Background

The prevalence of non-communicable diseases (NCDs) including type 2 diabetes mellitus (T2DM), cardiovascular disease (CVD), and chronic respiratory disease experienced by children and adolescents aged 0 to 18-years is increasing [1, 2]. Four hundred new cases of T2DM are diagnosed annually in Australians aged 10–24-years [3]. Hypertension, a risk factor of CVD, is present in 6–7% of children and adolescents in Australia, the United Kingdom, and the United States of America (USA) [4,5,6]. Of concern, NCDs adversely affect growth, development, and maturation in childhood and adolescence [7], leading to compromised adult health and reduced life expectancy [8]. Hence, the prevention of NCDs in childhood is a global priority, requiring a multi-pronged approach to address major NCD risk factors [9]. These risk factors include diet quality, healthcare access, and substance abuse, which affect physical growth and mental development [10], with poor diet quality identified as one of the largest contributors to the global burden of NCDs [11].

Diet quality is broadly defined as a dietary pattern or an indicator of variety across key food groups relative to those recommended in dietary guidelines [12]. High diet quality thereby reflects achieving more optimal nutrient intake profiles and a lower risk of diet-related NCDs [13]. Diet quality can be influenced by confounding factors, including cultural and food environment, socio-economic status, child and family food preferences, and nutrition recommendations relevant to age, sex, country, and/or culture of the individual [14]. Diet Quality Indices (DQIs) are assessment tools that can be used to quantify the overall quality of an individual’s dietary intake by scoring food and/or nutrient intakes, and sometimes lifestyle factors, according to how closely they align with dietary guidelines [12]. There are a variety of DQIs which utilise a range of scoring matrices. Some use frequency of food or food group consumption, others use nutrient intakes which require estimation prior to scoring, and some include both.

Due to the link between dietary intake in childhood and NCDs in both childhood and adulthood, the accurate measurement of paediatric diet quality is essential both to understand current intakes as well as evaluate the effect of interventions [15, 16]. Reflecting this need, the use of DQIs is increasing not only in research and epidemiology, but also in community health and clinical settings where DQIs may form part of dietary education and self-monitoring interventions [14, 17,18,19,20]. A systematic review of paediatric DQIs which included papers published up until October 2013 identified 80 individual DQIs used in paediatric population samples, some of which identified cross-sectional associations with growth and health outcomes such as body weight, early onset puberty, and blood pressure [14].

Given the increasing number of DQIs identified in the previous review used or created for research, the diversity in the tools, and the different settings, age groups, and countries they are used amongst, there is a need to update the previous systematic review to identify valid DQIs and their associations with health outcomes [14]. Therefore, the aims of this systematic review update are to; 1) summarise a-priori DQIs used in child and adolescents; 2) appraise the validity and reliability of paediatric diet quality indices; and 3) synthesise the evidence on the relationship between diet quality and physical health, mental health, and growth-related outcomes among paediatric samples.

Methods

Study design

A systematic literature review was conducted and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [21] and registered prospectively with the International Prospective Register of Systematic Reviews (PROSPERO number: CRD42018107630).

Search strategy

The search was designed as an update of the 2014 systematic review [14]. Medline (PubMed) and CINAHL were searched from 31 October 2013 to 11 January 2019. To broaden the search, the current review also searched Embase, Web of Science, and CENTRAL from database inception to 11 January 2019. The strategy used both controlled-vocabulary and keywords, and was designed for PubMed and translated for use in other databases using Polyglot Search Translator [22]. The translated search strategies were checked for accuracy by a librarian, and two authors (PD and SM), then further adapted for each database after examination of sensitivity and specificity by using a target of one eligible study per 100 records retrieved, with an estimated 150 eligible studies (Appendix). To support the systematic search update, snowball searching of reference lists of identified papers was conducted and the previous review [14] was examined to include any eligible studies the current search strategy didn’t identify.

Eligibility criteria

Table 1 describes the eligibility criteria used to identify studies to answer the research questions; a study was included if it addressed one or more of the research questions. Studies published in English and Mandarin (translated to English by colleagues) were included. Studies published in other languages were included if they could be translated using Google translate [23]. For this review, a DQI was defined as any assessment tool which applied a quantitative score to food (i.e., frequency of consumption) or nutrient intake, where the scoring system reflected pre-defined national dietary or nutrient guideline/s (i.e., the DQI scoring system was developed a-priori). Diversity and variety indices that score or count the variety of foods consumed without regards to a dietary standard were excluded. Excluded lifestyle indices were any scoring system which had ≥2 scoring components on behaviours such as exercise, sedentary activities, or smoking.

