PROTOCOL: Evidence and gap map protocol: Understanding pathways between agriculture and nutrition: An evidence and gap map of tools, metrics and methods developed and applied in the last 10 years

The global food price crises of 2007–2008 and 2010–2011 drew attention to the need for addressing the underlying determinants of malnutrition in low‐ and middle‐income countries (LMICs; Brinkman, de Pee, Sanogo, Subran, & Bloem, 2010; Webb, 2010). Specifically, as the primary source of food and income in LMICs, agriculture received renewed focus. Making agriculture work for nutrition—nutrition‐sensitive agriculture—has climbed the international development agenda (Ruel, Alderman, & Maternal and Child Nutrition Study Group, 2013). More recently, given the sharp increase in diet‐related chronic diseases underpinned by overweight and obesity in LMICs and the threats of climate change to diets, attention has expanded to leverage food systems to optimize nutrition, health and environmental outcomes (Johnston, Fanzo, & Cogill, 2014). Donors, researchers and implementers mobilized research agendas to invest in understanding how to strengthen agriculture and food systems to realize nutrition outcomes sustainably. Progress in this field in the last decade included three key developments:


Data Expansion Project [INDDEX], 2018; Sustainable and Healthy
Food Systems [SHEFs], 2018) have built on existing theoretical underpinnings and have helped to refine hypothesized pathways, illuminating additional aspects and dynamics between agriculture or food systems and nutrition outcomes, such as food environments, environmental factors and food safety. The Food and Agriculture Organization (FAO) of the United Nations (UN) adopted the High-Level Panel of Experts (HPLE) definition of a food system: "all the elements (environment, people, inputs, processes, infrastructures, institutions, etc.) and activities that relate to the production, processing, distribution, preparation and consumption of food, and the output of these activities, including socio-economic and environmental outcomes" (HLPE, 2017).
Agriculture and health are part of the broader milieu of a food system. As such, there has been a proliferation of innovations in programme design and implementation, as well as in metrics and methods and their application. While the body of evidence on effectiveness of food systems to improve nutrition has been recently summarized (Ruel et al., 2018), the portfolio of new methods and metrics has not. It is now necessary to take stock of these developments and plan for the future to support the production of effective and relevant research.

| Scope of the evidence and gap maps (EGM)
Recognition of the multiple pathways through which nutrition impact is achieved brought about new conceptual frameworks, and along with them, new thinking in how to measure the complexity and dynamism within these systems. The innovations that emerged range from new technology to new indices to the application of methods from other fields. New metrics and methods have been developed throughout the pathways (household production, decision-making, income, etc.) linking agriculture and nutrition.
In a standard effectiveness map, the row headings are interventions, and the column headings outcomes. In this map, those thematic pathways or domains will be considered the "interventions" through which nutrition is improved. We consider our "outcome" to be tools, metrics or methodological innovations, which are the columns of this map. Some innovations have been widely adopted across settings, and others are still in development. Therefore, each example innovation will be mapped using the studies that pertain to the innovation. Innovations have taken place at every level of measurement (individual, household, national, etc.) and correspond to certain cross-cutting themes. These additional aspects will be coded internally on the map.
As an example of technology application at an individual and household level, researchers have utilized accelerometers to measure calorie expenditure in new ways to address intra-household food allocation and gender roles (Zanello, Srinivasan, & Nkegbe, 2017). At a community level, researchers have employed wearable cameras and GIS technology to map changing food environments in urban areas (Schrempft, van Jaarsveld, & Fisher, 2017). A methodological innovation at a sub-national level has been to use Bayesian theory and decision-analysis for making policy that affects nutrition (Yet et al., 2016). More thorough data collection and new indices to capture prices of nutritious foods in markets at the regional level have led to better estimates of cost of nutritious diets in Ghana . New innovation in this field also includes tools to conceptualize and operationalize food systems, including how to frame and measure cost-effectiveness of complex interventions, which have a range of outcomes . In this EGM, the columns will be types of innovation or novel application.
The aim of the gap map is to articulate and summarize the innovation in tools, metrics and methods that have been created and applied to understand food systems and agriculture-nutrition linkages in the last ten years. We have chosen the ten-year period based on the focus on and funding for agriculture-nutrition linkages that emerged following the global food price crisis in 2007-2008, as well as wanting to focus on new innovations, which, by definition, would no longer be novel if developed more than a decade ago. We also aim to highlight gaps and opportunities for future development.

