The long-list of the 375 potential indicators identified from Step 1 can be found in Online Resource 1.

Participant Characteristics

Participant characteristics from Steps 1 and 2 are not described to protect confidentiality, given the small sample size. In the online survey in Step 3, participants’ affiliate institutions were based in all WHO regions (Western Pacific n = 20; Americas n = 14; Europe n = 8; Africa n = 3; Eastern Mediterranean n = 2; South-East Asia n = 2, Unknown n = 2). When asked what INFORMAS domains they had been involved with or methods they had used in the past, participants were most frequently familiar with the food promotion (40%), food retail (34%) and food labelling (32%) domains, followed by food composition (24%), food prices (16%), and food provision (14%), with only one having experience with the trade and investment domain (2%). In addition, 32% of participants indicated that they had been involved with or used the public sector domain and 24% the private sector domain.

Short-List of Indicators

Table 1 describes the refined short-list of 16 indicators that were identified in Step 2, incorporating feedback on content and framing from open-ended feedback in Step 3 and from the workshop and authorship team.

Table 1 Short-list of indicators for monitoring and assessment of food environmentsComposition

The Composition domain considers the nutritional quality of the food supply, comprising both packaged foods and food prepared ‘out-of-home’ (e.g., restaurant/fast-food), including both formal and informal food markets [25]. Four indicators were short-listed in relation to this domain (Table 1). Three indicators focused on packaged foods, and one focused on fast-food chains.

For packaged foods (indicators COMP1, COMP2 and COMP3), the monitoring data underpinning the prioritised indicators would likely be limited to products available in supermarket chains with the largest market share. These indicators rely upon information being available for products (e.g., nutrient declarations and/or ingredients lists), which may differ between countries depending on national labelling regulations and voluntary food industry practices. Across the three indicators related to packaged food, participants recognised the importance of focusing monitoring efforts on a list of priority food categories or developing sampling strategies to minimize the burden of data collection (e.g., regular rolling data collection over time).

The first indicator (COMP1) focuses on the proportion of packaged food and drinks products that are classified as unhealthy. Two potential sub-indicators are proposed, each adopting a different measure related to product healthiness: (a) a national nutrient profiling system; and/or (b) degree of food processing according to Nova. The two different measures of product healthiness were selected to indicate different characteristics of the available products, and in response to feedback from participants (described further below) that indicated that there was no universal agreement on the most appropriate food classification system to use across all indicators, and that simultaneous use of multiple systems would help meet multiple objectives such as national relevance and international comparisons. Participants noted additional value in applying both to understand the nutritional value and degree of processing among products for a comprehensive understanding of the food supply.

Two indicators (COMP2 and COMP3) focused on specific nutrients in the packaged food supply. COMP2 examines the extent to which products meet relevant government formulation targets for nutrients to limit. Depending on the government targets in a particular country, it may be necessary to have separate indicators for each nutrient (e.g., the proportion of products that meet sodium, sugar and saturated fat targets). Given that relatively few countries have established country-level targets, targets from the WHO [48] for sodium could be useful and applied across countries, and the elimination of trans fat in the food supply provides a benchmark for evaluating levels of trans fat. It was noted that these reformulation targets would likely change over time, which may require consideration in the monitoring of this indicator.

COMP3 examines mean nutrient levels in food and drink products in major healthy and unhealthy food categories. While these are generally and preferably reported per 100mL–100 g, it was noted that the mean level of nutrients could be considered per serve, which is applicable in some country contexts and should be considered if relevant. Across COMP2 and COMP3, when data are available, monitoring of added sugars or free sugars was considered important in addition to total sugars. Participants highlighted that other aspects of food composition could also be considered, including fibre, calcium and non-nutritive sweeteners.

The final indicator, COMP4, related to foods available in major fast food chains. With regards to systems for classifying the healthiness of foods in the ‘out-of-home’ setting, we noted that national nutrient profiling systems specific to this setting are relatively uncommon. The required information to conduct nutrient profiling is not usually publicly available for these foods, and many existing approaches often do not incorporate important elements, such as portion size. We identified that the Health Star Rating has been evaluated for use in restaurant foods [49].