Table 1 Eligibility criteria of original studies included in this review according to the population, indicator, comparator, outcomes, and study design (PICOS) format.
Full size table

Study selection and data extraction

Identified records were de-duplicated using Systematic Review Assistant-Deduplication [24] followed by a manual search in Endnote [25]. Titles and abstracts of papers were screened independently to assess their potential eligibility by two researchers (PD and SM) using Covidence [26], which further removed duplicates. The full texts of potentially eligible records were acquired and screened for eligibility by two researchers independently (PD and SM), with disagreements managed by consensus. Data were extracted from included papers by one researcher (PD) into three standardised tables; with random quality checks by a second researcher SM). For studies which measured prospective health-related outcomes, data were reported in their standard international units at baseline and follow-up, as well as mean change over time where possible.

Health-related outcomes

Any prospective outcome related to physical health, mental health, or growth was included if the variable was reported relative to DQI score or categories. Health-related outcomes used to describe the sample, but not linked to a DQI score were not considered. Health-related outcomes in adults were considered if they were related to a DQI assessment when the sample was aged < 18 years. In order to assess the ability of the DQI to predict health-related outcomes, outcomes were considered from 1-week after the DQI assessment with no further restriction on timeframe of follow-up. Health-related outcomes reported as the result of an intervention study were not considered as outcomes are likely to reflect the intervention rather than baseline diet quality.

Study quality

Any study which reported on the validity of a paediatric DQI or health-related outcomes was critically appraised using The Academy of Nutrition and Dietetics Quality Criteria Checklist (QCC) [27], independently by two authors (PD, SM, TB, or CC). Studies which reported the use of a DQI but didn’t report validity, reliability, or health-related outcomes were not critically appraised as study quality was not relevant to research question 1. Any disagreements in study quality were settled by consensus. The Academy QCC is a critical appraisal tool suitable to evaluate the risk of bias for any study design, including diagnostic, intervention, or observational. The QCC rates the quality of the study as positive, negative, or neutral reflecting risk of bias in participant selection, generalisability, data collection, and analysis [27]. Studies found to have negative study quality were not excluded.

Results

Of 15,577 records identified in the search, 4896 were duplicates. After title and abstract screening, 312 full texts were assessed against the eligibility criteria, with 132 papers included, including 22 identified through snowball searching (Fig. 1). The main reasons for exclusion were use of a non-a-priori diversity or variety index (n = 127), study design (n = 48), or study outcomes (n = 48). From the 132 included studies, 81 diet quality indices were identified by the current search strategy in addition to those identified in the original systematic review [14]. Of the 80 indices described in the original review [14], 47 were eligible in the current review update and were primarily identified from the current search strategy but was supported by the snowball search (Fig. 1), leading to a combined total of 128 unique indices designed for and/or used among children and adolescents. Of these, 39 included papers had evaluated the validity and/or reliability of 37 DQIs, while 12 evaluated the association of 12 DQIs with prospective health outcomes.

Fig. 1
figure 1

PRISMA flow diagram demonstrating selection of studies.

Full size image

Characteristics of diet quality indices developed for or used in paediatric samples

The 128 DQIs were developed across 33 countries, with most being designed for the USA (n = 23), Australia (n = 16), Germany (n = 11), and Brazil (n = 8) (Table 2). There were 23 DQIs created outside of the USA such as Australia, Belgium, Canada, and Gaza with scoring methods based on the Dietary Guidelines for Americans (Table 2). Very few indices were identified in developing countries (n = 7) [262]. Those identified were from India, Indonesia, and Guatemala [134, 138, 141] and were typically brief tools more appropriate for field work, assessing frequency of consumption or dietary patterns and used dietary guidelines from other countries such as the USA to assess diet quality [134, 138, 141]. Thirteen (10%) DQIs were adaptations of the Diet Quality Index (DQI) [250], and 22 (17%) were adaptations of the Health Eating Index (HEI) [227]. These adaptions reflected changes to the scoring system to be more applicable to different countries or age groups. Four identified DQIs were designed for adults and subsequently used among children and adolescents without being adapted [89, 106, 127, 250].

