The data in this study are from a cross-sectional survey of dietary intake, with anthropometric measurements collected from 720 primary school aged children (nine-12 yrs) from 29 randomly selected government schools in the Hunter Region of New South Wales (NSW), Australia in 2005 [²³].

Body weight measurements were taken in light clothing without shoes using A&D Personal Precision Scale, UC 321, Accurate to 50 g (A&D Engineering, Inc. San Jose, California). Height was measured without shoes using a Harpenden Portable Stadiometer (98.603) classified as Class 1 under EC directive 93/42/EWG (Holtain Ltd. Crosswell, Crymych, UK-developed in collaboration with the Institute of Child Health at the University of London). Two measurements were taken for weight and height and the average calculated. A third measurement was taken for height and weight when the difference between the first two measurements was more than 0.5 centimetres or 0.5 kilograms, respectively, with the closest two measurements used to calculate the average. Average height and weight measures were used to calculate BMI using the formula: BMI = weight (kg)/[height (m)]². Anthropometric data was transformed into BMI z-scores using the LMS method [²⁴,²⁵] based on reference data from the 1990 British Growth Reference [²⁶].Weight status was classified as healthy weight, overweight or obese using the UK age and gender-specific BMI z-score cut-off points identified by Cole et al. [²⁷] corresponding to a BMI of 25 for overweight and 30 for obese at 18 years of age.

Dietary intake was measured using the Australian Child and Adolescent Eating Survey (ACAES), a 135-item semi-quantitative FFQ with 120 food items and 15 supplementary questions addressing age, food behaviours and hours spent in sedentary behaviour. ACAES had been previously evaluated for reliability and relative validity and demonstrated acceptable accuracy for ranking nutrient intakes in Australian youth aged nine to 16 years [^28-30]. Portion sizes for individual food items were accessed from the Australian Bureau of Statistics (ABS) [³¹] and unpublished data from the 1995 Australian National Nutrition Survey or the “natural” serving size for common items such as a slice of bread. Subjects were asked about frequency of their consumption over the previous 6 months. The frequency options ranged from ‘never’ to ‘4 or more times per day’ but up to ‘7 or more glasses per day’ for drinks. Of these frequency response questions, 24 related directly to intake of vegetables or legumes and 11 to fruit with seasonal availability of some fruits considered in the nutrient analysis, nine questions related to breads and cereals, nine to dairy foods, 32 to main meal or lunch items, nine to beverages, 20 to snack foods/dessert and six to sandwich spreads/dressing/ sauce. Nutrient intakes from the FFQ were computed from the most current food composition database of Australian foods available, the Australian AusNut 1999 database (All Foods) Revision 17 primarily and AusFoods (Brands) Revision five (Australian Government Publishing Service, Canberra) [³²] to generate individual mean daily macro-and micronutrient intakes. The estimated daily intake for twenty macro- and micronutrients was calculated using FoodWorks (version 3.02.581 Xyris Software Australia, Highgate Hill, Queensland). In the current analysis FFQs that had greater than five unanswered items were excluded. For all FFQs included missing responses considered to be items that were never consumed.

This study was approved by the University of Newcastle Human Research Ethics Committee (approval number H-949-0205) and the New South Wales (NSW) Department of Education and Training. Written informed consent was obtained from all participants’ parents/guardians and participants gave informed assent prior to inclusion in the study.

