Identification and Evaluation of Tools Utilised for Measuring Food Provision in Childcare Centres and Primary Schools: A Systematic Review

Background: Children aged 2–11 years spend significant hours per week in early childhood education and care (ECEC) and primary schools. Whilst considered important environments to influence children’s food intake, there is heterogeneity in the tools utilised to assess food provision in these settings. This systematic review aimed to identify and evaluate tools used to measure food provision in ECEC and primary schools. Methods: The Preferred Reporting Items for Systematic Reviews (PRISMA) was followed. Publications (2003–2020) that implemented, validated, or developed measurement tools to assess food provision within ECEC or primary schools were included. Two reviewers extracted and evaluated studies, cross checked by a third reviewer and verified by all authors. The Academy of Nutrition and Dietetics Quality Criteria Checklist (QCC) was used to critically appraise each study. Results: Eighty-two studies were included in the review. Seven measurement tools were identified, namely, Menu review; Observation; Weighed food protocol; Questionnaire/survey; Digital photography; Quick menu audit; and Web-based menu assessment. An evidence-based evaluation was conducted for each tool. Conclusions: The weighed food protocol was found to be the most popular and accurate measurement tool to assess individual-level intake. Future research is recommended to develop and validate a tool to assess service-level food provision.


Introduction
Early childhood provides a unique window of opportunity to influence nutrition and dietary habits, as this is when food preferences and habits are formed, often tracking into adolescence and adulthood [1][2][3] and influencing health outcomes throughout the life course, in particular the risk of developing obesity [4]. Dietary patterns in low-middle income (LMIC) countries indicate that children's dietary intakes do not meet nutrition guidelines, with an overall under-consumption of the core food groups, particularly vegetables and wholegrains, and over-consumption of discretionary foods, defined as processed foods high in fat, sugar and sodium [5][6][7][8]. According to the Global Burden of Disease Study across 195 countries, suboptimal dietary habits, (low intakes of wholegrains, fruit and vegetables and high intakes of sodium fat and sugar) account for more deaths than any other risk factor [9]. In addition, according to the World Health Organisation Global Strategy

Materials and Methods
This review follows the Preferred Reporting Items for Systematic Reviews (PRISMA [38] protocol, which includes a 27-item checklist and a four phase flow chart ensuring rigour and quality ( Figure 1). The protocol for this review is registered with the Interna tional Prospective Register of Systematic Reviews (PROSPERO) (registration CRD42018109719).

Search Strategy
Five health databases (MEDLINE, Scopus, CINAHL, Cochrane Library and Web o Science) were searched for full-text English-language publications published between Jan uary 2003 and 18 November 2020. Search terms were developed in consultation with an experienced university librarian, to inform the development of a PICO (population, inter vention, comparison, outcome) derived framework with related key search terms. Key terms included ("early childhood education and care" OR "childcare" OR "child care" OR "long day care cent*" OR "day care" OR "kindergarten" OR "K-6" OR "pre-school" OR "primary school" OR "elementary school") AND ("measur*" OR "survey" OR "assess*"

Search Strategy
Five health databases (MEDLINE, Scopus, CINAHL, Cochrane Library and Web of Science) were searched for full-text English-language publications published between January 2003 and 18 November 2020. Search terms were developed in consultation with an experienced university librarian, to inform the development of a PICO (population, intervention, comparison, outcome) derived framework with related key search terms. Key terms included ("early childhood education and care" OR "childcare" OR "child care" OR "long day care cent*" OR "day care" OR "kindergarten" OR "K-6" OR "pre-school" OR "primary school" OR "elementary school") AND ("measur*" OR "survey" OR "assess*" OR "evaluat*" OR "tool*") AND ("food" OR "menu") AND ("food provision" OR "food service" OR "food waste" OR "plate waste" OR "wastage" OR "menu compliance" OR "nutrition guidelines" OR "nutrition policy").

Selection Criteria
The search strategy for this review aimed to identify publications that implemented, validated and/or developed measurement tools that were used to assess menu compliance, food provision and food wastage at a service level within ECEC [defined as long day care (LDC), kindergarten and preschool services where food is provided onsite) and primary school (where food is provided through a food service or school canteens) settings. No limitations were applied to study design, and articles were eligible if they were published in English-language. Studies were excluded if they were not based in ECEC or primary school settings, did not measure menu compliance, food provision and/or food wastage at a service level, or simply measured individual dietary intakes or eating behaviours of children.

