In implementing the project, we will adopt a multi-method cross-sectional study design that applies quantitative and qualitative methods. The various methods to be deployed are described under field procedures.
Study tool development and adaptation processes
The protocol for this project was developed iteratively and collaboratively by project partners. A series of virtual engagements between the project team and leaders of the adapted INFORMAS Modules led to the development of initial drafts of the protocols. Following this, a face-to-face protocol adaptation workshop, facilitated by project co-investigators and Module Leaders was conducted in August 2019, at Aburi, Ghana. Three work teams (each team focusing on one of the work packages) were assigned the task of revising the initial draft protocol. WP1 Food provision received additional inputs from the INFORMAS School Health Group (part of the wider INFORMAS Network). The CRM component (which is not part of INFORMAS methods) adapted an existing Net-Map interview tool  and CRM protocol , building on our previous use of CRM in Ghana in deprived communities. The various modifications/adaptations made to the original protocols are highlighted in Table 1. The final MEALS4NCDs protocol comprised 15 tools. Altogether, the INFORMAS protocols contributed significantly to creating a project protocol that is tailored to the Ghanaian context. Details of the field procedures are outlined below.
TV monitoring (Nationwide)
The following sections describe the planned implementation of the TV advertising monitoring.
Selecting TV channels, duration of monitoring and days of the week to record
Sampling of TV programming will be conducted in three stages; 1. Selection of TV channels to monitor; 2. Duration of monitoring (Months, Days and hours to be recorded), and; 3. Days of the week to record. Ghanaian TV services includes free-to-air (analog and digital services) and paid satellite. By the end of June 2017, there were 51 free-to-air TV channels in Ghana; all channels have some time allocated for children’s programming . No data are available on child TV viewing audiences in Ghana. As such, we plan to consult with experts in the Ghana media industry including media measurement service provider – GeoPoll  to ensure that we select the most appropriate channels. Channels that do not broadcast adverts will be excluded. Also, pay-to-air TV channels will be excluded due to low household subscription. Number of pay TV households in Ghana in 2018 was 0.6 million and is expected to increase to 0.7 million in 2023 . At a minimum, the top 3 free-to-air local/national television channels – in terms of children’s viewership, will be included in the data collection process.
The data collection period will cover one school term (3 months) and one-month of vacation/recess. National holidays and special events will be excluded to ensure that data represents typical broadcasting. For each advert broadcast during the sampled period, details will be collected on brand and products, channel, time shown, nature of the product (food, non-food) and food type. All programming for selected days will be recorded from 6:00 am through to 00:00 am (midnight).
In the final stage of sampling, we will adopt the composite week sampling method to identify the days of the week to be monitored. The work of Riffe et al.  and Laar  show that this sampling approach is consistent with the INFORMAS’ stratified sampling design requirement of a minimum of four days in both the working week and at weekends. Thus, the content of the three TV channels will be recorded as follows: a random start date for TV Channel 1 - Sunday for week 1, Monday will be recorded for week 2; Tuesday will be recorded for week 3, Wednesday for week 4, etc. The other two TV channels will be recorded on the same days as the first to avoid any differences in advertising patterns between weekdays and weekends. In the event that a sampled day is not eligible (e.g. national holiday or special event), the next available date is included. Irrespective of start day/date, efforts will be made to ensure that a minimum number of four weekdays and four weekend days will be recorded.
Data will be collected manually by recording live television programmes as per existing protocols [37, 59]. All programming will be recorded onto external hard drives. The recorded data will be coded using a coding sheet (Appendix 1).
Assessments at the sub-national levels
Sampling study districts and child-serving institutions (public sector basic schools)
A multistage sampling approach will be employed to select the study’s geographic region, districts, and the main Primary Sampling Units (PSUs) – public basic schools. The first stage of the sampling process will entail purposive selection of a geographic region. Of 10 regions in total, Greater Accra, which hosts the national capital was/and remains the most urbanised and most marketed to region of Ghana, will be purposively chosen. The region is sub-divided into 16 administrative districts categorized as Metropolitan, Municipal, and Districts Assemblies (MMDAs). A representative sample of six districts will be selected using both probabilistic and non-probabilistic sampling approaches as detailed below.