The ACARFS was designed as a brief, culture specific, food based diet quality tool for children and adolescents. It focused on dietary variety within recommended food groups for use in populations and individuals and took 10 to 15 minutes to complete. It was modelled on the Recommended Food Score [³³] and the Australian Recommended Food Score (ARFS) [³⁴] which evaluated adherence to National Dietary Guidelines for Adults. The ACARFS utilises a subsample of questions in the ACAES FFQ [²⁸] which are consistent with the eating patterns recommended in the Australian Dietary Guidelines for Children and Adolescents [³⁵]. The ACARFS has eight food group components, seventy questions and a score ranging from zero to 73 (Table 1) with 20 questions related directly to vegetable intake, 12 to fruit, 13 to protein foods (seven to meat and six to non-meat protein foods), 12 to breads/cereals, 10 to dairy foods, one to water and two to spreads/sauce. The scoring for each food group component is detailed in Table 1.The maximum possible score for each component was determined by the number of suitable ACAES FFQ items in each food group with scoring based on how frequently the included items were consumed. Most foods were awarded one point for a reported consumption of ≥ once per week, but differed for some items depending on national dietary guideline recommendations [³⁵] with consideration of the Australian Guide to Healthy Eating [³⁶]. Some food items had a limit placed on their score for higher intakes due to higher intakes being associated with potentially higher saturated fat or disease risk and these capped items received only one point for the following; minced meat consumed > never to < once per month; beef and/or lamb, chicken without crumbing or batter or pork consumed one to four times per week; flavoured milk, ice-cream or frozen yoghurt consumed ≥ once per week to ≤ once per day; cheese, cheese spread or cream cheese consumed ≥ once per week but < four per day; and water if consumed ≥ four glasses per day. A scoring cap was applied to meats as only 0.5 to 1 serve of meat (including fish and poultry) is recommended per day in national guidelines for children and adolescents [³⁵,³⁶]. Some dairy foods including flavoured milk, ice-cream, frozen yoghurt and cheese were capped due to their high saturated fat and/or sugar content [³⁵,³⁷]. Additional points were awarded (Table 1) for consuming evening meals with vegetables ≥ three times per week, ≥ one pieces of fruit per day, ≥ two serves of milk, yoghurt or cheese per day, and for using reduced fat milk and wholegrain or wholemeal bread. The type of fat spreads was not asked as children are unlikely to know this. Yeast extract spread and tomato ketchup were included in the score as they contain a significant amount of B-group vitamins or β-carotene respectively [³⁷].

Health-related variables potentially associated with food intake and adjusted for in the multiple regression were gender, hours spent in sedentary behaviour, BMI z-score [³⁸] and age. Meeting sedentary behaviour recommendations (<0–1 h spent watching television and <0–1 h spent at the computer or playing video games), height, weight, BMI category, gender, school year, and Socio-economic status (SES) were also described. SES was assessed using the Socio-Economic Index for Areas (SEIFA) [³⁹] at the level of the school, based on school postcode, but was not included in the regression as it was not available at the individual level.

Health-related variables were described for the whole group and by school year. Correlation and multiple regression analysis were used to test for significant associations with the potential confounders: age, BMI z-scores and hours spent in sedentary activity. Descriptive statistics were used to summarize the ACARFS for each school year and by gender and also to describe the contribution of each of the core food component scores to the total ACARFS. Normality of data was tested using the Shapiro-Wilk test, the χ² test was used for testing association between categorical variables, the Two Sample t-tests (two-tailed) for parametric continuous variables, and the Wilcoxon test (two-tailed) for non-parametric continuous variables.