Study Selection
The reviewers used Covidence, a Cochrane-developed software, to search and select relevant studies for both the title and abstract screening and full-text screening for inclusion. Following the initial search, two reviewers (Y.G., M.V.) independently screened the eligibility of titles and abstracts against the established inclusion/exclusion criteria, with further eligibility screening using full text. After each independent assessment, the two reviewers discussed individual perspectives for each study, and if no consensus could be made, both a third and fourth reviewer independently screened the relevant study to resolve the decision (R.J., S.K.). The first author (A.E.) reviewed all included studies and cross-checked interpretation of the findings. All procedures associated with study selection were completed in Covidence.

Data Extraction
Two reviewers (Y.G., M.V.) extracted information from the included studies into a study-developed data extraction table, cross checked by the first author (A.E.) and verified by all authors (Tables 1 and 2). Data extracted from each study included: (1): author, (2): year of study, (3): study design, (4): country of origin, (5): setting, (6): study aims, (7) measurement method, (8) limitations and strengths, and (9) quality appraisal. Where studies measured both food provision and intake, only the method utilised for food provision was reported. The Academy of Nutrition and Dietetics Quality Criteria Checklist (QCC) was used to critically appraise each study [39], as our review included multiple types of study designs and this tool comprehensively critiques original research across all types of study designs. The validity questions within the QCC focus on study research questions, participant selection protocol, comparability between study groups where applicable, handling of withdrawals, blinding of groups, where applicable, details on the intervention protocol, validity and reliability of measurement outcomes, data analytical approach, justification of study findings and biases. The outcome of this appraisal was either positive (+) (the report clearly addressed issues of inclusion/exclusion, bias, generalisability, analysis and data collection), neutral (Ø) (the report was neither exceptionally weak or exceptionally strong) or negative (−) (issues were not adequately addressed) [39]. Two authors (Y.G., M.V.) independently appraised each study, verified by a third and fourth author (R.J., A.E.) (Supplementary File S1). Cohen's kappa (κ) was calculated to assess inter-rater agreement, with values interpreted as: <0 no agreement, 0.00-0.20 slight, 0.21-0.40 fair, 0.41-0.60 moderate, 0.61-0.80 substantial and 0.81-1.00 perfect agreement [40]. Cohen's kappa result was within substantial agreement (κ = 0.647; 95% CI: 0.558-0.736).

Search Results
Database searching identified 1687 studies, with 963 studies remaining after removal of duplicates. After abstract screening, a total of 129 articles were identified for fulltext screening with 47 articles excluded because they (a) did not measure food provision and/or compliance at a service level (n = 40), (b) were published in a language other than English (n = 4), (c) were not based within ECEC and/or primary school settings (n = 2) or (d) were other review articles (n = 1). In total, 82 articles were included in this systematic review ( Figure 1).
Forty-seven studies were conducted within the ECEC setting and 35 studies in the primary school setting. Across both settings, seven methods for measuring food provision were identified, namely: (1) Menu review; (2) Visual observation; (3) Weighed food protocol; (4) Questionnaire/survey; (5) Digital photography (6) Quick menu audit and (7) Web-based menu assessment. Figure 2 outlines the number of studies utilising each measurement method in each setting. d) were other review articles (n = 1). In total, 82 articles were included in this systematic review ( Figure 1).
Forty-seven studies were conducted within the ECEC setting and 35 studies in the primary school setting. Across both settings, seven methods for measuring food provision were identified, namely: (1) Menu review; (2) Visual observation; (3) Weighed food protocol; (4) Questionnaire/survey; (5) Digital photography (6) Quick menu audit and (7) Web-based menu assessment. Figure 2 outlines the number of studies utilising each measurement method in each setting.  There were several differences in how these methods were utilised. Menu reviews were undertaken by either entering menu data into nutrition software to compare to guidelines, categorising menu items into food groups to compare to guidelines, listing foods available at each meal to compare to guidelines, or using a scoring tool to assess compliance with guidelines. Similarly visual observational data were either categorised into food groups or used nutrition software to assess against guidelines; weighed food protocol data were analysed via nutrition software; and digital photography categorised data into proportions of food groups to compare to guidelines. Questionnaires/surveys varied, generally asking questions about availability of foods on menus rather than actual served quantities. Quick menu audits classified food and drinks as consumed every day, sometimes or occasionally. Web-based menu assessment tools enabled data entry by staff at the setting (as opposed to researchers), producing a comparison of menu and recipe data to guidelines. Tables 1 and 2 provide detailed information regarding categorisation of identified measurement methods for the ECEC setting (n = 45) and primary school setting (n = 35). To describe foods and beverages served in childcare centres, assess provider behaviours related to feeding, and compare these practices to national guidelines self-reported questionnaire 4

