (grouped into quintiles) was calculated as the adjusted ratio of total household income to the low-income cut-off corresponding to the household and community size. Immigrant status categorised respondents as either non-immigrant, recent immigrant (immigrated < 10 years ago), or long-term immigrant (immigrated to Canada 10 years ago or more). Household food security status was assessed using the Household Food Security Survey Module, which consists of eighteen questions that assess the income-related food security situation in the household in the previous 12 months. Respondents’ household food security status was categorised as food secure, moderately food insecure or severely food insecure. Region of residence was categorised as Atlantic provinces (New Brunswick, Nova Scotia, Prince Edward Island, Newfoundland and Labrador), Quebec, Ontario, Prairie provinces (Manitoba, Saskatchewan, Alberta) and British Columbia. Data for residents of the Atlantic and Prairie provinces were collapsed because of small sample sizes.

The mean energy contribution (% of total daily energy intake) according to NOVA groups and UPF subgroups were generated overall and by respondent socio-demographic characteristics. Only the first 24-hour dietary recall was used to estimate the mean energy contributions from NOVA groups/subgroups. This is because mean intakes estimated from a single dietary recall are equivalent to usual (i.e. habitual) intakes at the population level⁽²⁴⁾.

Associations between all socio-demographic characteristics and the mean energy contribution of UPF were assessed using multivariable linear regression models. Total UPF and each UPF subgroup were modelled separately. In addition, predicted mean energy contributions of UPF subgroups were generated for selected socio-demographic variables from the fully adjusted linear regression models. All analyses were conducted in SAS 9.4 and applied survey sampling weights provided by Statistics Canada to account for the complex sampling design and unequal probability of selection. Bootstrap weights were used to calculate robust se using the Balanced Repeated Replication method. Statistical significance was set at an alpha level of 0·05.

In 2015, 40·6 % of total daily energy intake, on average, came from unprocessed or minimally processed foods and 7·1 % came from processed culinary ingredients (Table 1). Processed foods contributed 7·4 % of total daily energy and UPF contributed 44·9 %. Within UPF subgroups, most energies came from commercial breads, which contributed 10·4 % of total daily energy, followed by margarine (3·6 %), commercial fruit juices and drinks (3·6 %), and sauces, spreads and salad dressings (3·4 %).

In the study sample, just over half of the respondents (50·7 %) were female and aged 31–54 years (50·1 %) (Table 2). Forty per cent of participants were in households where the highest level of education attained was a university degree or higher, followed by the certificate or diploma (37·4 %). Approximately three-quarters of participants (76·1 %) were non-immigrants to Canada, 7·1 % were recent immigrants (<10 years since immigration) and 16·8 % were long-term immigrants (10+ years since immigration). Household food insecurity (moderate or severe) was reported by 8·4 % of survey participants.

Table 2 also shows the mean energy contribution (% of total daily energy) according to NOVA groups by socio-demographic characteristics. There was more variation across several socio-demographic groups in the mean percentage of energy from unprocessed or minimally processed foods and UPF than from culinary ingredients and processed foods. The mean percentage of energy from unprocessed or minimally processed foods ranged from 35·0 % among respondents living in households with severe food insecurity to 50·8 % among recent immigrants. Conversely, the energy contribution from UPF was lowest among recent immigrants (33·3 %) and highest among the severely food insecure (53·3 %). There was also some variation across age groups, education level and region of residence, but little variation by sex and household income adequacy.

Tables 3(a) and (b) present mutually adjusted associations of socio-demographic characteristics with total UPF consumption (mean % of total daily energy) and subgroups of UPF. For total UPF consumption, there were statistically significant associations for all socio-demographic characteristics examined. For example, compared to children aged 2–5 years, children aged 6–12 years on average consumed 5·3 percentage points more total daily energy from UPF (β = 5·3; P-value < 0·0001), while adults aged 31–54 consumed 2·7 fewer percentage points (β = –2·7; P-value < 0·0001). Differences in total UPF consumption were most notable between immigrants v. non-immigrants and between those living in food-secure v. food-insecure households.

Mutually adjusted associations of socio-demographic characteristics and the mean percentage of energy derived from 13 subgroups of UPF revealed differences across several characteristics, most notably, age group, immigrant status and region of residence (Tables 3(a) and (b)). The relative energy contribution of UPF subgroups varied substantially by age group. For example, soft drinks contributed, on average, substantially more relative energy for children, adolescents and adults under age 55 compared with the youngest children aged 2–5, while fruit juices and fruit drinks contributed significantly less energy for adults. Notable differences between adults and children were also seen for relative energy from sauces, spreads and salad dressings; sweetened milk and soy-based products; cakes, cookies and other pastries; and sweetened breakfast cereals. Fast-food and frozen dishes contributed significantly more relative energy for adolescents and younger adults, and commercial breads and margarine contributed more relative energy for older adults aged 55+ compared to the youngest children.