| Conceptual framework of the EGM
Although the intervention-outcome framework is most common for maps on effectiveness studies, this framework will be organized differently. We will take an approach that considers tools, metrics and methods (types of innovation/application in the columns) for agriculture-food systems-nutrition research (thematic domain in the rows). The map will be organized around a combination of conceptual frameworks that include the definition of food systems offered by the HPLE report on nutrition (HLPE, 2017), predefined pathways to improved nutrition (Global Panel, 2015;Hawkes et al., 2012;Herforth, Nicolò, Veillerette, & Dufour, 2016;Kadiyala et al., 2014), as well as additional themes that have been identified as more research is being undertaken on this topic (Grace et al., 2018;SHEFs, 2018) These conceptual frameworks (illustrated in Appendix A) overlap a great deal. For instance, each highlights the role of on-farm production as a means for direct consumption as well as a potential income source, both which influence food availability and diet quality, and thus contribute to nutrition outcomes. Frameworks 1 (Kadiyala et al., 2014), 2 (Hawkes et al., 2012) and 3 (Herforth et al., 2016) are very similar-in fact the most substantive difference is the visual organization of components. These frameworks include aspects of women's time, income and employment. The same three highlight interacting aspects of care, education or knowledge, as well as overall health as drivers of nutrition outcomes.
Each framework also has differences, both in how it is visualized and the content. Each represents "indirect" determinants of nutrition outcomes, such as the role of climate, the environment, policy, governance and culture, inequity, and so forth, but some are shown as an outside layer of influence, whereas some of these are considered within the central framework. For instance, Framework 4 shows the interactions between environment or sustainability aspects and food production, highlighting that human health should always be balanced against planetary health, since they are symbiotic in the long run (Tuomisto et al., 2017). Several highlight important domains that are not equally represented on the other. Frameworks 5  and 6 (Global Panel, 2015) propose the most current thinking about markets and the economic role of nutrition.
These support the idea that production will lead to consumption only where, when and for whom markets are missing. The Global Panel Metrics and Methods Framework (Framework 6) puts the food environment as the central milieu into which other dynamics feed, and diet diversity, adequacy and safety as general by-products of that food environment. In contrast, Framework 2 (Hawkes et al., 2012) specifically articulates the subsequent layers of the food environment that progressively lead to nutritional status. We will use all frameworks generally to ensure that the EGM is comprehensive and that the domains within conceptual pathways in agriculture to nutrition literature are represented and categorized logically, while maintaining iterative methods of refining the domains based on search results.

| Why it is important to develop the EGM?
Governments, non-government donors, implementing agencies and academia have all made significant investments, both intellectually and financially, in improving agriculture or food systems for nutrition outcomes. This investment has gone beyond scholarship and documentation and taken the form of application and innovation of tools, metrics and methods at every level. Stakeholders have called for a synthesis project on this topic in order to visualize the current portfolio of these developments, strategically plan the next wave of investment and shape the next generation of agriculture-nutrition research.

| Existing EGMs and/or relevant systematic reviews
There are no current gap maps on the topic of metrics and methods on the topic of agriculture and food systems for nutrition (or to improve nutrition outcomes). Some mapping exercises have been undertaken on pathways between agriculture and nutrition, namely the 2012 LCIRAH "Current and planned research on agriculture for improved nutrition: a mapping and a gap analysis", which led the way to the IMMANA programme (IMMANA, 2018). The FAO Compendium of nutrition-sensitive indicators also summarizes the most wellestablished indicators on the subject (Herforth et al., 2016), but does not fully capture innovative tools and methodologies, as well as metrics that are in development currently. Furthermore, to our knowledge, none of these synthesis projects have been systematic or published as a formal EGM, and overall there have been no EGM of tools, metrics, or methodologies; rather most existing gap maps focus on effectiveness studies.

