PROTOCOL: Impact of the food environment on diet-related health outcomes in school-age children and adolescents in low- and middle-income countries: a systematic review.

Abstract

Current estimates indicate that some form of malnutrition (undernutrition, overweight or obesity, and poor dietary habits) affects one in three people worldwide (International Food Policy Research Institute, 2016), and the 2016 Global Burden of Disease study has placed poor dietary habits as one of the leading risk factors for mortality globally (Collaborators GBDRF, 2017). Over the past decade, there has been great momentum around addressing malnutrition in all its forms and commitment to actions that can accelerate progress to reduce its associated burden of morbidity and mortality. In 2012, the World Health Assembly (WHA) adopted the 2025 Global Targets for Maternal, Infant and Young Child Nutrition and in 2013, WHA adopted targets for non-communicable diseases (NCDs), including several nutrition-relevant targets (International Food Policy Research Institute, 2016). More recently, the United Nations elevated its efforts through a global declaration of 17 Sustainable Development Goals (SDGs), where at least 12 of the 17 goals feature indicators relevant to nutrition. In line with these targets, the decade of 2016-2025 has been declared the Decade of Action on Nutrition (International Food Policy Research Institute, 2016). To this end, prioritizing critical actions to address school-age children and adolescent nutrition, is necessary to achieve these milestones. At the forefront of malnutrition and poor dietary intake is the food system. According to the Food and Agriculture Organization (FAO) High Level Panel of Global Food and Nutrition Security, the food system is defined as ‘a system that embraces all the elements (environment, people, inputs, processes, infrastructure, institutions, markets and trade) and activities that relate to the production, processing, distribution and marketing, preparation and consumption of food and the outputs of these activities, including socio-economic and environmental outcomes' (High Level Panel of Experts, 2017). Importantly, this group identified three food system typologies (i.e. traditional, mixed and modern), based on distinct inputs (natural resources, human capital, physical capital, agriculture and food technology), outputs (food purchasing patterns, diet, health and environmental sustainability) and processes (food production, supply chains and the food environment) Table 1. The transition from traditional to industrial food systems has been linked to urbanization, policy liberalization, agricultural productivity and income growth. In addition, the Global Nutrition Report, (International Food Policy Research Institute, 2015) defined two additional food system typologies (emerging and transitioning food systems), which are variations of the mixed food system, often observed in low- and middle-income countries (LMICs). Importantly, multiple types of food systems, and their associated food supply chains and food environments can co-exist within a single country simultaneously. Within traditional (or rural) food systems, there is a greater proportion of informal food markets (i.e. wet markets, mobile street vendors), compared to formal food outlets, as food is produced locally by rural smallholder farmers, is minimally processed and dependent on seasons. Moreover, there is minimal food promotion through advertising, limited food quality control and increase food affordability as a significant portion of monthly household expenditure is allocated to food. Examples of countries with such food system type are: Bangladesh, Ethiopia, Indonesia, Nepal, Senegal and Zimbabwe. Emerging food systems are characterized by low urbanization and agricultural productivity, though there is still reliance on staple foods and household food budgets are considered moderate to high. Examples of countries with such food system type are: Cameroon, China, Honduras, Pakistan, Namibia, Philippines and Thailand. In transitioning food systems, there is a shift towards moderate agricultural productivity and lower dependence on staples as compared to emerging food systems. As well, household food budgets are still considered high compared to mixed food systems. Country examples from this food system type are: Brazil, Ecuador, Guyana, Malaysia, Mauritius, Russia, Suriname and Ukraine. In mixed food systems, there is improved infrastructure to support formal markets, though informal markets are still prominent. Furthermore, there is increased diversity of foods available, including staples, fresh produce, animal-sourced foods and highly-processed packaged foods. There is increased food promotion through advertisements and weakly implemented regulations on food safety and quality. Examples of countries with such food system type are: Barbados, Bulgaria, Estonia, Germany, Hungary, Italy and Switzerland. In modern (or industrial) food systems, the highly urbanized environment promotes increased access and reliance on highly diverse and dense formal food markets (i.e. large supermarkets, hypermarkets, fast-food and fine dining restaurants). A high level of food information and promotion exists through various media platforms, including marketing and labelling on food packages, advertisements and public health campaigns. Economically, household food expenditure is reduced, though fresh produce and animal-sourced foods are more expensive than staples and packaged products. Examples of countries with such food system type are: Australia, Canada, Denmark, Lebanon, Republic of Korea, Sweden and the US. In 2012, a global network of public-interest organisations and researchers aiming to monitor, benchmark and support public and private sector actions to create healthy food environments and reduce obesity and NCDs was established. This network, named The International Network for Food and Obesity/Non-Communicable Diseases Research, Monitoring and