Sampling Procedure
Work Package 1
a) Sampling design
Purposive and snowball sampling techniques was used to identify eligible participants for the key informant interviews. Relevant government departments and agencies that have responsibilities related to development and implementation of policies on nutrition, the food environment and NR-NCDs were purposively identified. This iincluded the government departments, civil societies and representatives of relevant organizations in the wider food system in Kenya. Purposive and snowball sampling was then used to select the key people (policy & decision makers) in each of the departments who best understand the issues of interest to this study) to be interviewed.
The initial sample commenced with 1) Government agencies with responsibilities related to policies on nutrition/food environment, and NR-NCDs: Finance, Industry/Commerce, Trade, Health, Agriculture; 2) Food industries and manufacturers; 3) Civil Society actors relevant to health and/or food; 4) Journalists; 5) the media.
Recruitment of interviewees was through formal approaches to the heads of relevant agencies. Once approval was obtained, we contacted relevant departments for identification of the individuals who best understand the issues of interest to this study either by their positions (senior/intermediate) or period worked in the department; and subsequently request for interviews. At the end of each interview, we asked interviewees to identify further relevant interviewees.
b) Sample size
Approximately 10 key informant interviews were conducted with key policy and decision makers relating to NR- NCDs as well as the food environment in Kenya. We anticipated that this number of experts from the various government departments and the wider food systems' stakeholders, would adequately present the issues relevant to policies and actions on nutrition, creating a healthy food environment and NR-NCDs in Kenya. However, there was flexibility to accommodate emerging issues that may need additional follow-up interviews, hence the final sample size was based on saturation of the ideas from the interviews.
Work Package 2
TELEVISION AND RADIO MONITORING
a) Sampling design
TV and Radio to be monitored were sampled in three stages including i) selection of the TV and radio channels to monitor, ii) the calendar period (duration) of the monitoring (months, days, hours) and iii) days of the week to record.
First, purposive sampling was used to select TV channels to be recorded. In the past years, Kenya has completely transitioned from analogue to digital television type with both free to air and paid for TV channels. According to the communication authority of Kenya (CAK)'s report (2019 -2020), the free to Air TV is the most popular form of television, enjoying a 95% popularity in Kenya [52]. As such only the free to air TV channels that broadcast advertisements were included. Paid for TV channels and those that do not broadcast advertisements were excluded.
After the identification of the free to air TV channels and Radio that broadcast advertisements, the most popular television and radio channels were then selected for monitoring, based on their levels of popularity with children aged <12, <16 or <18 years during peak viewing times. Popularity was assessed based on the updated local ratings data obtained from TV and radio audience measurement institutions such as GeoPoll [53], the CAK. For instance, KARF report of 2020 rated Citizen and KTN televisions as the most preferred channels, with children 15 to 17 years as the age group that watched most TV [52]. However, this may change with time, and therefore the prevailing TV and Radio ratings at the time of data collection were used. Where local ratings data were not available, we consulted with experts in the media industry in Kenya.
Second, based on the recommendation for recording over at least a 3 month period [48], random sampling was used to select three consecutive months for recording. In a typical school (primary and secondary) calendar in Kenya, school term runs from January to March, May to July and September to mid-November, with April, August and December as vacation Months. The selection of the recording month ensured inclusion of both school term months and a vacation month in order to capture any variations in food advertisement or promotion during school days and vacations
The data collection period excluded national holidays and special events, to ensure that data represents typical broadcasting.
Thirdly, stratified sampling design was applied to select the days for recording, with weekdays and weekend days comprising the two strata. The INFORMAS protocol recommends at least four week days and four weekends as optimal for recording [48]. Thus, the random sampling was used to select four week days and four weekend days in each of the three months selected for recording. Each channel was recorded for 18 hours a day, from 0600hrs to 2400hrs (midnight), which entail the peak viewing times for children.
b) Sample size
Given that the focus of the monitoring was children/adolescents, TV and radio channels were chosen based on their levels of popularity with children aged <12, <16 or <18 years during peak viewing times. At minimum, the top three most popular television and radio channels respectively were included for monitoring. Monitoring was conducted on at least four week days and four weekends over a three month period, hence a total of twelve week days and twelve weekend days. These number of channels, days of monitoring and duration of monitoring have been documented as optimal to assess the nature and frequency of advertisement and promotion of foods to children [48].
OUTDOOR ADVERTISING
a) Sampling design
A multistage sampling approach was employed. At the first stage, purposive sampling was applied in the selection of the three counties, Nairobi, representing an urban and major capital city, Nairobi representing a metropolitan city, Mombasa, a coastal tourist city and Baringo, a rural setting and all the three counties are in different geo-political areas. Inclusion of these three counties ensures that we capture any variations in outdoor advertising and marketing of unhealthy food to children in different geographical and social contexts.
In the second stage, Nairobi and Mombasa Counties were stratified further into urban formal and informal settlements according to deprivation levels and socio-economic differences defined by the county government and one area (sub-county) was selected from each of these strata, while in Baringo County, one rural sub county was randomly selected from a list of all rural sub-counties in Baringo.
In the third stage, a random sample of schools was drawn from each area selected in stage 2. The schools were further stratified by type of school (public and private, primary and secondary). The number of schools selected in each location was proportionate to the size of the area (in terms of the total number of schools in the area).
This selection of different sub counties/ strata within the three counties captured the range of schools by deprivation levels and socio-economic differences to show the variations in advertisements in each of these contexts.
In order to undertake this process of sampling, we approached the counties, sub-counties and the ministry of education to provide the list of schools with the level of detailed information needed for sampling.
b) Sample size
We estimated approximately 149 schools per county to be sampled based on prevalence of 23% unhealthy advertising in Nairobi low income settings [29], a 10% higher or lower unhealthy food advertising around schools depending on location, a power of 80% and a margin of error of 0.05. Considering the three counties, yields a total sample of 447 (? 450) schools.
MARKETING IN-STORES
a) Sampling strategy
Purposive sampling was used to select the major supermarkets or grocery stores selling foods and non-alcoholic drinks in the selected sub-counties. The major supermarkets or food outlets were identified through local or international market research reports such as Euromonitor International. Where such information is unavailable, a rapid assessment was conducted to identify the major food outlets in the study area after which purposive sampling was applied to select the most popular outlets for inclusion in the survey.
b) Sample size
To implement this, the main supermarkets were identified through a local market research reports or rapid assessment of the main supermarkets in the selected sub counties/ neighborhoods. All the main supermarkets identified were included in the survey.
QUALITATIVE INTERVIEWS
a) Sampling design
For the key informant interviews with policy and decision makers, purposive and snowball sampling were then used to select the key people (policy & decision makers) in each of the departments who best understand the issues of interest to this study, to be interviewed.
Purposive sampling was used to select parents and children to participate in the in-depth interviews and focus group discussions respectively. Participant characteristics on gender, location (urban/rural) SEC status (informal and formal settings), and type of school (public/private) were considered during sampling and participant selection to ensure representation and collection of information from a wide variety of participants and perspectives.
b) Samples size
An estimated 15 key informant interviews were conducted with various stakeholders representing policy makers, school authorities, media houses, store managers, other food service vendors, and sports event managers.
For children in schools, three focus group discussions with six to eight participants each were conducted in each sub county/neighborhood, one male-only, one female-only and a mixed group; and hence an estimated 90 participants / children).
Similarly, In-depth interviews with parents representing at least 2 to 3 parents with children schools in the sub-counties/ neighborhoods selected, and hence an estimated 15 parents. Gender representation were considered while selecting the parents to participate in the interviews in each school.