| OBJECTIVES
The main objective of this EGM is to guide funders and researchers in the most promising areas of innovation within the study of food systems or agriculture to nutrition pathways, and demonstrate their phase of development and other thematic trends. We also will be able to demonstrate where there are gaps in existing innovative tools, metrics and methods that correspond to key domains identified in these conceptual frameworks. Empty cells in the map will indicate where no new methods, metrics or tools either exist or have been developed in the last decade within those domains. Furthermore, we intend that this EGM will then be used to shape future investments in this field, both by pursuing opportunities to take the most promising developments to the next level, and focusing attention on where there are gaps in available tools, metrics and methods.
A secondary objective of this EGM is to identify trends in investigation and application that would be suitable for further synthesis.

| Defining EGM
All previous EGMs published to date have been compilations of effectiveness studies, therefore this EGM will be novel in many aspects. We are not aware of any protocols or published EGMs on which to model this project. There are several synthesis reports on this topic (Hawkes et al., 2012;Herforth et al., 2016), but as mentioned previously, none of them are current, systematic or are formal EGMs. We will use published, well-established conceptual frameworks to define thematic domains of agriculture-to-nutrition in order to categorize the identified tools, metrics and methods. Therefore, our "intervention" will be each broad domain on the food systems or agriculture to nutrition pathway. The columns of our map, (outcome in effectiveness maps) will be each item of innovation (tool, metric or method) that has been developed or applied to capture or measure these domains.
While "innovation" is hard to categorically define, we will take a pragmatic, data-driven approach to selecting what is new, novel or innovative in this dynamic field. By a data-driven approach, we mean that we will use existing knowledge and data to inform the process, both through the published record and expert consultation. We will adopt several strategies and guidelines in selecting new, novel or innovative tools, metrics or methods: • Limit the search to work published after 2008.
• Identify completely new tools, metrics or methods that were introduced after 2008 with no previous iterations.
• Identify tools, metrics or methods that existed prior to 2008 but have been significantly revised or modified since. As a "significant" change or modification is difficult to determine, SPARLING ET AL.

| 3 of 18
we will rely on the group or authors' own assertions and explanations, and make an expert judgement as a group when unclear.
• Identify new, novel or innovative applications of existing tools and methods. This will mostly entail applying these cross-disciplinarily.
This will be the most difficult aspect of "newness" to define, and we therefore will also rely on the authors' description and justification, and secondarily make a collective expert decision.
Because some of the tools, metrics and methods are in their infancy, while others are globally adopted and have become standard practice, we will code each study based on the current (e.g. in 2019) stage of development of the innovation. We will further code and categorize innovations by several thematic filters (e.g., gender, equity, economics, technology, private-sector engagement, conflict or political fragility), geographical application, and level of measurement.

| EGM framework
The "intervention" (rows), will be defined using agriculture-tonutrition conceptual frameworks mentioned previously, divided into "domains" of influence on agriculture and food systems or nutrition, such as household or on-farm production, food policy and governance, or food environments and markets. The columns, or categories of innovations/applications, will be grouped by different classifications of tools (technology application and instruments to capture data on a range of agriculture-nutrition topics), metrics (new indices and measures to quantify agriculturenutrition linkages) and methods (research design and analytical approaches applied to agriculture and nutrition research). We will code each study related to an item and group the items iteratively once all items have been mapped. We have chosen to do this since some well-adopted innovations will have many papers that use the tool, metric or method, while others will have only a few, and some will apply the tool, metric or method in different ways. Users will be able to see each individual item as well as grouping by tool, metric or method.
Traditional EGMs include a quality assessment of each item, such as a risk of bias rating, which are not designed to evaluate tools, methods or metrics. In place of a quality assessment, this EGM will summarize the stage of development or application (explained below). We will add filter codes for certain cross-cutting themes such as gender or private-sector engagement. It will also categorize the measurement level (individual, household, district, national, etc.), and setting or geographical application (Asia, Africa, global, etc.). We also may add other filters as the search progresses. The framework structure is shown in Appendix B.

| Population
This map will only include tools, metrics and methods that have been applied to agriculture-nutrition pathways in any country at any level: individual, household, community or district, sub-national, national and global.

| Intervention
The problem we are considering is any domain that exists on the conceptual pathway between agriculture and/or food systems and nutrition outcomes. These domains have been grouped (through using frameworks and extensive rounds of expert consultation) by broad themes around food production, food safety, value-chains, markets and food environments, food policy and governance, environment and climate, among others. We have organized the domains around broad themes in order to group items with minimal double-or triple-coding, but we do envision that some items will appear in more than one domain. Whenever possible we will select a "primary" domain and use the filters and codes to indicate other aspects of the tools that are cross-cutting, such as gender, technology or economics.