Action Support (INFORMAS), defines the food environment as ‘the collective physical, economic, policy and sociocultural surroundings, opportunities and conditions that influence people's food and beverage choices and nutritional status' (Swinburn et al., 2013). In the same light, the FAO High Level Panel of Experts for Food Security and Nutrition characterizes the food environment as ‘food entry points’ (i.e. the physical spaces where food is obtained; the built environment that allows consumers to access these spaces); personal determinants of food choices (including income, education, values, skills, etc.); and the political, social and cultural norms that underlie these interactions. The key elements of the food environment that influence food choices, food acceptability and diets are: physical and economic access to food (proximity and affordability); food promotion, advertising and information; and food quality and safety (High Level Panel of Experts, 2017). It is important to note that there are several proposed conceptual frameworks with a global lens available from Health Canada, Centre for Disease Control (CDC), USA, the FAO High Level Panel of Experts for Food Security and Nutrition, and the Global Panel on Agriculture and Food Systems for Nutrition (Health Canada, 2013; CDC, 2013; Development Initiatives, 2017; High Level Panel of Experts, 2017; Global Panel on Agriculture and Food Systems, 2016), which borrow from the early work of Penchansky and Thomas (1981). These food environment models have been conceptualised to possess five dimensions that impact food choice and dietary intake: availability, spatial accessibility, affordability, accommodation and acceptability. Others have proposed a food environment model with four domains: community, consumer, organisational and information (Glanz, Sallis, Saelens and Frank, 2005) or two domains: the external food environment and the personal food environment (Turner et al., 2017). Importantly, this latter model differs from previous models as it aims to position exogenous dimensions at a structural level (i.e. vendor and product properties, marketing and regulation, availability and prices) with endogenous dimensions (i.e. accessibility, affordability, convenience and desirability) at the individual level. Though these models have laid the foundation within the food environment space, none are tailored to the complexity of LMICs or are specific to school-age children and adolescents. Rapid globalisation and urbanisation in LMICs has incited major changes to the landscape of food. This is observed upstream with trade liberalization and foreign direct investments by transnational food and beverage companies, including grocery and fast food retailers. As well, transformations in food processing and modernised supply chains have caused a systemic shift in the composition and packaging of foods, as well as the ease of acquiring these foods. In parallel, traditional domestic channels of food acquisition, such as through informal unregistered vendors (wet markets and street vendors) still account for a portion of the market share in LMICs. This dynamic and opportunistic environment is in a constant flux, posing a challenge in accurately estimating food availability, accessibility and affordability. Downstream, food preferences, dietary patterns, and habits have been negatively impacted as economical, convenience foods are inexpensive compared to healthy options. For school-age children and adolescents, this corresponds to increased reliance on unhealthy, nutrient-poor ultra-processed foods and sugar-sweetened beverages, as well as, increased snacking and eating away-from-home. Taken together, energy imbalances have led to population-wide nutrition transition with continued burden of stunting and wasting. This double burden of malnutrition has been widely studied in LMICs, though there have been insufficient efforts to study the association to the food environment (Popkin and Reardon, 2018; Baker and Friel, 2016; International Food Policy Research Institute, 2017). Our conceptual framework aims to fill the current gap, by focusing on extrinsic factors within the food environment that influence school-age children and adolescents in LMICs. Importantly, Figure 1 is not a definitive model of existing evidence. Rather, it is a proposed logic model to help guide the implementation of this review. It is believed that the food environment is a complex adaptive system, influenced by the wider food system, whereby various industries and actors operate interdependently and adaptively, and their interaction is often shaped through spatial and temporal complexity. Though others have depicted specific micro-food environments (i.e. home, school, consumer food environments or external versus personal food environments), we are not attempting to categorise these micro-food environments. We acknowledge these micro-food environments exist and interact with one another, especially in terms of the translation of food and nutrition knowledge, perceptions and desirability. For example, education on healthy food choices learned in the school food environment would influence an adolescent in the market food environment. Conceptual Framework of the Food Environment in School-age Children and Adolescents in LMICs As seen in Figure 1, it is believed that the food environment innately possesses the following dimensions: Accessibility: This refers to proximity, density and presence of retailers relative to individuals or organisations. Availability: This refers to both retailers and product availability within a given context. Pricing: From a market perspective, this is indicative of the market price of products. From a household or individual perspective, this equates to affordability (purchasing power). Our review will focus on market price, as this is most relevant to environment level interventions. Promotion: Promotional advertising, marketing and branding directed at individuals. These dimensions