•
The questionnaire was modified for the United Kingdom from 3 instruments of which 2 were tested for validity and 1 for reliability • Unable to assess all foods and beverages served (only selected) and therefore unable to assess entire dietary quality • Centres received access to the web-based menu tool and training on how to use it. • There were improvements in provision of fruit, vegetables, dairy and meat, and reduction in discretionary food, but no improvement in full compliance goes guidelines • The tool improved food provision but did not translate into full menu compliance  To assess the efficacy of a food service implementation intervention designed to increase provision of foods Quantities of food as well as quality of menus analysed against guidelines • Self-reported data were validated with on-site observation 1(a) Food items on the menu are extracted and/or analysed with nutrition analysis software to compare against setting specific guidelines. 1(b) Food items in the menu were analysed into food groups and compared against setting specific guidelines. 1(c) A scoring tool was utilised to assess menu compliance against setting specific guidelines. 1(d) Menu reviewed and compared to a list of foods available on the menu, for example, vegetables in every meal, but actual amounts of foods on menu not assessed. 2 Observation of foods served by nutrition trained researcher/s and compared to posted menus for comparison of foods served to foods on menu. 2(a) Observation by nutrition trained researcher/s and analysed with nutrition analysis software to compare against setting specific guidelines. 2(b) Observation of foods served by nutrition trained researcher/s and food analysed into food groups to compare against setting specific guidelines. 3 Weighed food method-food served, (and in some cases plate waste measured to closest gram to calculate actual intakes). Data entered onto nutrition analysis software and compared against setting specific guidelines. 4 A questionnaire includes questions related to food provided, nutritional practices and the nutrition environment. Menu compared to setting specific guidelines. 5 Foods consumed were photographed with a digital camera mounted on a tripod with standardised measures for distance between lens and centre of meal plate and camera angle. The photographs were compared to photographs of weighed reference portions of the food to estimate the percentage of food served and consumed and then compared to guidelines. 6 Web-based instruments designed for centres to enter their menus and receive results comparing menus to guidelines.        1(a) Food items on the menu are extracted and/or analysed with nutrition analysis software to compare against setting specific guidelines. 1(b) Food items in the menu were analysed into food groups and compared against setting specific guidelines. 1(c) A scoring tool was utilised to assess menu compliance against setting specific guidelines. 1(d) Menu reviewed and compared to a list of foods available on the menu, for example, vegetables in every meal, but actual amounts of foods on menu not assessed. 2 Observation of foods served by nutrition trained researcher/s and compared to posted menus for comparison of foods served to foods on menu. 2(a) Observation by nutrition trained researcher/s and analysed with nutrition analysis software to compare against setting specific guidelines. 2(b) Observation of foods served by nutrition trained researcher/s and food analysed into food groups to compare against setting specific guidelines. 3 Weighed food method-food served, (and in some cases plate waste measured to closest gram to calculate actual intakes). Data entered onto nutrition analysis software and compared against setting specific guidelines. 4 A questionnaire includes questions related to food provided, nutritional practices and the nutrition environment. Menu compared to setting specific guidelines. 5 Foods consumed were photographed with a digital camera mounted on a tripod with standardised measures for distance between lens and centre of meal plate and camera angle. The photographs were compared to photographs of weighed reference portions of the food to estimate the percentage of food served and consumed and then compared to guidelines. 6 Web-based instruments designed for centres to enter their menus and receive results comparing menus to guidelines. 7 School canteen quick menu audit: This tool assigns product information and serve sizes for each item based on common canteen menu items, eliminating the need to obtain additional information from canteen managers. Table 3 summarises an evidence-based evaluation of the included studies, which is described in further detail in the discussion. • Gold standard for measuring intakes [123,124] and has been adapted for accurately assessing food provision in ECEC [36] and primary school settings [106] • Valid method [123], and whilst not validated in ECEC or primary school settings, studies of other measures utilises weighed food measures as a reference for validating measurement tools due to the accuracy of this method [77,104,106] • Generally used in smaller sample sizes over a shorter period • Eight out of 21 studies (38%) utilising this method received a neutral rating, due to criteria on validity being unclear in the study descriptions