Immigrants reported consistently lower shares of energy from virtually all UPF subgroups than non-immigrants, particularly from sauces, spreads and salad dressings; fast-food and frozen dishes; and chips, crackers and other salty snacks. Among recent immigrants, the mean energy contribution from commercial breads, sweetened milk- and soy-based products, processed meat products and ‘other’ UPF (e.g. meal replacements, protein shakes, sweeteners, imitation fish/meats) was notably lower than among non-immigrants.

For region of residence, residents of Atlantic provinces consumed a higher share of energy from nearly all UPF subgroups compared to residents of other regions, although differences were not always statistically significant.

There were some differences between males and females in the mean energy contribution from UPF subgroups, particularly for processed meat products (β = –1·1 for females v. males; P-value < 0·001). For household food security status, the share of energy from UPF subgroups was typically higher among those in food-insecure households v. food-secure ones, although few differences were statistically significant. Few differences were seen across household education or income adequacy strata.

As a complement to the multivariable linear regression results shown in Tables 3(a) and (b), Fig. 1 panels (a) through (c) present the predicted mean percentages of total energy derived from UPF subgroups for selected socio-demographic variables, adjusted for all other socio-demographic covariates under study. Children aged 6–12 and adolescents aged 13–18 consumed over half of total daily energy from UPF (adjusted means of 51·9 % and 50·7 %, respectively; Fig. 1, panel A). The largest absolute differences in the mean percentage of energy from UPF (of at least 9 percentage points), adjusted for all other socio-demographic variables, were observed between recent and long-term immigrants v. non-immigrants (Fig. 1, panel B).

This study is strengthened by the use of a large, national-level sample and examination of a range of socio-demographic characteristics. However, several limitations deserve mention. During the classification of food items into NOVA groups, some foods and drinks may have been misclassified because of the lack of details available about product brands and type of food processing. For example, some mixed dishes should be disaggregated into underlying ingredients and classified as minimally processed if they were prepared at home from scratch ingredients v. ultra-processed if prepared industrially (e.g. pizza). In this study, such dishes were classified as UPF only if they were consumed in a quick-service setting (e.g. a fast-food restaurant) but not if consumed outside of the food service setting (e.g. as takeout or delivery) because information on the type of food processing and place of food preparation was not consistently available. This may have attenuated the level of UPF consumption and the associations under study. Furthermore, social desirability bias may have caused underreporting of certain foods (e.g. fast foods, salty snacks, soft drinks), leading to underestimation of UPF consumption. If this misreporting differed across socio-demographic groups, this could have led to either underestimation or overestimation of the associations under study. While data from a single 24-hour dietary recall are useful for estimating group means, as was done in this study, they do not fully capture all intra-person variability and thus may not represent the usual dietary intakes of individuals. Further, as with all observational research, we cannot rule out the presence of residual confounding by unmeasured or mismeasured characteristics. Finally, the results are based on data collected in 2015. While these data represent the most recently available national-level nutrition survey data for Canada, it will be important to examine any changes in socio-economic patterning of UPF consumption using more recent data, particularly following the release of an updated Canada Food Guide in 2019⁽⁴⁹⁾ and the rising cost of living since 2021⁽⁵⁰⁾.

This large, national-level study examined variations in UPF intake among Canadian children and adults across socio-demographic characteristics, namely age, sex, education, income adequacy, immigrant status, household food security status and region of residence. Although UPF consumption was pervasive among all socio-demographic groups, the most substantive differences in the mean energy contribution of UPF were observed by age group, immigrant status and household food security status. Because the traditional food practices of many immigrant communities in Canada are known to be healthy and low in UPF, this points to the need for efforts to slow down the process of dietary acculturation and to protect and promote these rich culinary traditions across the population at large. Our findings also suggest that policies and programmes to promote healthier eating among young persons and to curb the prevalence of household food insecurity could contribute to not only reducing UPF intake but also mitigating social inequalities in the consumption of these products. Given the overall high consumption of UPF in Canada, the results of this study support policies and interventions to promote environments favouring healthy diets and to reduce exposure to UPF among all socio-demographic segments of the population.