This is the first study in NZ that has used a diet index to assess diet quality in adolescents, and one of the first to have simultaneously addressed the relative and construct validity of an FQ-derived DQI by utilising different dietary assessment methods. Unlike the majority of studies which have used the same instrument for constructing and validating a diet index, the present study has the advantage of having examined the validity of the NZDQI-A using an independent measure of nutrient intakes, i.e. 4DFR data.
Comparisons with other studies
Although an increasing number of studies have employed diet indices to describe diet quality in children and adolescents [12–14, 16–20, 22–24, 28, 51, 54–56], none have been conducted in NZ. Six studies have reported the validity of a diet index against an independent reference method among populations which included adolescents aged 13 years and above [17, 20, 23, 28, 51, 54]. Among these six studies, three studies validated their indices using nutritional biomarkers [20, 23, 54] including the large-scale European HELENA study ; whilst the other three studies validated against diet records  and 24-hour diet recalls [17, 28].
The first study by Torheim et al. compared the validity of the Food Variety Score and Diet Diversity Score calculated from two FFQs (69 and 164 food items) to those from 2-day weighed records. Correlation coefficients between the two methods ranged from 0.2 to 0.5 for the two indices of diet variety. In another study by Schroder et al., dietary intakes of 11 selected nutrients reported by multiple 24-hour recalls were found to be positively associated with the three diet quality indices (Diet Quality Index, modified Mediterranean Diet Score and Antioxidant Score) derived from two short questionnaires containing 15 to 18 food items. The validity correlation coefficients between the two methods ranged from 0.32 to 0.45. In these two studies, validation work was completed in relatively large samples with wide age ranges (n = 145 aged 15 to 59 years and n = 102 aged 3 to 80 years, respectively). As results were not reported separately for adolescents, a direct comparison with these studies was not possible. In spite of this, we found that our NZDQI-A calculated from a FQ produced similar if not better correlations plus reasonable agreement in ranking scores compared to a 4DFR. In the validation study by Serra-Majem et al. among 3166 children and adolescents (aged 6 to 24 years), an increased mean intake for a majority of vitamins and minerals calculated from 24-hour diet recalls was found with increasing Mediterrannean diet adherence score using the KIDMED. Employing a similar analytical approach in our study, the significant trends towards more optimal intakes of selected nutrients are also suggestive of the NZDQI-A’s construct validity.
Following expert recommendations [5, 7], the current NZDQI-A was constructed based on five major food groups to reflect the fundamental premise of the NZ food-based dietary guidelines, which emphasised that nutrient needs should be met primarily through ‘eating different kind of foods each day’ . To address the equal importance of having variety in addition to adequacy in diets, a total score was calculated by multiplying both construct elements for the five equally weighted index components.
The methodology used for constructing the NZDQI-A may have implications for validity estimates in this study. First, as this index was constructed within the boundaries of the FQ data, the limited number of questions in the questionnaire may have restricted the variation in food variety and serving intake responses. In particular, our NZAFFQ data does not distinguish between wholegrain and refined grains and between lean and fatty meats; hence the index scoring was not specific to carbohydrate and fat quality. In any case, participants who scored higher in this study seemed to consume less fat and better fat quality, but no relationship was seen for dietary fibre. To improve the ability of the NZDQI-A to detect fibre intake, a possible index modification may include a component that gives merit to a higher intake of wholegrain bread relative to white bread, given that breads are reported to be the main dietary fibre source for NZ adolescents .
Secondly, the current NZDQI-A components were selected to reflect positive food choices rather than negative ones. Therefore in measuring favourable food group intakes, the index discounts excess intake of foods high in fat, sugar and salt. This may then mean that it is possible that a higher quality index score is reflective of a higher energy intake. However, results from this study suggest that this is not the case. Some authors have used a deduction approach to demerit foods considered detrimental such as sausages, pastries, confectionery, soft drinks and fast foods in their composite index [23, 28]. Nevertheless, unless population-specific evidence-based recommendations for these discretionary foods are available, the decision on what is considered ‘an acceptable intake level’ of such foods remains subjective.
With regard to scoring of the index, we used trichotomous cut-off points (0 for non-consumption, 10 for intakes below recommendations and 20 for intakes in line with recommendations) based on the suggested minimum servings of the dietary guidelines . We are conscious that this scoring approach has disadvantages. Mainly, the discriminating power of the NZDQI-A may potentially be reduced when most participants have low intakes of certain food groups , especially when the sample size is small. For instance, only 22% of participants achieved the suggested daily intake of milk and dairy products in this study. The low intakes of this food group resulted in more than 80% of participants scoring 10 and below for the DAIRY component. The right-skewed distribution of this component score may have diluted the resultant total score and therefore attenuated the association between the NZDQI-A and calcium intakes in the sample.
Lastly, the validity of a diet index is reflective of the dietary guidelines upon which it is based [50, 58]. When possible, quantitative criteria were used in establishing the cut-off points for scoring the NZDQI-A. This was however not feasible for all foods, as some intake recommendations were not quantified in the NZFNG . An example of a non-explicit recommendation is ‘choose food low in fat, sugar and salt’. An age-specific food-based dietary guideline that is formulated in quantitative terms  will facilitate the interpretation of dietary guidelines more objectively and reduce the various arbitrary choices involved in the construct of an index.
Although some authors suggested that the validity of a diet index should be compared to nutritional biomarkers , the use of this ‘gold standard’ is often prohibitive due to its invasive nature and expense. We chose a 4DFR as our reference method as this prospective method is not memory-dependent and hence has less correlated errors with an FFQ compared to a 24-hour diet recall . The main limitation of estimated records is the higher participant burden due to multiple-day recording which may discourage completion [60, 61]. Despite careful preparation of the food record as an easy-to-carry booklet and provision of portion aids to facilitate accurate recordings, we acknowledge that misreporting may still occur given the limited motivation and possible poor portion size estimation among adolescents [62, 63].
The main limitation of this study was the small sample size (n = 41). The findings of this study must be interpreted with caution as there is a possibility of type one errors due to multiple testing in a small sample. On the other hand, the relatively narrow range of NZDQI-A scores attributed by the low variation in food intakes with the small sample may have restricted the ability to detect true correlations between the two methods for some nutrients. To eliminate learning effects from food recording , data from the first administration of the FQ was used to compare with the 4DFR. The reference period of the first FQ (i.e. past seven days or 4 weeks) spanned differently from the 4DFR  and may have led to underestimation of the relative validity of NZDQI-A. Nevertheless, the present study yielded some positive findings that suggested that the NZDQI-A is valid as an indicator of diet quality.
The major advantage of the NZDQI-A lies in its simplicity and practicality, as neither nutrient quantification nor food composition data are required for its score derivation. Based on summary questions and frequency questions from the FQ, diet quality was assessed based on intakes of variety and servings of foods recommended for adolescents. Further to extending the good use of dietary information from a brief dietary assessment tool, this index may be applied to assess diet quality in studies of a broad range of adolescent populations, including those where study resources are limited.