Digital photography
Foods provided were photographed with a digital camera mounted on a tripod with standardised measures for distance between lens and centre of meal plate and camera angle.
The photographs were compared to photographs of weighed reference portions of the food to estimate the percentage of food served and consumed and then compared to guidelines.
• Validated tool in food consumption studies, but no validation studies for food provision at service level • Validated by Kenney et al. [104] in a study examining food consumption in after school care facilities with good correlation between photography and weighed food methods (0.92-0.94) and good inter-rater reliability (0.84-0.95) • Validated by Taylor et al. [119] against visual observation and found 96% agreement with an intraclass correlation of 0.92 • Four out of the five studies that utilised this measurement tool had a positive QCC rating Table 3. Cont.

Method Description Evidence-Based Evaluation
Quick menu audit This tool assigns product information and serve sizes for each item based on common canteen menu items, eliminating the need to obtain additional information from canteen managers. Foods and drinks are colour coded based on classification of every day (green), sometimes (orange) or occasional foods (red).
• Specific tool for assessing primary school canteens in settings such as Australian schools where children can bring food/lunchboxes from home or purchase foods from their school canteen • Not an appropriate measuring tool for ECEC or primary school settings where all food is served • Validity study found agreement between quick menu audit tool and observations to be 84% with Kappa of 0.68 [112] • Three out of the four studies that utilised this method had positive QCC ratings, with the fourth receiving a neutral rating due to some validity questions not clearly articulated in the studies Web based menu self-assessment tool Designed for centres to enter their menus and receive results comparing menus to guidelines.

•
Validation study conducted by Patterson et al. in a primary school setting [114] • Sensitivity ranged from 0.85 to 1, specificity from 0.45-1.00 and accuracy 0.67-1.00, therefore found to be a feasible instrument for self-assessment of menus • Clinical trial by Grady et al. [65] found that use of the tool by centres for self-measurement resulted in improvement in food group provision but not full compliance • Both the studies utilising this method had a positive QCC rating Measurement methods differed in validation and accuracy. Some visual observation methods were either validated (n = 4 out of 17) [54,67,82,108], adjusted from validated methods against weighed food records (n = 1 out of 17) [54], or tested for reliability between the observers (n = 5 out of 17) [45,82,90,91,94]. Some questionnaires were validated (n = 6 out of 13) [61,66,68,74,105,110] or adapted from validated questionnaires against visual observations or menu reviews (n = 2 out of 15) [50,54]. Digital photography was validated against weighed food measures in two studies [104,119] or adapted from a validated method in one study [107]. One study used a quick menu audit validated against visual observation in an Australian school canteen setting [112]. Web-based menu selfassessment was validated against menu items in one study [114].

Discussion
To the authors' knowledge, this is the first systematic review to identify and evaluate measurement methods used to assess food provision and menu compliance in ECEC and primary school settings. Overall, 70% of the studies in this review had a positive rating, assessed according to the Academy of Nutrition and Dietetics' Quality Criteria Checklist (QCC) [39], with the remainder being assigned a neutral rating, due to lower scoring on validity screening questions. It is important to note that some of the validity screening questions in the QCC, such as bias in subject selection and blinding of subjects, were not applicable for a number of included studies as it is unclear whether participating services are biased towards better food provision and blinding to food provision assessment at a service level is not possible. Given this, the use of an unvalidated measurement method contributed to a lower study quality rating (QCC) score. This indicates that the quality rating of food provision research in ECEC and primary school settings could be strengthened through the utilisation of validated measurement tools and by improving internal validity in research studies.
This discussion will outline the evidence, including strengths and weaknesses, for each of the seven food provision methods/tools identified across ECEC and primary school settings, with recommendations to inform future research and practice.