Participants noted that analyses could be disaggregated by company (manufacturer or restaurant chain), which would provide an opportunity to assess individual company progress towards achieving healthier food offerings. Some experts suggested that these data could be reported as sales-weighted to better represent the influence of the food supply on diets; however, others suggested that data required for such analysis would likely be cost prohibitive or unavailable in many countries and settings. It was noted that sampling may need to be adapted in countries where informal markets are larger sources of food acquisition.

Price

The Price domain includes the analysis of the price and affordability of individual foods and standardised diets [23]. Three indicators were short-listed, including one in relation to the price of individual foods, one relating to the price of diets, and one relating to the affordability of diets. Sampling (for data collection) was identified as a key element of this indicator, as prices are known to be affected by seasonality, regional differences, systems shocks and price-chain disruptions, as well as price differences between different types of outlets and retailers.

PRICE1 focuses on changes over time in the price of key healthy and unhealthy food and beverages. Given the volatility of food prices and secular trends, participants noted that this would likely need to be reported on a regular (e.g., annual) basis to be most valuable.

PRICE2 describes the cost of different diets. This indicator typically involves integration of product or category-level price estimates with nationally representative dietary intake data to estimate the cost of the current average diet compared to a diet that adheres to national dietary guidelines or a globally accepted healthier dietary pattern. This indicator thus requires regularly collected food price data and dietary intake data upon which an estimate of the current diet can be based.

PRICE3 uses the same methodology as the PRICE2 indicator but presents the estimated cost of the current and recommended diet as a proportion of household income to demonstrate the affordability of diets. Multiple scenarios would be used to describe the affordability of the diet among different population groups, with differing household incomes (households with an income based on minimum wage, households on social assistance, etc.). This approach requires national data on household income and the establishment of a reference household or households (e.g., two adults and two children).

Participants noted the potential for data harvesting methods, such as web-scraping, to support regular price data collection from various regions. Others noted that such methods may not be suitable in all contexts, for example, in countries or regions where online data are not available or there are limited resources for webscraping and associated data cleaning practices. Several participants discussed the potential for using consumer price index (CPI) data that are routinely collected by countries. Participants indicated that, while potentially valuable, there were several limitations to these data [50, 51] which differ between countries and often include a small range of key foods that were not developed with health considerations in mind and are not sampled from all areas. It was noted that international comparisons would require exchange rate parity.

Participants also recognized that data are available on food affordability from the FAO, including a measure on the cost of a healthy diet [50]. The goal of this FAO work is to provide global estimates of food security and monitoring for the Sustainable Development Goals and the eradication of hunger, food insecurity and malnutrition, including the cost of a healthy diet, focusing on regional estimates and changes over time. Some participants indicated that the FAO data are less granular than what is proposed herein, do not compare with the cost of current diets, and may not provide the breadth of information to inform policy at a national level.

Promotion

The Promotion domain explores the extent to which children are exposed to the promotion of unhealthy foods [28]. Children are typically exposed to commercial marketing across many media and other settings with which they engage. Participants recognised that it can be challenging to monitor all relevant media/settings, and, thus, monitoring efforts may need to prioritise key media/settings [36, 52]. The proposed strategy puts forth seven media/settings that can be prioritised for monitoring depending on the country context, noting that marketing in retail settings is discussed in the Retail module. Of the seven settings (TV advertising, digital advertising, outdoor advertising, product packaging, community sport-based advertising, sport sponsorship and schools), countries undertaking monitoring were suggested to prioritise at least two for inclusion. For most countries, participants anticipated that digital advertising would be most relevant, as this has been identified in the literature as an increasingly important contributor to children’s advertising exposure [52]. TV advertising was also identified as likely to be of high importance, although trends in children’s engagement with this media should be considered. Participants noted that there may be contexts for which these media are less relevant, such as contexts where few children use digital devices or have internet access, where there are declining rates of commercial TV use, where existing policies are focussed on specific sub-group of children (e.g. aged under 12 years) or where particularly settings, such as sport settings, have important cultural relevance. The selection of media or settings to be monitored may also be based on resources available for data collection and/or existing data on food promotion within the region or neighbouring countries, including industry data on advertising expenditure and others.