To select the six administrative districts, the 16 districts will be grouped into three strata/sub-administrative units of districts, municipalities, metropolises. This administrative categorization is a marker of urbanization status and gives an idea of district-level socio-economic deprivation. Of the 16, two are Metropolises with comparable poverty headcount (the proportion of population living below the national poverty line). The most urbanized will be selected. Nine are Municipalities (three will be randomly selected after further stratification by poverty headcount); and five are Districts (two will be selected using the same criteria as in the selection of Municipalities).
With an overall sample of 200 schools, the required number of basic schools in each of the six districts will be selected using probability proportional to size, with size of the school defined as the number of pupils in the school. Schools will be stratified by location (rural or urban); level (primary, or junior high), and sorted within each stratum based on roll size data from the Ghana Education Service. Following this, a systematic random sampling will be applied in their selection. According to the records of the Education Management Information System (EMIS) of the Ghana Education Service, the Greater Accra region had a total of 862 public primary schools and 812 public junior high schools, in 2018 . Sampling 200 schools will give us 80% statistical power to compare a hypothesized outdoor unhealthy food marketing rate of 47% to the recently observed rate of 57% .
Mapping and assessing outdoor advertising in and around selected schools
All outdoor adverts in stipulated areas, surrounding all sampled schools will be observed and recorded manually, using camera and geo-positioning systems-enabled tablets. The stipulated area will be defined using a walkable road network distance from the entrance of the school. A walkable distance is herein defined as limited by 250m road network boundaries, within which pupils are able to purchase from food vendors during short breaks. To implement this, research staff (working in pairs) will walk and observe the area within the established perimeter, around the sampled schools, to identify and record coordinates as well as pictures of all adverts (food and non-food adverts). All captured food adverts will be coded using a predesigned coding sheet (see Appendix 2). In cases where multiple adverts are included per picture, the type of advert, the type of food as well as the brand will be the focus in the photographed image. Since adverts in outlet windows count as outdoor advertising, we will capture all such adverts associated with both mobile/ephemeral and non-mobile outlets.
Mapping food provision and food retail outlets within stipulated school zones
At the same time as the outdoor advertising assessment, we will conduct a census of all food retail outlets (a place/structure used primarily for the preparation and sale of food and non-alcoholic beverages mainly for consumption off the premises) and food provision outlets (a place/structure used for the preparation, sale and serving of ready-to-eat food to customers) within the immediate school compound - using a road network distance of 250m from the school entrance. This mapping exercise will geo-locate all kinds of outlets categorized as “Supermarket; School canteen; Shop; Kiosk; Vegetable/fruit/food stand/Table top; Local vendor; Restaurant; Chop bars; Cold stores; Open market; Bakery; Drinking bar/pub as in Green et al . All such outlets may be operated as completely self-service; both self- and assisted-service or full assisted service. In addition, the name of the outlet (if available) will be recorded and a picture taken. The unique foods available will be assessed and categorized using the NOVA food classification system  or the INFORMAS’ core, non-core, miscellaneous approach . Our experience implementing similar exercises reveal a phenomenon of shared outlets/vending spaces and the tendency for outlets to appear and disappear at certain times of the day . Therefore, efforts will be made to map fixed, mobile, as well as non-mobile, yet ephemeral outlets (see Appendix 3).
Conducting in-store/supermarket assessment and post-purchase exit interview
The supermarket/in-store promotional assessment will be implemented simultaneously with administration of a short exit survey (see Appendix 5), which will be administered to eligible adults customers exiting from shopping from the identified vendor. Shoppers will be eligible if they are parents/guardians with a child(ren) aged 6-17 years and have bought at least a food/beverage product from supermarket. A consecutive sample of 10 interviews per supermarket is expected. The short survey will assess overall shopping experience in the store, what was purchased, as well as promotional activities they encountered at the supermarket. In addition, relevant questions from Green and Glanz  perceived nutrition environment measures survey will be fielded.
Food provision policy, programme, and practice assessment and monitoring
Per existing INFORMAS protocols, the exercise will aim to, first, describe the nutrition standards or guidelines that are in place to implement specific policies or programmes within public basic schools in Ghana. Second, it will evaluate the nutritional quality of the foods and beverages sold by vendors (e.g. in school cafeterias, and canteens) or provided (e.g. foods provided by school feeding programmes) in these settings. In addition, the study will explore key considerations in making decisions around what foods are provided in such settings by engaging relevant stakeholders. We will adopt L’Abbe et al  two-component framework.