Menu Reviews
Neary a quarter of studies (23%; n = 19), with 32% in ECEC and 11% in primary school, were found to use menu reviews as a food provision measurement method [31,34,37,43,[46][47][48]56,58,[62][63][64]66,67,71,72,74,80,83]. Menu reviews were usually conducted to determine the quantity of food, and in some cases quality of food (n = 6 out of 20) and variety of food served (n = 2 out of 20) compared to set standards or guidelines. Food quantity was either assessed by analysing all items on the menu and comparing to the guidelines, which required detailed recipes with exact quantities of each item in the recipe; or by analysing the menu based on a list of foods that are available on the menu, for example vegetables in every meal, without the analysis of the exact amounts of food. These differences in determining quantity would influence accuracy of menu review, as those studies that analysed based on a list of food items available, would not accurately determine amounts of food provided.
On average, menu reviews were conducted over 2 weeks or more, compared to observations which were conducted for 1 day or more. Menu reviews therefore have a unique advantage of capturing average food provision, as well as analysis of quantity, quality and variety of food provided over time. Only two studies in the ECEC setting focused on quantity as well as quality and variety of menus [81,106]. A major disadvantage of menu reviews is that planned menus may not always reflect actual food provision. Four studies compared planned menus to actual food served, with varying results [42,48,51,60]. A crosssectional study (n = 6 LDC centres, children aged 3-5 years) conducted by Fleischhacker et al. [60] in the United States found planned menus were inconsistently followed by the childcare centres, with only 28% of food served matching the planned menu. Similarly, Alves et al. [42] (n = 5 LDC centres, children aged 7-24 months) found only 20% of food served matched the planned menus. Conversely, Benjamin-Neelon et al. [49] (n = 84, children aged under 6) found an 86.6% match between food items served and planned menus, and Breck et al. [51] (95 LDC centres, children's ages not stated) found an 87% match. While the studies that found higher matches between planned menus and actual food served were larger studies, they were also conducted over a shorter time of 1-2 days [48,51]. The studies with a lower match between planned menus and food served were conducted over a longer time period, with Alves et al. comparing data over a 6 week period [42] and Fleishhacker over 6 months [60].
This suggests that menu reviews may not be an accurate indicator of actual food provision in ECEC and primary school settings over a longer time. Moreover, menu reviews may be compromised if insufficient information is available such as portion sizes, types of foods (e.g., low, or high fat milk), and methods of preparation, leading to researchers being tasked with making assumptions and potentially adding to inaccuracy of reporting [63,74]. Menu reviews also rely on skilled professionals for menu coding and nutrient analysis [46,59,109]. Additionally, evaluation testing for validity and reliability for menu reviews appears to be lacking. To the authors' knowledge, there are currently no validated menu review tools for the ECEC or primary school setting.

Visual Observation
Twenty-one per cent (n = 17) of included studies used visual observation to assess food provision in ECEC (25%) and primary school (14%) settings. This method requires trained observers to visually estimate the amount of food served to (and in some studies also consumed by) children, including visual estimation of portion sizes of foods before and after consumption [125]. A major limitation of visual observation is that food provision is determined through estimation rather than calculating the exact amount [104,106], therefore this method is highly reliant on well trained observers and a standardised protocol for data collection. The approach varied across studies with one study using a 5-point scale where an untouched plate scored 5, if at least one bite was consumed scored 4, if three-quarters of the food remained scored 3, if half the food remained scored 2, if a quarter of the food remained scored 1, and if no food remained scored 0 [106].
Visual observation may be less costly and time-consuming to implement as only training of observers is required, with one validation study (n = 111 primary school children) reporting this method as being lower in cost compared to weighed food method and digital photography [104]. A limitation of this method is the number of children that can be observed at one time, which was commonly reported as 4 children per observer at a time in most ECEC and primary school studies, suggesting this method is better suited to small scale settings [42,45,54,60,67,85]. In contrast, larger cohort studies (between 20 and 95 ECEC or primary school settings) either only observed 1 meal (ECEC/schools) or snack (ECEC) [44,48,51,57,82], with data collection not representative of usual food provision.
Visual observation demonstrated an intraclass correlation of >0.92, indicating excellent reliability, when compared with weighed food measurement in a validation study in the primary school setting [104]; however, only snack consumption was measured. Conversely, another validation study conducted in the primary school setting found poor correlation (5.5-24.7%) for visual observation compared to the weighed food method; however, this study used a point scale for observation rather than estimating portion sizes [106]. A high variation was also noted in terms of number of days of observation, ranging from one day to a few weeks [42,48,51,60]. Studies conducted over fewer days may reduce the ability to collect representative data of usual food provision by not capturing day to day variations and therefore may not accurately reflect actual food provision in the ECEC/primary school setting. This lack of a recommended study length to demonstrate usual food provision therefore requires further investigation.