Current methods for collecting data include recording television or digital screens for lengthy periods or collecting large volumes of images (e.g. for product packaging marketing), posing challenges with resource intensive data-coding. Participants noted that there were emerging methods whereby data collection and coding methods can be streamlined with web-scraping and machine-learning techniques using image recognition [52], and that additional validation of will support their use in monitoring. Lastly, it was noted that emerging that the best practice approach is to use methods that collect data on ‘actual’ exposures where resource and capacity allows, but ‘potential’ exposure is an alternative.

Labelling

The Labelling domain monitors the provision of nutrition information on product labels. LABEL1 includes two sub-indicators relating to the proportion of products that carry nutrition information on the front and the back of package, and the extent to which those products are healthy. Participants noted that monitoring priorities may differ based on whether labelling is mandatory or voluntary. In many countries, some food labelling components are mandatory (e.g., ingredients lists, nutrient declarations, and, in some cases, interpretive front-of-package labelling schemes). In principle, compliance with mandatory labelling is likely to be a lower priority for monitoring efforts than uptake of voluntary measures, although participants noted that implementation of mandatory labelling may vary substantially according to the country context. This indicator requires up-to-date data or images of product packaging to assess the extent of implementation of food labelling, and/or to build the case for strengthening existing measures. Participants noted that this indicator was designed to focus on surveillance of government-endorsed labelling only, and was not intended to include monitoring of industry-led initiatives (such as ‘Guideline Daily Amounts’ labelling or health symbols developed by food companies).

The second labelling indicator (LABEL2) relates to the use of nutrition and health claims on products, and the healthiness of the products carrying claims. Participant feedback highlighted the ability of this indicator to identify the prevalence of potentially misleading uses of nutrition and health claims.

Provision

The Provision domain monitors the foods available and the way they are promoted in public sector settings [26]. Input from participants identified that schools were the priority public sector setting for surveillance. As such, the indicators proposed focus on school environments; however, it was noted that these same indicators could be applied to other public sector contexts, such as hospitals, early childhood education centres, correctional facilities, recreation centers, community centres or spaces, libraries and others. The proposed indicators align with a recent report from the UNICEF East Asia and the Pacific Regional Office: the Nutrition Environment Assessment Toolkit for Schools (NEAT-S) [53].

The first provision indicator (PROV1) aims to monitor the proportion of schools that provide or sell unhealthy foods, including foods provided or available for sale in school, nutrition/meal/snack programs and those sold in cafeterias/canteens, school stores and vending machines. It was noted that, when relevant, indicators could be disaggregated by source (school meals/cafeterias/canteens/stores/vending machines) or by the nature of provision (provided for free or sold). The indicator refers to items available ‘regularly’, which would typically be defined as more than once per week. For rotating menus, availability could be assessed on the day of data collection.

The PROV2 indicator examines the proportion of schools that provide or sell healthy foods (including unprocessed or minimally processed fruits and vegetables, and safe drinking water) every day that their food services are operational. Participants noted that a category of whole grain foods may be particularly relevant in some contexts and may require additional criteria for nutrients or other food components in whole grain products. Participants also noted the importance of monitoring school food and nutrition policies, but this was not included in this short-list as policy monitoring was considered out of scope for this work. We also note that monitoring of marketing in schools was included as part of the Promotion domain. Several participants highlighted the importance of monitoring food outlets immediately surrounding schools as these foods may ‘compete’ with school foods, which would fall within the Retail domain [54].

Sampling was identified as a challenge for monitoring in this domain, particularly in countries with multiple sub-national governments and/or large geographic areas. If sampling strategies include a site visit to conduct direct observation, indicators could be adapted for ‘the day of the survey’. Participants also indicated that ethics clearance for in-person data collection in schools can prove a barrier to monitoring efforts, and likely requires governmental support and coordination, although participants noted that some menus may be available online. It was noted that, in some contexts, there are likely to be very few (if any) schools that would have no unhealthy items available; in these contexts, improvements over time would likely be a better indicator than the absolute percentages.