and their associated measures are useful for understanding the food environment, especially in urbanised settings of LMICs. As well, the intersection of these dimensions directly or indirectly involves primary and secondary sectors such as agriculture, food manufacturing and food processing. We recognise policy is an important component of the food environment, however, we have not included policy as its own dimension. This is because we believe policies, regulations and guidelines (i.e. national, regional, local) underlie all dimensions of the food environment. Likewise, we also acknowledge that conceptually food quality and food safety (including food handling, sanitation and hygiene) are integral components of the food environment. However, food safety and food quality are in a constant flux, making it difficult to establish metrics of the food environment, particularly for school-age children and adolescents. For this reason, we have chosen to place food quality and food safety under the food system and supply chain in our model (Figure 1). Additionally, it is hypothesised that the food environment is linked to consumer behaviour and purchasing, which includes knowledge, awareness, attitudes and preferences. As seen in Figure 1, consumer behaviour and purchasing are extrinsic to the food environment, and subsequently affect consumption (preparation & transformation, quantity, quality, diversity and safety) and ultimately, diet-related health outcomes in school-age children and adolescents in LMICs. We acknowledge the limitation in developing a single conceptual framework for both school-age children and adolescents. School-age children differ from adolescents in that, school-age children have limited autonomy in the acquisition of food. Instead, parents and caretakers influence their consumption by making decisions on what is purchased for the home, what is eaten outside of the home and setting limits on the quantity of food consumed. On the other hand, adolescents have the ability to make their own decisions and purchase on their own. However, the focus of our review is to understand the impact of food environment interventions on diet-related health outcomes in school-age children and adolescents, where food purchasing is beyond the scope of what we have considered to be intrinsic properties of the food environment, whether this is done by parents, providers or adolescents. With global increases in obesity prevalence, especially in LMICs, there are various interventions that have been categorised and implemented with intent to improve the food environment (Lobstein et al., 2015). These interventions are often best evaluated in randomised controlled trials (RCTs) whereby the efficacy and degree to which interventions produce an impact of change are tested under optimally controlled conditions, minimizing bias and confounding factors (Marchand, Stice, Rohde and Becker, 2011). However, there are interventions (e.g. policy-based or food-retail interventions) that cannot be so rigorously randomised and controlled, and as such, the RCT is not an appropriate study design. In these cases, interrupted-time-series (ITS) and quasi-experimental designs are used for impact evaluation (Taillie et al., 2017). ITS and quasi-experimental studies evaluate a change or exposure that is outside the control of the researchers, proving more difficult to manipulate the exposure and randomise participants into intervention and control groups (Craig et al., 2012). These limitations present several challenges, such as ensuring strong internal and external validity, minimizing potential unmeasured confounders, mitigating pre-existing baseline differences in the outcome between groups, and being flexible with uncontrolled timing and seasonality, etc. Despite this, ITS and quasi-experimental studies offer greater generalisability than RCTs. As well, they provide critical insight and information about real-world settings, such as how feasible the implementation of an intervention is in situations where a considerable lack of control exists (Taillie et al., 2017). In addition, food environment interventions can be classified as preventive or management-based (Swinburn and Egger, 2002; Harvard, 2017). Preventive interventions intend to improve the food environment by minimizing an individual's exposure to unhealthy environments (e.g. limiting unhealthy food options in school cafeterias) or by preventing and mitigating poor health with proactive strategies (e.g. with programs targeting increased physical activity and minimized television or ‘sitting time’) (Harvard, 2017). Management-based interventions involve long-term strategies to improve population health, such as reducing the population's prevalence of obesity. Examples of these interventions in school-age children and adolescents include public education campaigns and policy interventions on taxation and television advertising of unhealthy foods (Swinburn and Egger, 2002). Interventions can also be defined as environmental or behavioural (Roberto et al., 2015; World Cancer Research Fund, 2017), whereby environmental interventions target the various spheres of an individual's environment that influence their choice in food and beverages. These include the built and natural physical environments, legal and political environments, socio-economic, and cultural environments. Examples of environmental interventions specific to school-age children and adolescents include school-feeding and school-meal programs or policies that incentivise vendors to provide healthier retail environments. Behavioural interventions place a greater focus on the individual, targeting their knowledge, attitudes, perceptions, preferences, and abilities in food and beverage choice and consumption patterns. Several theories and models of behavioural change such as the theory of planned behaviour, diffusion of innovation theory, the social cognitive theory, and the health belief model, offer insight into the behavioural