Table 2 Description and purpose of diet quality indices which have been designed for use or used in paediatric populations presented alphabetically by country (n = 128 indices)
Full size table

Most indices were scored by considering both food and nutrient intakes (n = 64 DQIs), while 34% (n = 44 DQIs) scored by considering food intake alone, and 6% (n = 7 DQIs [111, 114, 115, 117, 146, 148, 189]) scored using nutrient intake data alone (Table 2). In addition, 10% (n = 13 DQIs [46, 73, 75, 80, 121, 128, 165, 195, 214, 224, 235, 248, 258]) assessed a single behaviour (e.g. physical activity levels) as well as food and/or nutrient intake. The most common methods of collecting dietary data in studies which reported the development of DQIs were 24-h dietary recalls (n = 44) and food frequency questionnaires (FFQ) (n = 43); while some studies used both methods (n = 18), others used alternative methods such as study specific questionnaires or multiple day food diaries or records (n = 23) (Table 2).

A number of studies utilised information from the same datasets, such as data from the National Health and Nutrition Examination Survey (NHANES) prospective population surveillance in the USA, or the Healthy Lifestyle by Nutrition in Adolescence (HELENA) in Europe [263, 264].

The quality and strength of papers identified

Of the 39 papers assessing validity and/or reliability of 37 DQIs, 22 papers had positive study quality, while 17 papers had neutral study quality (Table 3). Of the papers assessing the relationship with health-related outcomes, 10 papers had positive study quality and two papers had neutral study quality (Table 4). None of papers evaluated had a negative study quality. The most prevalent reasons for papers to be downgraded to neutral study quality was due to authors not reporting the eligibility criteria of participants, sampling method, or reasons for attrition.

Table 3 Studies evaluating the validity and/or reliability of paediatric a-priori diet quality indices (n = 37).
Full size table
Table 4 Association of diet quality indices with prospective health-related outcomes in paediatric populations (n = 12).
Full size table

The validation of diet quality indices

Only 28% (n = 37) of the DQIs identified were evaluated for validity (n = 35) and/or reliability (n = 11) (Table 3). Validity was assessed by construct validity (n = 21), concurrent or convergent validity (n = 8), relative validity (n = 8), content validity (n = 4), predictive validity (n = 4), or comparative validity (n = 1), and eight DQIs were assessed for more than one type of validity [46, 107, 152, 177, 192, 240, 268, 272]. Reference standards used to evaluate the validity of indices were other validated tools, serum biomarkers (n = 9) [45, 50, 74, 80, 88, 130, 194, 265, 266], food intake (n = 18) [19, 33, 45, 69, 73, 74, 80, 87, 88, 107, 121, 130, 152, 172, 175, 177, 178, 265, 266], nutrient intake (n = 30) [19, 33, 43, 45, 46, 53, 69, 73, 74, 80, 87, 88, 104, 107, 121, 130, 151, 152, 172, 175, 177, 178, 189, 192, 194, 240, 265, 266, 270,271,272] and energy intake (n = 9) [43, 69, 73, 104, 107, 172, 194, 200, 240, 272]. Cross-sectional health markers including blood pressure (n = 1) [80], weight (n = 3) [45, 87, 122], BMI (n = 11) [33, 45, 73, 80, 87, 121, 122, 200, 240, 270, 271], and waist circumference (n = 1) [271], percent body fat (n = 2) [122, 270] were used to evaluate validity (Table 3). Although assessed, the Modified revised children’s diet quality index (M-RCDQI) [151] and the Revised Brazilian Healthy Eating Index (BHEI-R) [266] were found to require further research to test the validity and reliability of these tools before they could be considered valid or reliable.

Health-related outcomes

Only 12 DQIs were evaluated for association with prospective health outcomes (n = 12 studies). Measured outcomes from these 12 studies included nutrient biomarkers (n = 7) [74, 88, 269, 273], IQ scores (n = 1) [269], blood pressure (n = 2) [269, 273], plasma cholesterol (n = 2) [269, 273], risk of metabolic syndrome (n = 1) [149], mental health (n = 1) [275], pre and post-menopausal breast cancer (n = 1) [244], and timing of puberty (n = 1) [117] (Table 4). Anthropometric values examined included BMI (n = 7) [38, 74, 85, 88, 252, 254, 269], changes in BMI or fat mass (n = 2) [117, 252], changes in weight (n = 1) [74], and body composition at onset of puberty (n = 1) [117].

Significant associations were found between high diet quality and serum vitamin D (β = 0.005, 95% CI = 0·002, 0·008, p < 0.0001), holo-transcobalamin (an indicator of B12) (β = 1·005, 95% CI = 1·002, 1·007, p = 0.0002), n-3 FS status (β = 0·376, 95% CI = 0·105, 0·646, p < 0·007) [88], and serum vitamin A (r = 0.128, p = 0.004) [74]. In adjusted models there were significant positive associations between CFUI score and total IQ (β = 1.92 [1.38, 2.47], p < 0.001), verbal IQ (β = 1.92 [1.37, 2.48], p < 0.001), and performance IQ (β = 1.33 [0.74, 1.92], p < 0.001) [269].