Retail

The Retail domain monitors the availability and promotion of foods in retail food environments, encompassing both the spatial distribution of retail outlets in the community as well as in-store environments [24].

The RETAIL1 and RETAIL2 indicators relate to in-store environments. RETAIL1 explores the proportion of price promotions, such as temporary discounts or multi-buys, in relation to unhealthy compared to healthy foods, expressed as an average over time (e.g. per week, per month, etc.). The indicator is designed to capture major grocery retailer chains within a country context, and types of retailers included could be based upon market share data. For example, in many countries, supermarkets would be the primary focus, whereas in some countries monitoring may include convenience stores if they make a prominent contribution to population-level food purchases.

The RETAIL2 indicator relates to the proportion of checkouts in grocery retail settings that are free from unhealthy foods. Challenges with the RETAIL1 and RETAIL2 indicators include sampling strategy, as there is likely to be considerable variation in these indicators over time, as well as between store locations or brands in some instances. While the potential for online data collection exists – and is increasingly relevant as more shoppers turn to digital grocery platforms – this may not be possible in all contexts and may not represent promotions in physical outlets where the majority of grocery shopping still currently takes place. Although there are a range of tools to monitor the healthiness of online grocery retailers under development, large-scale monitoring in these settings will still require advancements in automated data collection to streamline the assessment process.

The RETAIL3 indicator assesses the geographic distribution of retail outlets in the community, and monitors the ratio of different types of food outlets per 100,000 people, by geographic location (such as between states or provinces, level of urbanicity, or area-level socio-economic status). Many participants queried the precision by which healthy and less healthy food outlets can be defined, noting that there are inconsistencies in the literature, and no consensus on methods for doing so, with multiple approaches available [55, 56]. For example, the healthiness and accessibility of some types of outlets (such as supermarkets) may differ between country contexts, and many restaurants offer a wide range of foods that vary in healthiness [13]. Industry-owned data for this indicator are likely to be available for some countries via proprietary data sources, such as the Euromonitor Passport database. Government business registries may also be available to inform this measure in some contexts, and web-scraping approaches to identify different types of food outlets are also emerging [57].

Retail indicators could potentially disaggregate data by geographic location, providing an opportunity to explore elements of inequities between areas of higher and lower socio-economic status, or between regions. RETAIL1 and RETAIL2 could equally be explored by grocery chain, providing understanding of company-level practices. It was also noted that there have been recent advances in government requirements for reporting of sales of healthy and unhealthy food items by food retailers and manufacturers [58], which could be another avenue for monitoring retail settings that requires little data collection effort.

Trade & Investment

The Trade and investment domain focuses on monitoring the impacts of international trade and investment liberalization, facilitated by trade and investment agreements, on food environments [29]. The TRADE1 indicator includes three sub indicators for import, production and export of key food categories. It was considered necessary for these three sub indicators to be monitored in tandem to provide an overall portrait of the food supply as linked to global markets. Given the nature of available food trade data that is typically classified at a high level of aggregation, the indicators in this domain are recommended to be assessed according to the level of processing, using Nova. Participants recommended that monitoring should assess time trends (for example, five year periods), and likely could be monitored retrospectively (as far back as 1990).

Participants highlighted the importance of investigating foreign direct investment into the food system. However, determining potential health-oriented indicators using the available data was identified a challenge for monitoring this area, as the data are generally not granular enough to examine food-specific trends.

Prioritisation of Key Indicators

The results from the prioritisation exercise within the online survey are shown in Online Resource 2, which shows how often indicators were selected within the top seven, and the distribution of the assigned rankings (from 1 to 7) for each indicator. The final set of prioritised indicators are presented in Box 2.

Box 2. Prioritised indicators for monitoring and assessment of the healthiness of food environments at the country level.