relationship between food environment interventions and diet-related health outcomes. Behavioural interventions targeting school-age children and adolescents include nutrition counselling and education for educators and students, physical activity programs, and public awareness campaigns (World Cancer Research Fund, 2017). It is acknowledged that overlap in interventional effects may occur whereby interventions, such as nutritional labelling policies, may alter both the environment and behaviour of the individual or consumer. Likewise, there may be instances where interventions could be categorised as both preventive (and or management-based), as well as environmental (and or behavioural). This behavioural-environmental approach is consistent with other models and frameworks designed to identify and categorise distinct areas of policy and intervention actions. Similarly, the NOURISHING framework, created by the World Cancer Research Fund International (WCRFI) identifies ten specific areas where policy and interventions can be used to improve food environments (World Cancer Research Fund, 2017; Hawkes, Jewell and Allen, 2013). While designed to be globally applicable, WCRFI recognises that the framework's key areas would need to be adapted to the specific context and population of different countries. Further, interventions can be classified by their level of implementation. Swinburn and Egger, 2002 suggest that interventions can be categorised into two main groups. A ‘settings-based’ intervention would focus primarily on micro-environments (micro-level) such as schools, workplaces and neighbourhoods, whereas ‘sector-based’ interventions would target macro- and meso-level environments such as the food industry and national supply chain (Swinburn and Egger, 2002). A simplified logic model from the CDC distinguishes four levels of implementation: policy-level, community-level, organisational-level and the individual-level. The policy-level includes interventions that influence legislation and policies that have a macro-level exposure, such as food labelling laws and supply chain regulation. Community-level interventions focus on increasing the awareness of the general population, while the organisational-level interventions are intended to influence organisations and systems such as health-care systems, industry players and community-based organisations. Lastly, the individual-level defines interventions as those that improve and enhance the knowledge, skills, attitudes, abilities and preferences of the individual (CDC, 2013). Specific to our review, school-aged children and adolescent engage with the food environment in various ways. They interact through micro-food environments such as the home, school and workplace, and consumer and retail spheres. While our review does not concern interventions implemented at the household level, we acknowledge that dietary intake in school-aged children and adolescents is affected by factors in which they have limited control. Adolescents, but especially children, may not have full autonomy and choice over what is made available to them and what they consume. These include: what is available in their homes, what their parents or providers purchase, and what the household can afford. The school and workplace can be important food environments for children and adolescents. At school and work, their food choices and consumption are influenced by what is available and accessible (e.g. in cafeterias, vending machines, food stalls on or near campus). Thus, organisational-level interventions that change the physical environment such as improving the quality of foods available at lunchtime could have an impact on the dietary intake and consumption, and thus health outcomes, of school aged children and adolescents. We recognise also that not all children and adolescents attend school, especially in LMICs. However, we aim to capture these children and adolescents through community-level interventions or organisational-level interventions specific to the workplace (for those children and adolescents who are in the workforce). One limitation we acknowledge is that some children and adolescents who do not attend school might be better captured at the household level. However, the food environment at the household level is out of this review's scope as it can be affected by other factors, such as provider's preference and household income, which are not specific to the child or adolescent him- or herself. This review will include environmental interventions, both preventive and management-based, targeted at school-age children and adolescents in LMICs, which are implemented through market, community and organisational platforms (Figure 2). For the purposes of this review, we will consider any policy, regulation or guideline that affects the food environment through each of the three platforms, as an intervention. For more details on specific intervention types, please see the section “Types of interventions”. Impact of Food Environment Interventions on Diet-related Health Outcomes in School-age Children and Adolescents in LMICs Today, many food environments promote access to unhealthy foods, influencing individuals' preferences and demands for foods with poor nutritional quality. This has contributed to the rise in obesity and non-communicable diseases globally, especially in low- and middle-income countries (Roberto et al., 2015). In response, food environment interventions are designed and implemented to improve these environments, which ultimately interact with other factors, affecting people's food and beverage perceptions, choices and intake. Many frameworks and models have conceptualised this complex relationship between the individual and the host of environmental and health factors that contribute to his/her dietary behaviour and nutritional status (Health Canada, 2013; Development Initiatives, 2017; High Level Panel of Experts, 2017). The intended effects of these interventions are thus to change