The weighted kappa (κ) statistic, a quantitative measure of agreement where chance is accounted for, was used to assess agreement between ACARFS quartiles with quartile distributions of the average daily intake of 13 nutrients: percent energy from saturated fatty acids (SFA), fibre, vitamin C, vitamin A, β-carotene, thiamin, riboflavin, niacin, folate, calcium, magnesium, iron and zinc. Quartile four represented the highest ACARFS score or highest intake of each nutrient except for percent energy from (saturated fatty acid) SFA in which quartile four represented the lowest intake. The macronutrients of total fat, mono-unsaturated fatty acids (MUFA), poly-unsaturated fatty acids (PUFA), protein and carbohydrate were not assessed using the weighted κ statistic as there is no definite recommendation for increasing or decreasing their intake, for example the recommended fat intake is 30% whereas SFA intake is less than 10% [⁴⁰]. The weighted κ statistic was chosen to compare ACARFS scores with nutrient intakes as it summarizes the agreement in categories of ranked ACARFS scores in reference to each nutrient assessed from the FFQ. This provided information to assess the usefulness of the ACARFS in individuals as well as populations. In addition, the non-weighted κ statistic was calculated for each ACARFS quartile and its corresponding quartile of nutrient intake to further investigate their agreement. Kappa values of 1.0 represent perfect agreement, 0.0 represents no more agreement than would occur due to chance and negative values suggest there is less agreement than would occur due to chance alone [⁴¹]. Arbitrary benchmarks for κ statistics have been set as 0.01 – 0.20 ‘slight’ agreement, 0.21 – 0.40 ‘fair’, 0.41–0.60 ‘moderate’, 0.61–0.80 ‘substantial’ and 0.81–0.99 ‘almost perfect’ agreement [⁴²].

However, the κ statistic may be an overly conservative assessment of agreement as the degree of chance is thought to be overestimated [⁴³]. To further explore the relationship between the ACARFS and nutrient intakes determined from the ACAES FFQ, Spearman’s correlation coefficients (as the nutrient intakes were skewed to the left) were calculated for the 13 nutrients listed previously as well as energy and percentage of energy from protein, carbohydrate, fat, SFA, MUFAs, PUFAs and sugar. Descriptive statistics were used to evaluate nutrient intakes by ACARFS quartile and in comparison to the Nutrient Reference Values (NRVs) of Recommended Dietary Intakes (RDIs) or Adequate Intakes (AI, used when the RDI cannot be determined), where appropriate, for children aged nine to 13 years [⁴⁰]. The RDI was chosen as the benchmark for comparison as it is the average daily intake level sufficient to meet the needs of 97 to 98% of the population, a more ambitious target than the Estimated Average Requirement (EAR) which is the level required to meet the needs of only 50% [⁴⁰]. The RDI was used to indicate the ACARFSs association with nutrient intakes.

For the analysis, a p-value of < 0.05 was considered statistically significant. Statistical analysis was undertaken using STATA version 8, StataCorp LP, College Station, Texas (2003) except for the unweighted kappa statistics which were evaluated using JMP v 8, SAS Institute Inc, Cary, North Carolina (2009) as this function was unavailable in STATA version 8.

The demographic characteristics of participants (n 720) are reported in Table 2. Boys had lower mean BMI z-score compared to girls (0.57 ± 1.16 versus 0.75 ± 1.18, P = 0.040). Twenty-nine (four percent) of participants were found to have five or more unanswered ACAES FFQ items and were therefore excluded. These excluded participants (n 29) were not significantly different (P > 0.05) from the remaining participants (n 691) in any of the variables considered in Table 2.

The ACARFS was calculated for 691 children and was slightly skewed to the left. Table 3 reports the ACARFS descriptive statistics overall, by gender and school year. From a possible maximum score of 73, the median ACARFS score was 25 with a maximum of 58 and a minimum of three. Females and older primary school students (mean age ≈ 12 years) had a significantly higher mean ACARFS than males and younger children (mean age ≈ 10 years) respectively (P < 0.001, P = 0.005). Table 4 reports the contribution of each ACARFS component to the overall score.

The correlation between BMI z-score and the ACARFS was not significant (r = 0.02, P = 0.610). Multiple linear regression showed that together gender, BMI z-score, age and hours/day spent in sedentary pursuits explained a small amount of the variation in the ACARFS (R² = 0.04, P < 0.001). Individual regression coefficients were gender = 0.1, BMI z-score = 0.03, age = 0.10 and sedentary behaviour = 0.12. Twenty-six percent of the participants met the recommendations for minimizing sedentary behaviour with girls more likely than boys (P <0.001) but the difference by school year was not significant (P > 0.05). Thirty-four percent of children reported that they spent ≥ 4 h/d in sedentary activity, and five percent reported ≥ 8 h/d. The SEIFA codes for all schools were below the NSW average for the Socio-Economic Index for Areas of disadvantage [²⁸,³⁹].