• COMPOSITION (COMP1): Proportion of packaged food and drinks products that are classified as unhealthy according to a nationally relevant nutrient profiling system and/or the Nova system

• PROMOTION (PROMO): Frequency of exposures to unhealthy food advertisements in 2 primary media/settings to monitor from television, digital, outdoor, elite sport, community sport, product packaging, and school environments.

• PRICE (PRICE3): Price of a healthy diet compared to the price of a current diet as a proportion of income 

• PROVISION (PROV1): Proportion of schools in which unhealthy foods (sugary or artificially sweetened drinks, fried foods, salty snacks, sugary snacks) are provided or available for sale regularly

• LABELLING (LABEL1): Application of back of package and front of package labelling systems

Cross-Cutting Implementation ConsiderationsClassifying the Healthiness of Foods

Most of the proposed indicators necessitate the categorical classification of foods as ‘healthy’ or ‘unhealthy’. Initial reflections as well as feedback from the broader group highlighted the lack of consensus regarding food classification systems that could be appropriately applied across settings [44,45,46].

Reflections from participants indicated the importance of applying a country-specific food classification system, ideally developed or endorsed by governments, when available, to enhance policy relevance in a local context. These systems are typically designed to align with national nutrition guidelines or policy objectives (e.g., Health Canada’s front-of-package labelling regulations [59], Food Standards Australia New Zealand Nutrient Profiling Scoring Criterion [60]). WHO regional nutrient profiling models [61,62,63,64,65] could also be used. It was noted, however, that application of country-specific tools or systems may not facilitate international comparisons.

Participants noted the potential value in applying the Nova classification system as a measure of the level of food processing for some of the selected indicators [66]. The Nova system was identified as particularly useful for some indicators (e.g., in relation to trade and investment) where ingredients or nutrient-based data are not available.

Feedback suggested the inclusion of a model that incorporates both nutrients and level of processing; however, there is still a limited number of existing classification systems that have incorporated both elements that would be suitable across country contexts [67]. In consideration of this feedback, we did not recommend the use of a single scheme across the set of prioritised indicators. For some indicators we proposed use of multiple different systems (e.g., refer to COMP1, COMP4), and for others we selected the system that was most fit-for-purpose (e.g., TRADE1). Overall, we think this balances the need for policy relevance, whilst facilitating cross country comparisons where possible. Development of a harmonized food profiling approach remains a priority.

Feedback also indicated the potential utility of algorithmic food classification schemes that provide an indicator of overall nutrition quality of foods, either across the board or within food categories. Prominent examples of such schemes include the Health Star Rating and Nutri-Score, although there was no consensus on which may be most appropriate, or if other systems may be preferred. The Health Star Rating system has been previously used for international comparative purposes in food environments [68, 69]; however, participants expressed that other systems could be used for international comparison purposes and that the application of such a system may not be policy relevant in some contexts.

Key Food Categories

For many of the policy domains examined, sampling (for data collection) and reporting of indicators relies on the inclusion of specific food categories (e.g., sugary drinks, breakfast cereals, etc.) considered most relevant for monitoring food environments. The selection of priority food products may differ depending on country context (e.g., dietary patterns, food culture, and local policy priorities) but likely fall into similar general groups. The identification of specific food categories for international monitoring was beyond the scope of this paper, although participants proposed considerations for identifying such categories. Suggestions included: food categories that are the biggest contributors to sodium, sugar or saturated/trans fat in national diets; food categories that make the largest contribution to the total dietary pattern; and/or food categories that align with national or international nutrition-related policies or strategies [70].

Benchmarks

Results suggested that there was generally strong support to apply benchmarking criteria to assess and communicate the extent to which the results from food environment monitoring indicate favorable or unfavorable conditions from a health perspective. For example, in Australia’s Food Environment Dashboard [18], indicators were assessed against set criteria to assist with the interpretation of data with ‘green’ used to indicate where food environments were likely to be promoting health, ‘amber’ used to indicate areas needing further improvement to promote health, and ‘red’ used to indicate aspects that were unhealthy/not supportive of health. In addition, there was support for benchmarks to be used to indicate trends over time (e.g., improving, unchanged, worsening). Benchmarks for each of the indicators were not proposed at this stage.