Briefly, Component I will involve compilation, description, and evaluation of any existing nutrition policies/programmes using the expanded INFORMAS step-wise approach. This component will involve three key steps:
Step 1 will identify food provision policies and programmes through a literature search and stakeholder interviews. The initial policy/programme identification will rely on our recent Food-EPI exercise , complemented with interviews with relevant stakeholders - using Appendix 6.
Step 2 will describe the characteristics of the food provision policies and programmes identified in step 1. The core characteristics to be identified will include the jurisdiction of the policy/programme; the availability of guidelines for the policy/programme; geographical coverage of the policy/programme; the school sector coverage of the policy/programme; the type of nutrition standards/guidelines associated with the policies/programmes. This will be implemented using a checklist of ideal characteristics of a school feeding policy/programme (see L’Abbe et al 2013).
Step 3 is dedicated to unpacking details of standards/guidelines applied to foods/beverages that are provided and/or sold in the identified policies/programmes and described in steps 1 and 2. For instance, step 3 will determine the basis of any applied standard (i.e. whether the standard is applied per serving or per 100g of food). With the help of a policy rating tool (see Appendix 7), we will describe existing policies/programmes, their associated standards/guidelines, and evaluate them against ideal benchmarks for school food policies/programmes.
Component II entails monitoring implementation of food provision policies and programmes and will involve three key steps as detailed in L’Abbe et al 2013. It will be implemented using three separate study tools (see Appendices 8, 9, 10, 11). The purpose of this component is to evaluate the nutritional quality of the foods/beverages sold or provided to children in these settings relative to the nutritional standards/guidelines in any identified nutrition policy/programme. As many schools in Ghana run a system that includes both foods sold on the school compound and food provided by school canteens/commercial food service canteens, we propose to conduct our evaluation of foods sold separately from foods provided to children. Nutritional quality here refers to compliance with nutritional standards/guidelines applied in the rated ‘strong’ policies/programmes.
We anticipate that in many of the schools, nutritional quality assessment will not be conducted as part of the monitoring of the school food/nutrition programme, as such we will conduct interviews with school heads, administrators, and/or food service providers such as head teachers, caterers, cooks and food service vendors. Second, in the event that there are no nutrition standards/guidelines to guide implementation of school food/nutrition programme, we will conduct our assessment of the nutritional quality of the provided and/or sold foods relative to the standards applied in the INFORMAS’ Core/Non-Core food classification  or the NOVA classification  , or locally generated classification.
Focus groups discussions (FGD) with school children and interviews with their caregivers
In total 18 FGDs will be conducted (three per district) with children (12– 17 years).. Each will comprise a maximum variation sample of 8-12 school children – selected with consideration given to gender, age, location, level of education (primary or junior high) and socio-economic status).. Focus groups will explore children’s nutritional knowledge and perspective on what a healthy diet is, their typical eating habits (what they eat on a regular day –weekends/weekdays), changes in diet over time, awareness of school nutrition policies/programmes, their perspectives on promoting healthy food (counter marketing),views on food adverts in general, media influences on food choices, general recall of features of advertised product and brand recall (see Appendix 12). Parents/caregivers of children who participate in the FGDs will be interviewed. Interviews will cover parental appreciation of children’s school food environment, TV influences on children’s nutrition and health etc (Appendix 13).