and improve nutritional intake and dietary behaviour in a global effort to reduce the rise in obesity and non-communicable diseases. To illustrate these effects, we developed a logic model that depicts how different types of interventions, at various levels of implementation and influence, address food and nutrition insecurity and thus, affect the diet and health-related outcomes at the individual level. Environmental interventions, such as those categorised in the NOURISHING framework, would primarily work at the policy and community levels of implementation within the CDC's logic model, or at the macro-level as defined by Swinburn (2002) (CDC, 2013). In our model, policies, regulations or guidelines at every level (i.e. national, regional and local) affect the food environment through all three platforms (market, community and organisational). These include macro-level legislation on taxation of unhealthy food, regulation of imports and exports, and policies on nutritional labelling and advertising, each of which would have an effect on the economy, various industries and society at-large. The direct effects of these interventions would include food price regulation, taxation on unhealthy food and beverage, improved labelling and informative food packaging, and decreased airtime for TV advertisements of unhealthy food during children's programs. Examples of micro-level legislation or policy would include school-based policies that limit the selling of sugar-sweetened beverages in a school cafeteria. Similar to other models and frameworks cited above, market-level interventions in our logic model include those that affect change within the formal and informal retail space (i.e. changing the placement of products within a store, changing the diversity of products available for purchase or store-directed discounts on healthy food.) Community-level interventions affect widespread changes in physical infrastructure within the community and public health. These interventions include public health campaigns promoting healthy food and nutrition, transportation infrastructure improvements to provide better access and availability of healthy food options. Direct effects of these interventions include increasing the number of food outlets in the neighbourhood and providing subsidies for healthier options in stores. At the organisational-level, interventions aim to influence institutions, such as work and school systems. These involve regulating the availability of food in work and school environments, such as increasing the number of healthy options at school canteens and cafeterias, and improving the physical environment, such as the implementation of water fountains in schools and removing advertisements (Figure 2). The combination of these direct effects from the market-, community- and organisational-level interventions work to improve the acquisition of healthier food and beverages for both households and individuals, in terms of both quantity and quality. Better availability and accessibility to healthy food and beverages allows children, adolescents and their families to make healthier food choices and improve their food intake. These intermediate effects lead to short-term outcomes, such as improved diet and eating practices, reduced overweight and obesity, and long-term outcomes, such as improved life expectancy and improved work and education productivity. As mentioned above, we acknowledge that school-aged children and adolescents may experience the food environment differently as they transition from child to adolescent, or adolescent to adult. As children get older, they gain independence, autonomy and a greater range of choice in their acquisition of food and drink. This would affect their consumption and subsequently, their health outcomes. We also acknowledge that different environments might provide greater autonomy and choice. For example, a child eating lunch at school might have unhealthier options provided to them than at home where their parents are able to better control their diets. However, the focus of our review is to understand the impact of interventions outside of the home environment, apart from the strong influence that providers and caretakers might have on the diets of children and adolescents. Understanding if and how food environment interventions impact diet-related health outcomes in school-age children and adolescents in LMICs is critical for offering strategic recommendations for food environment improvement and encourages advocacy. Ultimately, this review will contribute to research within this field, with the overarching goal to improve health outcomes through recommended interventions. As mentioned, food environments in LMICs differ from those in high-income countries in part due to the lack of appropriate infrastructure to access and distribute healthy foods. The lack of registered and formal outlets complicates measuring food environments in LMICs through geospatial and observational methodologies. As such, we acknowledge there are limitations associated with synthesizing information on food environments in LMICs, as food environments are not static. Despite this, we will attempt to capture information on the transitory nature of the food environment. To date, there is paucity of high-quality summative evidence to link the food environment to diet-related health outcomes in school-age children and adolescents in LMICs. In fact, a lack of consistency in the definition of a food environment has presented challenges in synthesizing data. Currently, the only attempt to develop standard indicators of the food environment, relates to policy indicators. The Healthy Food Environment Policy Index (Food-EPI), was developed by the INFORMAS group to assess the level of policy implementation against international best practices. Food-EPI consists of seven domains in relation to policy (food composition, labelling, marketing, provision, retail, prices and trade) and six domains in relation to infrastr