The quality of the diets of Canadian children is often poor as they fail to meet the recommended intakes for fruit, vegetables, and sodium, with a significant portion of their energy intake coming from high-fat, sugary foods (1–3). It is well established that consuming diets of good quality is essential for growth and physical development, whereas emerging evidence points to their importance for mental health and well-being, as well as academic performance (4–6). Good quality diets have been associated with better health outcomes, including body weights, blood pressure, metabolic syndrome, mental health and academic performance (7–9). An accurate measure of diet quality is essential to study the importance of good diet quality and to assess the effectiveness of dietary interventions.

Several indices have been used to assess the quality of the whole diet of Canadian children (8, 10, 11). The most commonly used indices are the Diet Quality Index-International (DQI-I) (10, 12, 13), Healthy Eating Index-Canada 2015 (HEI-C 2015) (11, 14, 15), and the more recently developed Healthy Eating Index-2019 (HEFI-2019) (16). While all three indices aim to summarize the quality of the whole diet using established dietary recommendations, they consider different sets of criteria and scoring systems (17), hindering the interpretation of dietary guidelines to improve diet quality. It is not clear which index best captures diet quality of different population subgroups and should be used for research, policy and/or practice to help support dietary research and policy development, and promote dietary guidelines implementation. An optimal diet quality index should have a wide range of scores to distinguish between individuals who consume diets of different quality. Additionally, a diet quality index should be independent of diet quantity and be able to effectively differentiate between different population subgroupings (e.g., gender, age, socioeconomic status (SES), rural/urban residence) (18). Indeed, while some studies found little to no difference in diet quality between girls and boys (10, 19), government reports suggest that Canadian girls consume, on average, more vegetables and fruit and less sodium, compared with boys (2, 3). Moreover, younger children consume diets of higher quality than older children (19, 20), and children from families with higher SES have healthier dietary patterns, characterized by lower consumption of energy-dense foods and higher consumption of fruit, vegetables, and dairy products, compared with their peers from lower SES households (21–23). To our knowledge, no previous studies compared these three commonly used diet quality indices with respect to their ability to capture diets of different quality. This exploratory study sought to assess the agreement between HEI-C 2015, DQI-I, and HEFI-2019, and describe the ability of each index to discriminate the diet quality of Canadian children across population subgroups.

Data from 336 (2016), 454 (2018), and 909 (2020/21) students were available for analysis. Student and school characteristics are shown in Table 1. Over half (51.4%) were girls. About one-third were in grade 4, one-third in grade 5, and one-third in grade 6 (30.6, 37.1, 32.3%, respectively). Of 25 participating schools, 60.8% were located in urban areas. HEI-C 2015 and DQI-I had similar average scores and trends over time in each cross-sectional sample. Between 2016 and 2020/2021 HEI-C 2015 declined from 54.7 (SD = 13.9) to 49.5 (SD = 12.9), and DQI-I from 55.6 (SD = 9.7) to 53.2 (SD = 9.9). The HEFI-2019 scores were markedly lower than HEI-C 2015 and DQI-I in each cross-sectional sample and showed little variation over time: 45.0 (SD = 13.9) in 2016, 44.7 (SD = 13.1) in 2018, and 44.9 (SD = 13.6) in 2020/2021 (Table 1). Overall, the distribution of DQI-I scores had the lowest variability, ranging from 19.9 to 83.6 (CV = 18.1%), compared to HEI-C 2015 that ranged from 11.7 to 95.3 (CV = 25.8%) and HEFI-2019 from 8.6 to 90.1 (CV = 30.1%), with the latter having the widest range of scores.

Percent agreement and weighted kappa scores varied across the survey years but were statistically significant for all comparisons (Table 2). In a combined sample of students who participated in any of the survey cycles, percent agreement between HEI-C 2015 and DQI-I was 0.88 (95% Confidence Interval [CI]: 0.87, 0.89), between HEI-C 2015 and HEFI-2019 – 0.83 (95% CI: 0.82, 0.84), and between DQI-I and HEFI-2019 – 0.80 (95% CI: 0.79, 0.81). For this combined sample, weighted Cohen’s kappa coefficient for agreement between HEI-C 2015 and DQI-I was 0.55 (95% CI: 0.52, 0.59), between HEI-C 2015 and HEFI-2019 0.38 (95% CI: 0.35, 0.42), and between DQI-I and HEFI-2019 0.29 (95% CI: 0.25, 0.33). These values of weighted Cohen’s kappa coefficients translate into fair to moderate agreement (41).

Relative to girls, boys reported diets of lower quality, with the difference being particularly pronounced for DQI-I in both univariate (β = −1.37, 95% CI: −2.31, −0.43) and multivariable models (β = −1.44, 95% CI: −2.38, −0.50) (Table 3). There were no statistically significant differences in diet quality scores across grade levels regardless of the index used. Students attending schools in more vs. less materially deprived areas appeared to have worse diet quality irrespective of the index used, and these differences remained robust after adjusting for covariates (sex, grade level, energy intake, social deprivation, and geographic region). However, diet quality was higher in more vs. less socially deprived areas when using HEI-C 2015 (β = 1.24, 95% CI -0.09, 2.57) and DQI-I (β = 1.03, 95% CI: 0.05, 2.02) and after adjusting for covariates (sex, grade level, energy intake, material deprivation, and geographic region). Differences in diet quality according to geographic region were found for HEI-C 2015 but not for DQI-I and HEFI-2019. Finally, energy intake was positively associated with DQI-I (β = 0.09, 95% CI: 0.04, 0.14) and negatively associated with HEFI-2019 (β = −0.19, 95% CI: −0.25, −0.12).