In adjusted models, significant inverse associations were found between diet quality and waist circumference (β = −0.15 [−0.31, − 0.002], p = 0.046), diastolic blood pressure (β = − 0.15 [− 0.31, − 0.002], p < 0.001) [269] and incidence of metabolic syndrome (OR: 0.35, 95%CI = 0.13,0.98, p < 0.05) [149] (Table 4). Significant inverse associations were found between diet quality and HbA1c levels in youth with type 1 diabetes (β = − 0.2, SE = 0.07, p = 0.0063). There was no association between diet quality and HbA1c in youth type 2 diabetes; however, there was a significant association for improved systolic blood pressure (β = −2.02, SE = 0.97, p = 0.0406) [273, 274].

Diet quality was positively associated with mental health-related quality of life [275] (Table 4). Female children and adolescents with the top three quintiles of diet quality and followed into adulthood had decreased risk of premenopausal breast cancer (HR: 0.78, 95%CI = 0.63,0.97; HR 0.86, 95%CI = 0.69,1.07; and HR 0.84, 95%CI = 0.67,1.04 respectively); but no association was found between AHEI score and pre- or postmenopausal breast cancer (Table 4) [244].

In addition to the above; three studies used prospective health outcomes to evaluate the predictive validity of DQIs (Table 3). The CFUI was associated with improved BMI, waist circumference, and blood pressure [269]; the E-KINDEX was associated with improved BMI, total body fat, and waist circumference [271]; and the Diet Quality Score for Preschool Children was associated with improved fat-free mass and fat mass [177].

Discussion

This review summarises 128 unique a-priori DQIs used in children and adolescents internationally; however, only 30% were assessed for validity and reliability, from which two were found to require refinement [151, 266] to achieve suitable accuracy and reliability. Additionally, only 15 DQIs were tested for association with prospective health outcomes; finding associations between high diet quality and improved nutrient status, IQ, body composition, risk of metabolic syndrome, blood pressure, HbA1c, mental-health related quality of life, and premenopausal breast cancer.

This systematic review update identified 81 novel paediatric a-priori DQIs (from 157 publication), a 172% increase over 7 years from the 47 identified in the original systematic review [14]. This steep increase in the development and use of DQIs demonstrates that this approach to assessing diet quality is well-utilised within research in children and adolescents internationally. The USA, Australia, Germany, and Brazil appear to be leading the development of paediatric DQIs, together producing 45% of all paediatric DQIs. Beyond these four countries, the vast majority of other DQIs were from other developed countries, possibly reflecting this review’s eligibility criteria. Dietary assessment in developing countries are often focused on assessing growth in an environment characterised by a high prevalence of undernutrition, and and is assessed using non-a-priori diet diversity indices (DDIs), diet diversity scores (DDSs), and food variety scores (FVSs) [14, 138, 167, 224] of which there were 127 excluded from this review (Fig. 1).

There were significant variations in DQIs methods. Simpler scoring methods awarded and summed points for foods which were or were not consumed over a specific frequency. This simple food-based scoring method reduces burden on both researchers, clinicians, and individual users as they can be easily applied to clinical practice. Food-based DQIs included the KIDMED, DGI-CA and ACARFS [33, 35, 214]. More complex DQI scoring methods involved quantification of nutrient intakes from reported food intakes which then undergoes a further step of calculating nutrient intakes relative to age-specific dietary guidelines or energy intake, which make such scores less applicable to the clinical setting or for individual use [141]. DQIs with complex nutrient-based scoring approaches included the NIS [114] and the NQI [115], with DQIs which used a combination of food and nutrient-based scoring methods being more common, such as the ARFS-P [19] and the DGI [36], which embody the same limitations as nutrient-only scoring methods.

Of concern, only 29% of the 128 unique DQIs identified were evaluated for validity and/or reliability, and only 12% evaluated associations with prospective health outcomes. Of the 35 DQIs which were evaluated for validity, 34 were stated to be validated tools by authors; however, due to inconsistent methodological approaches the validity of the DQIs could not be consistently evaluated. Only five DQIs (5%; DQI-A [88], diet quality score for preschool children [177], CFUI [269], E-KINDEX [75] and HNSP [74]) were both evaluated for validity and found to be positively associated with prospective nutrient biomarkers, blood pressure, IQ, and body composition. This suggests these DQIs are the most rigorous in terms of accuracy, reliability, and relevance to health. While the use of DQIs to measure the diet quality of children and adolescents is a highly utilised assessment method, further research is required to address the current paucity of evaluation studies of currently available tools.