The ACARFS quartiles were assessed for agreement with quartiles of nutrient intakes (Table 5). The percentage of ACARFS scores and nutrient intakes classified into the same quartile, the percentage classified in the same or adjacent quartile, and the percentage grossly misclassified (i.e. those classified as quartile four for ACARFS but classified as quartile one for nutrients and vice-versa) were calculated. The κ statistic for each quartile was calculated as was the overall weighted κ statistic, standard error and p-value. The percent energy intake from SFA gave the least overall agreement of all the nutrients (κ = 0.13) and demonstrated ‘slight’ agreement, followed by riboflavin (κ = 0.36) which showed ‘fair’ agreement. Vitamin C (κ = 0.64), fibre (κ = 0.62) and β-carotene (κ = 0.62) had the strongest ‘substantial’ agreement. All other nutrients showed ‘moderate’ agreement (κ = 0.42 – 0.56). Within quartiles, fibre, vitamin C and β-carotene had the lowest percentages grossly misclassified. With the exception of SFA, all other nutrients had less than five percent grossly misclassified. The strongest agreement amongst the quartiles was quartile one, where nine of the nutrients showed ‘moderate’ agreement. Quartile four showed the next strongest agreement with agreement in quartiles two and three rated as very slight.

ACARFS demonstrated statistically significant positive correlations with all vitamins and minerals tested (Table 5). The strongest correlations were with vitamin C (r = 0.70, P < 0.001), β-carotene (r = 0.67, P < 0.001) and fibre (r = 0.67, P < 0.001). ACARFS also had a moderately strong positive correlation with total energy (r = 0.51, P < 0.001). When the ACARFS was correlated with macronutrients adjusted for energy intake there was a weak positive correlation with protein (r = 0.18, P < 0.001) and weak negative correlation with total fat (r = −0.12, P = 0.003) and SFA (r = −0.15, P < 0.001). Associations between the ACARFS and percent energy intake from MUFA, PUFA, carbohydrate and sugar intake were not significant.

Table 6 describes the NRVs appropriate for individual children aged nine to 12 for the nutrients considered in the analysis, and the median and IQR of nutrient intakes as calculated by the ACAES FFQ for each of the ACARFS quartiles. In both quartile three and quartile four (highest ACARFS score) all of the median nutrient intakes met the corresponding NRV. In both quartile one (lowest ACARFS quartile) and quartile two the median nutrient intakes of the population sample for fibre, folate and calcium did not meet the corresponding NRV. Table 6 also shows the proportion of the sample population not meeting the corresponding NRV. This was the greatest for fibre (45%), folate (45%) and calcium (40% <1000 mg, 62% <1300 mg) and the lowest for niacin equivalents (1%) and riboflavin (0.3%).

The sample population were aged nine to 12 years only and had a lower SES than the NSW average which may reduce how generalisable it is to other populations. While parents can fill in the ACARFS on behalf of their child, this may introduce bias as parents have been reported to overestimate child diet quality [⁴⁸]. The relative contribution of each component to the final score was dependent on the questions in the ACAES FFQ and is not necessarily representative of the Australian Guide to Healthy Eating [³⁶]. However, this may be viewed as a strength and a more realistic representation of the food group proportions available in the food supply. As the dietary guidelines for children in Australia do not provide specific recommendations for amounts to be consumed within food groups, the scoring contains an additional degree of subjectivity. This potentially means the ACARFS could overestimate usual diet quality. Further, given that biomarkers to objectively verify components of dietary intake were not measured, the results should be interpreted with caution.