This study compared diet quality derived using three commonly used summary measures (HEI-C 2015, DQI-I, HEFI-2019) among grade 4–6 students from 25 elementary schools in western Canada. The three indices have different properties (e.g., dietary components assessed, range of values, coefficient of variation), with HEFI-2019 demonstrating the widest range of scores and DQI-I the narrowest variation in the scores. The three indices demonstrated fair to moderate agreement (41). The ability of the indices to discriminate the quality of diets between different population subgroups of Canadian children varied, yielding different conclusions with respect to the determinants of children’s diet quality. Also, higher energy consumption was associated with higher DQI-I and lower HEFI-2019 scores, with the strongest association for HEFI-2019.

This study revealed that, compared with the international index (DQI-I), the two Canadian indices (HEI-C 2015 and HEFI-2019) appear to have more variation in scores. In particular, the index that was developed specifically for Canadian diets (HEFI-2019) showed the widest variation in scores, suggesting it may better capture diets of lower and higher quality. No HEFI-19 scores have been previously reported specifically for children (16). It is therefore not possible to assess whether our scores align with the literature, albeit it is feasible the scores for all three indices may be lower in our sample derived from socioeconomically disadvantaged communities in western Canada. Fair agreement was found between HEI-C 2015 and HEFI-2019 and between DQI-I and HEFI-2019, while a moderate agreement was found between HEI-C 2015 and DQI-I. The latter finding is not surprising as both HEI-C 2015 and DQI-I use similar dietary components (adequacy, moderation) as opposed to HEFI-2019. The correlation between HEI-C 2015 and HEFI-2019 has been previously reported to be 0.79 (16), while in our sample it was as low as 0.6 (data not shown), which may be attributed to our use of a HEI-2015 version that was adapted for the Canadian population (28–30, 34). Finally, higher calorie intake was found to be associated with higher DQI-I but lower HEFI-2019 scores, with HEFI-2019 having the strongest association with energy intake. Brassard et al. also reported the inverse relationship between energy intake and HEFI-2019 scores, proposing that it may be driven by two components which had the highest inverse correlation with energy intake: beverages and vegetables and fruit (16). HEI-C 2015 had no statistically significant association with energy intake in our study. Brassard et al. used the US HEI-2015 and also found no relationship with energy intake since each of its components is divided by total energy intake.

Consistent with government reports (2, 3), our results revealed that boys had worse diet quality than girls regardless of the index used, with DQI-I being the most robust at differentiating diet quality between girls vs. boys. The fact that, unlike HEI-C 2015 and HEFI-2019, DQI-I includes certain nutrients and dietary components (cholesterol, vitamin C, and macronutrient ratio) may have contributed to this finding. Our comparisons across grade levels revealed no statistically significant differences in diet quality regardless of the index used. This could be due to the narrow age range of children in our sample (9–12 years old), whereas previous studies in samples with a wider age range of children demonstrated statistically significant differences (1, 15, 19).

It has been previously demonstrated that children from lower SES families consume less fruit, vegetables and fibre, and more added sugar and energy drinks (42, 43). Our findings corroborate this evidence and show that regardless of the diet quality index used, students from more materially deprived neighbourhoods report worse diet quality, with HEI-C 2015 and HEFI-2019 better capturing these differences. Although diet quality appeared to be higher in more vs. less socially deprived areas, previous studies that reported on the association between social deprivation and diet yielded inconsistent findings (15, 44), possibly due to differences in diet quality indices used, covariates adjusted for, and characteristics of the study sample.

Except when using HEI-C 2015, no difference in diet quality was found between students residing in rural vs. urban areas. Similarly, Tugault et al. reported no significant differences in diet quality between rural vs. urban areas in a sample of 4,728 students 6–17 years old (15). While HEI-C 2015 was able to detect differences in diets between geographic regions, this could be due to smaller sample sizes and surveying participants from smaller geographic areas and no participants from metropolitan areas in our study.

To the best of our knowledge, this is the first study to compare established diet quality indices in a population-based sample of Canadian school-aged children. Data were collected through 24-h dietary recall with a sample size large enough to capture sufficient variation in diet quality across population subgroups. However, there are few limitations to consider. Collecting a single 24-h dietary recall on weekdays does not capture participants’ usual dietary intake; yet collecting data from multiple 24-h dietary recalls and including weekend days is not feasible in school-based studies. The findings of the present study should therefore be interpreted as being based on a single 24-h recall rather than being based on a comparison of usual intake. Since participating schools are located in socioeconomically disadvantaged communities and have an active health promotion intervention in place, children’s diets may differ from those of the general population. However, this does not affect the comparison of the three indices, the indices were able to capture variability in children’s diets as well as variation across population subgroups. All data were self-reported which may be subject to social desirability and measurement bias.

In sum, this study shows that the choice of a diet quality index affects the interpretation of the results and practical considerations. Therefore, researchers, practitioners, and policymakers must seek consensus on which index to use and under which circumstances. Of the three indices examined, HEFI-2019 has been developed most recently specifically for Canadian diets. It reflects adherence to the dietary recommendation outlined in the latest Canada Food Guide and our current understanding of diet quality and how it should be measured. However, adjustments to HEFI-2019 may be needed to circumvent its dependency on diet quantity.