Further, the large number of new yet non-validated paediatric a-priori DQIs suggests new DQIs are developed prior to evaluating existing DQIs, and therefore may have been unnecessary. The use of DQIs which have not been rigorously developed and evaluated may compromise the research in which they were used and lead to inaccurate and/or unreliable results. This is particularly the case for DQIs which were developed specifically to evaluate outcomes of a particular study, where the development of the tool was minimally described and not intended for re-use or replication; therefore, limiting confidence in the study results.

Approximately half of the identified DQIs were modified forms of the DQI or HEI [227, 250]. However, only 16 of these modified DQIs were validated in the new population (e.g. age, culture, country) group, where the remaining studies assumed validity based upon the tool being valid in the original population. Non-validated tools, even if adapted from a valid tool, should be used with caution as the modified DQI may not accurately assess diet quality or be appropriately extrapolated to the diet and cultural context of the new population sample. This is particularly the case for modified DQIs in which the scoring system was still based on national dietary guidelines of the original country (e.g. The USA), and not the new population (e.g. Brazil, Canada) [50, 59, 70]. Similar cautions should apply for DQIs such as the Healthy Diet Indicator and the Alternative Healthy Eating Index used in paediatric populations that were designed for adults as these indices may not accurately assess children and adolescent’s diet quality [89, 245].

A factor that varied between papers was the method of dietary data collection, with some DQIs able to be calculated using a variety of dietary assessment methods such as the Diet Quality Index – International [252]. This variety is a strength as it allows flexibility in the application of DQIs in future research and clinical practice. A 24-h recall was the most frequently used dietary assessment tool; however, it is unclear if the 24-h recalls were repeated over several days to improve its accuracy in reporting usual intake. Although most remaining DQIs used FFQs, a substantial number of papers did not use validated methods to collect dietary data [39]. There should also be a caution for the use of single 24-h recalls in studies with small sample sizes or in clinical practice as this one-off measure does not accurately represent usual dietary intake. Although a DQI may be valid, the method of dietary intake assessment must also be accurate and relevant if results are to be interpreted with confidence.

Limitations and future directions

The present review may be limited by publication bias, particularly in the fields of a null or negative result relating to the validity of DQIs and their association with health-related outcomes; however, publication bias was unable to be assessed as funnel plots were not able to be generated. Although this review reported validity, reliability, and associations with health-related outcomes; it did not evaluate other aspects of assessment tool utility such as sensitivity to change and participant burden nor did it evaluate the validity and reliability of dietary intake assessment methods.

Limitations in the existing literature highlight the need for future research to validate existing paediatric a-priori DQIs and to test their associations with prospective health-related outcomes. This will allow determination of the effect of diet quality during childhood and adolescence on physical health, mental health, and growth which is of increasing importance as the prevalence of diet-related NCDs continues to rise. The application of any DQI should appropriately assess dietary intake using validated methodology and researchers developing new DQIs should ensure that tools reflect indicators of alignment with an appropriate national dietary guideline or nutrient target specific to the culture, country, and age-group of the intended population, and rigorously describe the tools development, scoring method, and validation procedures. Researchers should consider applying existing valid DQIs to their data and undertaking reliability and validity studies in their population groups. For research reporting associations with health-related outcomes, researchers should fully describe the demographic and medical characteristics of the sample, information about dataset used, and transparently detail the results.

Implications for practice

DQIs present an important opportunity to measure the quality of the total diet of individuals and groups. The current review can be used as a resource to assist health professionals in identifying relevant and valid DQIs for their clinical setting. When selecting a DQI, health professionals should consider: i) whether the DQI demonstrated validity and/or reliability, ii) does the DQI reflect a nutritional reference standard which is relevant to the population in which it will be applied, iii) can the DQI be easily calculated in the clinical setting, and finally iv) can the DQI be calculated by a dietary assessment method which can be performed efficiently in the clinical setting? Although it would be ideal to select a DQI which is associated with prospective health outcomes; due to the paucity of research in this area, this is not yet a feasible consideration.

Conclusion

Research examining diet quality among children and adolescents is of increasing interest globally. However, few indices have been evaluated for validity or reliability or examined for a relationship with prospective health outcomes. Rigorously developed DQIs which have been evaluated have shown good validity, reliability, and association with a range of physical and mental health outcomes. Longitudinal studies are needed to determine the ability of diet quality indices to predict optimal growth and diet-related health-related outcomes among children and adolescents.