Self-Reported Questionnaire/Survey
Self-reported questionnaires were used in 17% of included studies, mainly in the ECEC setting (11 studies, 23%) [32,38,47,51,[53][54][55][56]59,63,76], with only three studies in the primary school setting [90,99,101]. Questionnaires mostly assessed compliance of nutrition policies, menus and/or feeding practices in relation to prescribed guidelines. About half of the questionnaires used were validated or adapted from a validated questionnaire [44,50,61,87], validating items against visual observation or menu reviews, with no validation studies using weighed food measurement. Studies using questionnaires had cohorts ranging between 29 and 4360 ECEC centres/primary schools, with most having a study cohort of 200 ECEC centres/primary schools or above (n = 7 out of 13) [51,53,59,90,98,101,104]. This may imply suitability for application in large-scale ECEC/primary school settings.
Questionnaires, however, can be associated with social desirability bias, linked with under-or over-reporting of certain foods [126]. Moreover, a respondent's lack of knowledge regarding various nutrition practices may result in reporting errors [44], as reported by Reilly et al. [112] who found poor agreement between self-reported data and data collected from on-site observations. Studies using self-reported questionnaires mostly collected data about the provision of certain items, for instance "vegetables served at each meal", without specific data on the type or number of vegetables served. Such data do not provide accurate detail on foods provided in ECEC or primary school settings. As a possible solution to this, three studies used multiple methods, such as a questionnaire alongside weighed food records [54], a questionnaire alongside a menu review [59] and a questionnaire alongside observation [101]. Completing a questionnaire alongside another valid and accurate measurement method warrants further investigation as this may allow researchers to triangulate data and more accurately determine food provision. However, associated time and financial costs need to be an integral part of determining the realistic application of this approach.

Weighed Food Protocol
Weighed food protocols were used by 27% of included studies (n = 22), with more frequent use in the primary school (37%) than the ECEC setting (19%). Weighed food protocols are considered "gold standard" for measuring individual-level food intake and, in some studies, have been adjusted for use in ECEC and primary school settings [36,127,128]. Stud-ies using this method tended to have smaller cohorts, with most ECEC studies assessing between 2 and 30 centres, and with only 3 of 13 primary school studies assessing cohorts of over 100 schools [92,98,116]. The limited use of this method for larger cohorts may be due to the labour intensiveness of weighing food served at the individual ('plate') level. As a potential solution to this, Sambell et al. [36] developed a weighed food protocol for service-level food provision, based on the 'gold standard', measuring raw ingredients in the preparation phase, and using the average portion of 3 'plates' served to children. This method appears less labour intensive to implement and in addition, there is little disruption to the children during mealtime as measurements are conducted in the kitchen/preparation area [36]. This protocol has yet to be validated against individual plate measures and should be a key research activity given the potential as a scalable option for larger cohort studies. Thirty eight percent of studies utilising this method received a neutral QCC rating, as the items for validity were not clearly articulated. Researchers using the weighed food protocol may be making an assumption that, as the 'gold standard', clear articulation of validity is not needed; however, this reporting needs to occur to strengthen the quality of such research studies.
Another limitation for several studies using the weighed food protocol is the time over which data can realistically be captured. The number of days over which food provision was analysed using this method varied between 1 and 5 days, with no studies measuring food provision for more than 5 days. Whilst 5 days can capture some variations in food provision, capturing seasonal variations needs further consideration. Research is therefore required to determine an acceptable length for data collection to ensure findings accurately represent food provision. It may be that data collection needs to occur over a designated number of days over several time points to better reflect the seasonality variation in menus and more accurately reflect food provision per se.
More recently, weighed food protocols have included a measure of weighed food waste, which can provide a more accurate calculation of food consumption data in addition to food provision data. At a service level, the aggregated plate waste method is considered more suitable for food provision studies compared to the individual plate waste method [129], where total amount of food consumed is calculated by deducting the total amount of food wasted from the total amount of food served (as average portions) and dividing this by the number of children. A validation study by Chapman et al. [130] found good agreement between individual and aggregated plate waste methods; however, both plate waste methods potentially underestimated vegetable consumption as some menu items, such as sandwich fillings, were not measured separately [129]. There is potential to combine aggregated plate waste with Sambell et al.'s method [36] to get an accurate measure at service level of food served, consumed, and wasted, but this will require further research and validation. Although aggregated plate waste methods may provide more accurate data, researchers are unable to identify which food groups make the largest contribution to food waste, and additional space is required for food waste collection [131,132]. Capturing food waste in addition to food served also has the potential to respond to an increased interest in the cost saving benefits of reducing food waste in ECEC and primary school settings and subsequent climate impacts [133]. Future research should aim to include food waste measurement when researching food provision, focusing on the validation and standardisation of both individual and aggregated plate waste methods as scalable options.