Implementation of Community Readiness mapping (CRM)
CRM will be implemented to gauge the readiness of and capacity of our studied community/stakeholders, as well as barriers and facilitators to implement food environment improvement interventions, which is not included in the INFORMAS protocol. The CRM serves as a theoretical framework for understanding and improving community readiness for interventions . Limited research has been conducted in African countries using the CRM. For example, it has been applied in Mali,  and in South Africa , and recently in Ghana . In the current study, the CRM will aim to answer the question of how ready the community is to accept and implement interventions to improve children’s food environments in the Greater Accra region of Ghana. The assessment will focus primarily on aspects of the food environment related to food marketing and the availability of unhealthy foods in and around basic schools. Here, unhealthy children’s food environments refer to: ‘food environments that promote the consumption of unhealthy food items, such as processed meats; sugar and sweet spreads; cakes and sweets; sodas and sweetened beverages; savory snacks; fried foods; and condiments high in sugar or salt’. It can also include food that is not safe to eat because it is unhygienic or contaminated. ‘The community’ is operationally defined both over geographic space (geographic community), and according to stakeholders of interest (interest community). Geographically, the focus will be the six sub-community of districts of the Greater Accra region of Ghana. The interest community will comprise individuals/key informants from both within and outside of the geographic community, adjudged to wield influence and/or to play a leadership role in promoting healthy food environments. It is expected that the selected key informants will be able to provide in-depth viewpoints on existing actions to improve children’s food environments and discuss their barriers and facilitators.
Individual CRM informants will be selected purposively to represent a wide range of sectors/programmes (regulatory authorities, schools, commerce, health (e.g. MoH, GHS, etc), religious institutions, school feeding programme, children/youth clubs, urban planners etc.) as well as policy areas in question – TV marketing, outdoor marketing, food provision etc. The actual list will be compiled with inputs from the study team and local people and interviewed using Appendix 14. To ensure that a maximum variation sample of key informants participate in the exercise, the CRM exercise will begin with a stakeholder mapping using the IFPRI Net-Map methodology . Two to three key informants will be identified per sector. We will be measuring the overall readiness of the community. At a minimum, 6-12 interviews will be conducted (actual number will be determined following the Net-Mapping process). By adapting an existing Net-Map interview tool  (see Appendix 15), participants will be asked to indicate, ‘Who are the stakeholders that influence children/ food environment and their dietary behaviour, at the National level?’ This question will then be followed by another question on ‘how do the identified actors interact with each other?’. The list that will be generated will be placed on a large sheet of drawing paper and with the help of participants, links will be drawn using the markers. Appropriate map generation tools (including large drawing paper, differently coloured markers, etc.) for drawing the map will be provided for the process. Once the links have been established, a determination of their relative influence in the food environment will be made by asking, ‘how influential each actor is in promoting school food environment’.
Figure 2. Overall sampling schema (about here)
Data quality assurance measures will be instituted pre – during, - post data collection – including identification of qualified field staff, rigorous training of field staff and pretesting of study tools.
For instance, during the pre-test, two principal researchers and all research assistants would code an identical area around one site independently. Their respective coding would then be compared and inter-rater reliability (IRR) computed using the formula:
The INFORMAS protocols we have adapted recommend a minimum of 80% inter-coder reliability. In the event that this minimum level is not achieved, the coding discrepancies would be discussed, and further training would be given to research assistants, if required.
Similarly, for TV monitoring data, the raw data (recorded videos) will be coded by trained members of the team with the help of the study-specific coding sheet, as per the INFORMAS protocol  Inter-coder reliability will be computed as follows: A random sample of 10% of all coded images will be selected and coded by an independent researcher. Individual scores for each attribute (all categorical variables) will be compared using Cohen's kappa statistic (κ).
Lessons from the pretest will inform modifications and finalization of the tools. There will be no pilot study as lessons from recently implemented studies [22, 70, 71] provide sufficient guidance on sample size calculations, community entry, and field etiquette.
Different approaches will be deployed to analyse data from this study. We describe below how the pooled quantitative, qualitative, policy assessment and GIS data will be analysed. In Table 2 is a summary of the various indicators that will be generated as part of this project.
Analysis of quantitative data
Advertising, publicity and sales promotion of food to children via Television, and outdoor and indoor advertising within school zones will be analyzed strictly as per INFORMAS’ protocols . For each school, two kinds of data will be generated: one related to the outlets observed, and the other, the foods items provided or sold. Density of outlets will be calculated per 100m2, to enable standardized comparisons within the 250m road network distance.