The first column lists 12 broad domains, and the second column
are examples of what types of work would fit in to these domains. All included tools, metrics or methods must explicitly relate to either agriculture/food systems, or to nutrition. Any tools, metrics and methods that are not related to either agriculture/food systems or nutrition will be excluded. Most of the domains could be measured at various levels (individual, national, global, etc.). We will not differentiate in the domains, but rather in the internal coding of each tool, metric or method innovation. As the initial results are identified, we may refine these domains and add sub-domains using an iterative methodology.
Specifically, we propose the following overarching categories based on the conceptual frameworks discussed previously ( Table 1).
The domains of food safety, food environments and economic evaluations each have systematic reviews on methods and metrics.
We will include the items that are included in those reviews and also use these items to test the framework.

| Outcomes
The primary "outcome", of the gap map (i.e., the columns in the map) is the innovative item (tool, metric or method) created and applied to studying and describing the broad agriculture to nutrition pathways.
We will define "tool" as is a vehicle or an aid to collect information and data (e.g., a survey module to collect data required to compute an index or a piece of technology). "Metrics" will be defined as the parameters (measures) or indices used for measurement, comparison or tracking performance (e.g., disability adjusted life years; household dietary diversity score and Women's Empowerment in Agriculture Index [WEAI]). We define "method" as the process and approach involved in a systematic inquiry of relationships between agriculture, nutrition and health and generally refer to study design or application of an analytical method to this topic (e.g., impact evaluations using various types of counterfactuals, pathway analyses, decision analyses). Several types of "innovation" are described above. We will identify each tool metric or method and group them together once identified. Some methods or tools will have slight variation in their application or analysis, but we will group these logically together as an item if the construction of the item is similar. We will include all of the studies on the map.
Each item might have multiple innovative components that fit distinctly within the categories of tools, metrics or methods. We propose that these will not be exclusive, rather the appropriate component of a paper or project will be categorized accordingly, and the total number of items summarized separately. In order to determine what is substantial enough for inclusion as a separate item (rather than a much smaller exploratory innovation and application), we will use the published or unpublished study objectives, considering only primary and secondary objectives. We will group tools, metrics and methods into the following categories, with some illustrative examples in the right-hand column ( Diet, nutrition and health Nutrition KAP, norms and behaviours, food consumption, nutritional status indicators (e.g., energy balance, micronutrient status, anthropometry); NCDs; food productionrelated labour burden, nutrition-related child illness; diet quality; bioavailability a These domains have existing reviews that summarize the topic. SPARLING ET AL.

| 5 of 18
• In grey literature, published and/or peer-reviewed sources, as well as known projects from grant databases that describe tools, metrics and methods in development.
• In any country.
• Developed for the purpose of and/or used to quantify or qualify a potential interaction between agriculture or food systems and nutrition. This will be defined by the tool, metric or method being related to either the broader agriculture field, food systems as defined in the High-level Panel of Experts (HPLE) Report or the broader nutrition field, since several methods and metrics used to study this interaction are only explicitly linked to one side of the pathway.
○ Nutrition or nutritional proxies on the outcome side of the pathway will be considered, including all forms of malnutrition • Tools, metrics or methods developed or applied prior to 2008.
• Projects with no reporting in English.
• In-vivo laboratory animal studies. If the subjects are animals for agricultural production, livelihoods or consumption, they will be considered. If the animals are subjects as a proxy for humans or models for general interests (i.e., if the animals aren't to sell or eat), they will be excluded.  or not mentioned in relationship to agriculture/ecology nexus, they will be excluded.
• If the paper is related to nutrition and nutritional proxies, food supplementation for communicable diseases (e.g., TB, HIV), special groups or niche populations (such as hospital patients, athletes, etc.) will all be excluded.