Digital Photography
Digital photography is based on visual estimation of food images, which are taken according to a standardised distance and angle from the food served [104]. Five studies used digital photography to measure food provision and menu compliance in ECEC (n = 3) and primary schools (n = 2) [62,71,89,92,103]. A validation study conducted by Kenney et al. (n = 111 primary school children) [104] compared digital photography with the weighed food protocol to assess the accuracy, time and costs involved in this method. Digital photography had good agreement with the weighed food protocol regarding accuracy of estimating food consumption, and implementation cost was less than the weighed food protocol [104]. However, this validation study only examined snack food consumption which may not be generalisable to other types of meals with more food components, and may therefore also be more costly when applied across all meals served [104]. Taylor et al. (n = 2 primary schools) [120] validated digital photography in the primary school setting against weighed food and food waste measures and found good validity (Pearson's correlation r = 0.91-0.96) and strong inter-rater reliability (0.92 (95% CI 0.90 to 0.94) in the assessment of fruit and vegetable consumption. The study highlighted slight underestimation of starting portions and waste of leafy greens (salads) [119].
As both the above studies focused on certain menu items (snacks, fruits, or vegetables), the use of digital photography in assessing mixed meals, defined as meals with more than one component (for instance a meat and vegetable stew), is unclear and may not be generalisable to all meal types. Furthermore, digital photography requires specialised equipment, may have high respondent burden due to interruptions during mealtime for photographs, and incurs human and resource costs capturing and analysing photographic data. One included study provided a possible solution to this by using iPads as a lower cost than digital cameras [107], supported by three training sessions with intra class coefficients improving through training sessions (0.86-0.98 by the 3rd training session). This review found that digital photography was used in smaller cohort studies, with the maximum number of ECEC centres assessed being seven centres. The use of this method for larger cohorts therefore warrants further research.

Quick Menu Audit
This method of measuring food provision was only used in primary school settings in countries where children have the option of bringing food from home or purchasing some or all their food from a school canteen [112,134]. Four included studies used the quick menu audit method to assess the healthfulness of items available for purchase in the school canteen [109,111,112,118], capturing product information and serve sizes for each item on the canteen menu, thereby eliminating the need to obtain additional information from canteen managers [112]. Foods and drinks were colour coded based on a classification of foods recommended for daily consumption (green), foods that should be consumed on some days (amber) and foods that that should be consumed only occasionally as they are highly processed, high in fat, sugar and/or sodium (red) [112,118]. This tool was validated against the visual observation method by nutrition trained researchers with good agreement (kappa = 0.68, 84% agreement) [112]. Studies using this method assessed cohorts between 53 and 168 schools, indicating that this measurement method may be suitable for larger settings and research studies. The tool, however, is only applicable to settings where food is available to be purchased/selected by children, and therefore may not be an applicable tool for ECEC or primary school settings where food is provided on-site.

Web-Based Menu Assessment
Whilst a web-based menu assessment could be classified as a menu review method, it is considered a stand-alone tool within this review due to its unique ability to be used by staff and health professionals within the ECEC/primary school setting for menu planning. One validation study in a primary school setting [114] found the web-based menu assessment to have good agreement (Cohen's cappa > 0.60) and reasonable reliability (intraclass correlation ranged from 0.33 to 0.99-fair to almost perfect) compared to an on-site menu review conducted by a nutrition researcher [114]. It is important to note that this study assessed nutrient components of meals (saturated fat, iron, vitamin D and fibre) rather than overall healthfulness of menus. In recent years, there has been a shift to promote foods instead of nutrients, as evidenced in Dietary Guidelines around the world [135]. At times, collecting nutrient provision is more relevant; however, with robust data collection methods, both food group and nutrient data could be obtained to support better translation for different sectors.
In addition, as discussed earlier, a menu review can be compromised by several issues, many of which would affect web-based menu assessments, such as menus not reflecting actual food provision. There is potential, however, for web-based menu assessments to assist ECEC and primary school staff in the planning of their menus to meet recommended guidelines. In Australia, two government-funded web-based menu assessment tools, namely FoodChecker [136] and FeedAustralia [137], are available to the ECEC setting. These websites, however, are based on jurisdictional guidelines and therefore measure menu compliance against different parameters within an environment that promotes national dietary guidance [138], which poses a barrier to wider (national) uptake of such a tool.
A randomised clinical trial on the use of FeedAustralia's menu planning tool and its impacts on food provision found that while no centres using the tool reached full menu compliance, use of the tool was associated with improved provision amongst most food groups [65]. This study, however, relied on self-reported menu data and observational child dietary intake data, rather than actual food provision at the service level, to determine compliance. The potential of web-based menu assessment tools to support menu planning and self-assessed menu compliance and enhance food provision in ECEC and primary school settings therefore warrants further investigation.