For in-store marketing assessment, we will present frequencies of all promotional strategies taking place in these outlets/stores including audiovisuals (music, loudspeakers, screens with displaying promotional characters), spatial distribution and product placement. Pictures of unique foods will be assessed and categorized using the NOVA system , and/or the core/non-core, food-based approach developed by INFORMAS . Frequency tabulation of post-shopping exit interview data detailing items purchased, overall shopping experience in the store, whether they did prior planning for the shopping at home; whether they bought any food or drink not previously contemplated; and the promotions they remember from the store. At the school level, frequencies of healthy food/beverage availability will be generated and compared by school level (primary or junior high school), district (district, municipality, metro). Ratios of healthy food outlets to total food outlets will be calculated for each provision outlet based on the total counts of each. Further, bivariate and multivariable analyses will be performed using appropriate approaches.
Analysis of GIS data
ArcGIS desktop software will be used to analyse spatial data and generate maps. Availability of various categories of products/adverts will be compared cartographically. Thus, the total number of food outlets and adverts within a 250m and 100m distance around rounds in each school will be mapped. The maps will be produced to help understand how the food environment varies spatially in our study locations. Also, to be mapped for each district and for pooled data, will be the descriptive data of frequency of food/beverage adverts, size of food advert, setting of adverts, type of advert, product type, and location of adverts or food provided/sold within the schools. Geographically weighted regression (GWR) technique will be used to model the local relationships between the predictor factors (e.g. rurality/urbanity, demographic factors; physical environment characteristics, type of school, roll size (number of student in the school), district type, poverty incidence etc.) and the outcomes of interest - intensity of outlets/adverts within 500m radius.
Analysis of qualitative data
Audio recordings of FGDs, the KIIs, as well as the IDIs will be transcribed verbatim. Transcriptions will be augmented with the researchers’ field notes. The data resulting from the transcriptions will be evaluated and coded using the constant comparative method of theme generation. Qualitative software (e.g. Nvivo SQ) will be used to assist with data storage and management, including development of data files, codes, codebooks, themes, and categories. The CRM-specific transcripts will be scored by two independent scorers within the research team using 9 anchored rating statements for each dimension . Data reporting will be undertaken according to the consolidated criteria for reporting qualitative research (COREQ) . To maintain confidentiality, all names will be reported as pseudonyms and identifying details removed.
Data from the Net-map (including the photographs, notes, and transcripts), on the list of actors (using actor identification codes), links between actors, and relative influence will be entered into Microsoft Excel (as two separate sheets: 1. list of individual actors, and 2. list of links connecting all actors). Actor identification codes and actor categories (the broad sectors that actors belong to) will be recoded for consistency across link list and actor list. The two excel sheets will be imported into Gephi, a social network analysis software. All analyses (statistics, filtering, network image generation and visualization) and image capturing will be conducted in the overview and preview features of Gephi. Overall, and link-specific network statistics (weighted average degrees, network diameter, graph density, modularity and connectedness components) will be computed to determine network characteristics. The network visualization images will be weighted by actor nodes using relative influence. The Yifan Hu algorithm will be used to visualize network images.
Overall, data reporting will be undertaken according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies.
Stakeholder validation, dissemination, and knowledge translation
The results from this study will be shared during a validation workshop to open to all the MEALS4NCDs stakeholders. We will adopt the ‘constructive confrontation design’. It is envisaged that the process will lead to an enhanced collaborative conversation on our initial findings, critique of, and possible re-interpretation of the results. Also, draft policy briefs will be tabled for inputs during the validation workshop. We believe that the views of multiple stakeholders, will strengthen the policy briefs.
Our proposed strategies for dissemination include standard dissemination activities such as publishing peer‐reviewed manuscripts (anticipate n=10 in reputable open-access journals) and presenting at local or global scientific conferences (n=6). We will also create an open-access project website, generate and disseminate project reports, produce policy briefs, conduct consultative meetings/dissemination workshops. Additional efforts to disseminate study data more widely and to support its ultimate application may entail packaging of the research findings into user‐friendly formats e.g. press releases, research briefs, guided by our research findings. We will also run “building project findings into policy and practice consultative meetings” (n=4). This will entail several consultative meetings with diverse, but relevant stakeholders including government ministries, departments, international organizations including the WHO, FAO, UNICEF, as well as civil society. Among others, these meetings will deliberate on the strategies for feeding the project’s findings into policies, practice, and scale up. We will produce policy briefs (n = 2 covering food promotion and provision) to facilitate dialogue with these stakeholders.