| Types of study designs
The gap map will include primary research of any design, as certain study designs may in fact be an innovative method of application to study the agriculture-nutrition linkages. Reviews will be excluded. newly developed impact evaluation methodology or an impact evaluation using well-established study designs but using new tools or metrics. We will use the four considerations described above to determine what is considered "innovative" or "new" (Table 3).

| Treatment of qualitative research
We will include qualitative research in the EGM as long as it fits within the agriculture/food systems to nutrition framework and meets the inclusion and exclusion criteria. Projects could be undertaken as part of a mixed-methods project, or be entirely

| Types of settings
Any innovation or application of tools, metrics and methods taking place anywhere will be considered. They could be from a specific country or region, or they could be applied globally, with relevance to LMICs (analytical approaches, trade issues, etc.).

| Status of studies
We will include ongoing studies and projects, identified through expert consultation, interviews, grant databases and unpublished documentation.

| Search strategy and status of studies
We will employ a comprehensive literature, trial and project database search that includes electronic screening with search terms (listed in Appendix C), consultation with subject-matter experts (outlined in Appendix D). We will also search various project and research databases and key websites and backward-track citations in the bibliographies of key papers, both listed below.

| 7 of 18
The following databases will be searched electronically: -CAB Abstracts.
-Web of Science (seven databases).
The following websites will be searched: -Some included articles, studies and reports

| Screening and selection of studies
We will use a traditional method of two independent researchers to search and then screen (both on title/abstract and full text) the first 10% of items, with a third researcher (Howard White, Denny John or Thalia Sparling) providing a decision in the case of disagreement. For the remaining search results, we will use single screening with 5% randomly checked by a senior researcher.
We do not plan to use any automation or text-mining. The Campbell Collaboration, in collaboration with the IMMANA team, will hire and manage an experienced team to complete the initial search, screen titles and abstracts according to the inclusion and exclusion criteria, and further screen full-text publications. We will then collaboratively extract the data, code the projects and complete the coding for stage of innovation or application, which will replace a traditional quality assessment in this EGM.
3.6.9 | Data extraction, coding and management Besides the rows of "intervention" and columns of "outcomes" described previously, we propose to code for several other factors that will act as filters in the EGM. Some filters will have predefined categories:

• Stage of innovation or application (definition provided in Quality
Appraisal): • Concept development and pilot Other filters will be used to identify cross-cutting themes, which will not have categories: ○ Private-sector engagement ○ Economics (e.g., cost-benefit, cost-effectiveness, cost of diets) ○ Gender dynamics and parity (Could be about men or womenanything about gender dynamics, e.g., decision-making power over productive resources; employment in agriculture; self-care and health decision-making; social gender cohesion; women's mobility; leadership) ○ Equity (e.g., caste/ethnicity, economic, geographic, age) The filters without categories will be applied if the authors of the item specifically state the relationship of the item to that theme or the research item is clearly linked to a particular filter. The same research team in charge of the search and screening will complete the coding of these filters and categories, which will have been decided through consultation with subject experts and a piloting one grant portfolio, but may also be modified once search results are identified. We will screen and code articles using EPPI Reviewer 4.

| Quality appraisal
Traditionally, effectiveness studies included would be evaluated for risk of bias or overall quality. These quality assessments are not applicable when evaluating a tool, metric or methodology.
However, we recognize that some formal assessment and/or ranking of each item included (tool, method or metric) will be useful to readers. In this map, some tools, metrics and methods will be too new to be widely adopted and tested across settings. As they may still be in development, we will draw on the traditional epidemiological lens of indicator development as well as stages of innovation to create four categories of development. During indicator development, various aspects of validity and reliability are explored and tested, which are all equal components in creating a successful indicator (Frongillo, 1999). In the Stages of Innovation, there is a hierarchical progression through ideas, proof-of-concept, design and wider application (Imperial College Health Partners, 2018). The overarching components of each are described below (Table 4).
By using both the principles of indicator development as well as the progressive stages of innovation, we propose four categories of development or application: 1. Concept development and pilot: there is well-defined problem which leads to a need for the innovation and a pilot innovation developed, the innovation is well-grounded in an understanding of the phenomenon. to new settings will be coded as Stage 3. More than five items mentioning a tool, metric or method developed or applied in a novel way after 2008 will be grouped into a simple count and used as evidence of "widespread" application, or Stage 4 of innovation. In these cases, the first five key papers will be coded and included.