In the ACAES validation study the FFQ demonstrated higher nutrient intakes compared to food records [²⁸]which may explain why the median intakes of niacin, vitamin A and magnesium were above the corresponding upper limit. However, the ACAES FFQ validation study demonstrated the ability of the ACAES FFQ to correctly classify participants into quintiles of nutrient intake and therefore not affect the assessment of agreement and correlations [²⁸]. This also suggests that participants in the first and second ACARFS quartiles may be at risk of inadequate intakes of nutrients other than fibre, folate and calcium.

As the ACARFS is derived from a validated FFQ for children it offers the opportunity for researchers to use it independently or to derive it secondarily from the FFQ as a measure of overall dietary quality as a single continuous variable. The calculation of the ACARFS from the full ACAES FFQ is less onerous than indices that include nutrient based sub-scales. Its use as a brief tool to assess diet quality using only the FFQ questions and relevant responses could extend its use by allowing the ACARFS to be used along with the provision of timely feedback.

To extend its usability further research should examine use of the ACARFS method applied to other FFQs, in other populations, age groups, as well as other settings such as a self-monitoring tool or within clinical practice. In order for the ACARFS to be of use clinically or for self-assessment, then cut points may need to be derived. However, the agreement between quartiles suggests that those with an ACARFS score of 32 and above have a good diet quality and consume a reasonably wide variety of nutritious foods and that they have the highest nutrient intakes. Those with an ACARFS score of 19 to 31 (quartiles two and three) have a moderate diet quality, and consume a moderate variety of nutritious foods, but are at risk of sub-optimal intakes of fibre, folate and calcium. Finally, those with an ACARFS score of 18 or less have a poorer diet quality, do not consume a wide variety of nutritious foods and have the lowest intakes of a range of nutrients.

‡Although the serves of vegetables consumed each day is of interest, the question was not available on the ACAES FFQ.

cm, centimetres; m, meters; SD, standard deviation. ^a Healthy weight, overweight or obese classified using the UK age and gender-specific BMI z-score cut-off points that correspond to a BMI of 25 for overweight and 30 for obese at 18 years of age. * P-values for the comparison of excluded participants and remaining participants using the students t-test (two-tailed) were not significant (P > 0.05). †P-values for the comparison of excluded participants and remaining participants using the Wilcoxon test (two-tailed) were not significant (P > 0.05).

IQR, inter-quartile range; Min, minimum; Max, maximum.

ACARFS, Australian Child and Adolescent Recommended Food Score; Max, maximum; Min, minimum.

Survey (ACAES) versus Australian Child and Adolescent Recommended Food Score (ACARFS). ACAES, Australian Child and Adolescent Eating Survey; ACARFS, Australian Child and Adolescent Recommended Food Score; κ, kappa statistic; SFA, saturated fatty acids. ‡Quartile 4 indicates the highest ACARFS (32–58) and nutrient intakes, quartile 3 indicates the second highest ACARFS (26–31) and nutrient intakes, quartile 2 indicates the second lowest ACARFS (19–25) and nutrient intakes, quartile 1 indicates the lowest ACARFS (3–18) and nutrient intakes. § Landis and Koch Classification [⁴²]. ∥Standard Error: 0.04 *P < 0.0005 **P < 0.001 *** Significantly different from 0, P < 0.001.

NRV, Nutrient Reference Value; ACAES, Australian Child and Adolescent Eating Survey; RDI, Recommended Dietary Intake; AI, Adequate Intake; IQR, inter-quartile range; eq, equivalent. ‡ The median intake of niacin is above the upper limit of 20 mg/day [⁴⁰]. § The median intake of vitamin A is above the upper limit of 1700 μg/day if the total vitamin A source was from retinol, however the vitamin A (retinol equivalents) estimated intake in the sample population includes β-carotene which is not known to result toxicity [⁴⁰]. ∥ The median intake of magnesium is above the upper limit of 350 mg/day [⁴⁰]. A median calcium intake of 1270 mg does not meet the RDI for children aged 12 to 13 years and children aged nine to 11 years who are growing at a greater rate than average. The RDI for these groups is 1300 mg [⁴⁰].