Implications for Research and Practice
There is a fundamental premise that children need to be provided with adequate serves of recommended food groups if they are expected to consume adequate serves, and in both ECEC and primary school settings, the assessment of this needs to be conducted through service-level food provision measurement.
This systematic review aimed to identify current methods/tools utilised for determining food provision at the service level in ECEC and primary school settings, and to provide a recommendation on a standardised approach based on these findings. This review found various degrees of validity and accuracy of measurement tools, and of note, there were varied benchmarks against which tools were validated. Utilising a standard protocol for the measurement of service-level food provision could potentially enhance research rigour, allow for the accurate comparison of research findings as well as monitor changes over time more accurately.
The weighed food protocol is considered the most accurate measurement of individuallevel food intake, and therefore the 'gold standard' [127,128]. While used most frequently to assess individual-level food provision, within ECEC and primary school settings, a servicelevel protocol, where each ingredient is weighed and recorded prior to the meal being prepared, has been adapted from the 'gold standard' and applied in the ECEC setting [36]. Future research should aim to validate this method for use in ECEC and primary school settings and explore its potential scalability for larger cohorts.
It is important to consider the differences in primary school food provision environments across countries, such as the United States and certain schools the United Kingdom [139] where food is provided to children, compared with Australia, New Zealand and the Netherlands, where children often have access to a school canteen to select and purchase items if food is not brought from home [111,134]. In the latter, a weighed food protocol may prove to be laborious and impractical to measure food provision as the proportion of food either purchased from the canteen or brought from home is unknown. In this scenario, a quick menu audit tool appears to offer a low cost, low burden, and validated tool to categorise the healthfulness of foods available for purchase at the school canteen in primary school settings [65,114].
Finally, a web-based menu assessment tool shows promise for ECEC and primary school settings in supporting self-directed menu planning, and evidence suggests that it does improve the menus of ECEC services [65]. Web-based menu assessment is subject to the same limitations as menu reviews in that it does not necessarily measure actual food provision. Future research should further investigate the uptake of web-based menu assessment tools to determine ease of access and usefulness, overall validity, and scalability.

Strengths and Limitations
A key strength of this systematic review is that it is the first study to examine types of measurement methods/tools used to assess service-level food provision in ECEC and primary school settings. Furthermore, this review offers recommendations to inform research and practice, and to guide the development and use of a standardised approach for the measurement of service-level food provision in ECEC and primary school settings. The process of data extraction and screening was overseen and cross checked by multiple authors, and the quality of each method/tool was also critiqued, thereby increasing the robustness of the review process. This study is not without limitations. Research articles may have been missed as no hand searching of articles was done in the review process and references of all included studies were not included in the search strategy. This review may be subject to publication bias as only peer-reviewed published English language studies were included. Finally, this review focused specifically on identifying food provision measurement methods/tools used within ECEC and primary school settings, consequently, recommendations may not be generalisable to other settings.

Conclusions
This is the first systematic review to identify and critique methods/tools used to assess service-level food provision within ECEC and primary school settings. Seven methods/tools were identified, with varying degrees of validity and accuracy, and varied benchmarks for which validity was measured against. This illustrates the importance of developing a standardised tool to measure and assess service-level food provision and menu compliance in ECEC and primary school settings. The review found the weighed food protocol to be the most commonly used and most accurate tool to measure individuallevel food intakes. The weighed food protocol has potential for adaption to measure food provision at a service level; however, future research will be needed, including validation. Validating a standardised weighed food protocol to measure food provision at a service level will allow for accurate comparison of findings across ECEC and primary school settings, providing reliable monitoring data and opportunities to enhance food provision.