| Unit of analyses
The unit of analysis will be the item (tool, metric or method) that develops the innovation or application. There may be several components of innovation, which will populate the subsections of tools, metrics or methods. Therefore, the total number of studies or projects will be calculated separately from the items of innovation, as one project may be categorized as having an innovative tool that is distinct from an innovative metric, for example, and several studies may be associated with a single item. Items will only be included in this EGM if they state the innovative tool, metric or method as either their primary or secondary objective, which will exclude minor or exploratory components of projects.

| Planned analyses
We will present the EGM with a numeric summary of included items, broken down by thematic domain (intervention) and innovation type (outcome), as well as other subgroups such as stage of development or unit of measurement. We will also present a narrative report synthesizing several aspects of the EGM, described below. In the case of "unsuccessful" methods, metrics and tools, which do not result in publications because of their flaws, these projects will be documented and included, but not analysed in depth. It would be useful to map these so as to learn from their shortcomings when we move to the evidence synthesis stage. We will attempt to identify these through expert consultation.

| Presentation
In this EGM, the rows of "intervention" will be various domains of influence along the agriculture or food systems to nutrition pathways, subdivided into several thematic domains such as household or farm production, women's role in agriculture, or food and farm policy and governance. These domains are derived from the conceptual frameworks in Appendix A, and listed fully in Appendix B. The columns of the map will group items by tools, metrics and methods, which will be further categorized into tools: technology application and survey and interview tools; metrics: measures and indices; and methods: research design and analysis.
Furthermore, we will internally code for the following characteristics, which will be modified or added to iteratively: -Stages of development (described in quality appraisal) in place of a traditional quality assessment: • Concept development and pilot Cross-cutting theme filters: • Private-sector engagement • Economics (e.g., cost-benefit, cost-effectiveness, cost of diets, purchasing power) • Gender dynamics and parity (Could be about men or women-anything about gender dynamics, e.g., decision-making power over productive resources; employment in agriculture; self-care and health decisionmaking; social gender cohesion; women's mobility; leadership).
• Equity (e.g., caste/ethnicity, economic, geographic, age) • Technology • Political fragility, conflict • Diets We will use the intervention (domain between agriculture and nutrition) and the outcomes (innovation items), as well as these domains to create the tables and figures, which will be decided finally based on our findings. Preliminarily, we propose a section of the report summarizing interventions, and another to provide a narrative summary of the outcomes. The report will include a section discussing the overall stage of innovation assessed in the map. We will include a section that highlights gaps and under-researched areas for future investment (but not prioritized as this is subjective based on the user), as well as compare the portfolio of innovation to the existing conceptual frameworks used. Tables and figures may include: • A PRISMA diagram researcher who has been working in the field of public health for over a decade.
• EGM methods: Howard White, CEO of the Campbell Collaboration and a veteran systematic review expert, will lead the EGM methodology and advise the content experts on framework and protocol, as well as helping train and coordinate the information retrieval.
• Information retrieval: Information retrieval will be done by experienced teams hired and managed by the Campbell Collaboration, in collaboration with Thalia Sparling and Suneetha Kadiyala.

SOURCES OF SUPPORT
This work is funded through the IMMANA programme, which is

DECLARATIONS OF INTEREST
None.

PRELIMINARY TIMEFRAME
Approximate date for submission of the EGM: June 15, 2019.

PLANS FOR UPDATING THE EGM
We are currently raising funds for a follow-on research programme to the first IMMANA programme. If IMMANA 2 is successfully funded, we will update the EGM at the conclusion of the project cycle, approximately at the end of 2024 or beginning of 2025.