MoreMilk Project: Milk Safety and Child Nutrition Impacts of a Training Scheme for Dairy Vendors in Nairobi, Kenya

The International Livestock Research Institute (ILRI), together with partners, has pioneered an approach to improving the safety of food sold in informal markets by professionalizing informal traders who operate in these markets in low- and middle-income countries. The approach involves training vendors in hygiene and business skills and improving their ability to market products through brand creation, building skills in sales and milk promotion, and/or social marketing. The approach appears sustainable and scalable, but the nutrition and health impacts, though potentially important, have not been evaluated. The objective of the study is to assess the impact of a Training, Certification and Marketing intervention for milk vendors in the informal sector on milk safety and nutrition outcomes in children in peri-urban areas of Kenya.

The study will be conducted in 3 sub-counties in peri-urban Nairobi, Kenya. The study area was selected for its peri-urban status, ongoing urbanization, and high density of livestock and dairy production. The target wards are characterized by neighborhoods with adequate infrastructure and multi-storey buildings and areas with tin-roofed houses and limited access to piped water, sanitation and electricity. Importantly, it is estimated that about 70% (Kenya Market Trust, 2016) of the milk consumed in the study area is sold through the informal dairy vendors-vendors that sell unpackaged or not-formally processed milk-who are the focus of this study.

A two-arm cluster-randomized, non-masked, community-based trial will be used to estimate the effectiveness of the MoreMilk intervention. The RCT aims at answering the following two research questions:

• Does a "training, certification and marketing" intervention for dairy vendors operating in the informal dairy markets in peri-urban Nairobi improves the microbiological quality (i.e. safety) of the unpacked milk sold in informal markets?
• Does a "training, certification and marketing" intervention for dairy vendors operating in the informal dairy markets in peri-urban Nairobi increase the natural log of mean dietary adequacy of protein, Ca, and vitamin B12 for children 12-48 months (at baseline) in consumer households?

The study population will include dairy vendors operating in the informal sector (businesses selling unpackaged or not-formally processed milk to individual consumers; this includes raw, boiled and unpacked pasteurized milk) in the study areas and consumer households that purchase most of their weekly milk from those vendors. The unit of randomization is a cluster of 1 to 5 vendors and the households consuming milk from those vendors. Clusters are identified using a hybrid approach based on a computer-based algorithm informed by the GPS locations of the informal milk vendors in the target areas. The resulting clusters ensure that vendors in one cluster are located at least 250m away from any vendor in a different cluster. Study clusters will be randomly assigned to one of two arms:

1. Treatment arm: Training, certification, and marketing scheme for milk vendors.
2. Control arm: Delayed intervention (implemented after endline survey). Study vendors in the intervention (treatment) clusters will be offered the possibility to participate in a 12-15hr training. Trainings will be offered free of charge. Upon completion of the endline survey, vendors in the control clusters will also be offered the training, which will follow the same format as the one used in the intervention clusters. The training will be offered a few hours per day over 5-days to accommodate vendors' time constraints, which we expect to be more severe for female vendors. The trainings will be given by business development service (BDS) providers. BDS correspond to a "wide range of non-financial services provided by public and private suppliers (BDS providers) to entrepreneurs to help them operate efficiently and to grow their business with the broader purpose of contributing to economic growth, employment generation and poverty alleviation" (Kimando, 2012). The Kenyan register of BDS providers will be used to identify available BDS providers. The selected BDS will receive a full training course, covering the intervention content and specific teaching methods to be used during the training of vendors.

The study will include 142 vendor clusters (with a minimum number of vendors per cluster of one and a maximum of five), 240 vendors and 852 households at baseline. For the health outcome, the study will be powered to detect a 0.4 standard deviation difference between the natural log of TBC in the treatment and control groups. This equates roughly to a 20 percentage point increase in the proportion of vendors selling milk meeting EAS one year after the intervention, off of a proportion of 37% in the data we collected in the study areas in 2017. For the nutrition outcomes, the study will be powered to detect a change in the natural log of mean dietary adequacy of Ca, protein, and vitamin B12 that corresponds to a daily increase of 75gr of milk (0.5 cups) consumed per child. This is one-third of the median daily milk consumption for children 24-48 months of age in the MilkMarkets data collected in Dagoretti in 2017.

Randomization of vendor cluster to treatment or control arm will be conducted immediately after completion of the baseline surveys for vendors and consumer households, and hence post-consent. After randomization, vendors in clusters allocated to the intervention will be contacted, informed of the training schedules available in their ward, and offered the possibility to attend 1 of up to 3 scheduled training options. Vendors in the control clusters will be informed that they will be given the option to participate in the trainings one year later.

Selection and recruitment of vendors - From the full list of identified clusters, we will randomly select the 142 to meet the study sample size requirements. In each cluster, a minimum of 1 and a maximum of 5 vendors will be recruited.

Selection and recruitment of households - For each participating cluster, we will recruit up to 6 consumer households that meet our inclusion criteria. Each recruited vendor will be given a small add to display in their business/premise advertising the study and inviting interested consumers to register interest to participate in the study by providing their name and telephone number in a registration form. Vendors will be encouraged to bring the study to the attention of their customers. The information will be kept at the vendor's premise for up to 10 days and then collected by field workers. Registered consumers will be contacted and information collected to check their eligibility (see criteria above). If not enough eligible households were identified, a second approach to recruitment will be used. A field worker will spend 2-3 hours in the morning or the evening of a given day at the recruited vendor's premise and approach customers that purchase liquid unprocessed milk during this time period. A brief set of questions will be asked to confirm whether the consumer household meets our inclusion criteria. If it does, the study will be presented and interest to participate in the study will be recorded. Basic contact details will be obtained to facilitate follow up. After identification of all eligible households in the cluster, 6 households will be randomly selected to participate in the study aiming at achieving equal number of household per vendor(whenever there is more than 1 vendor in a cluster) and ensuring there is at least one household per vendor in each cluster.

Selection of the index child - In each recruited household one child 12 months to 48 months of age will be selected. If only one child falls into this age window at the time of baseline, this child will be the index child. In case there are multiple children in the household meeting the age criterion, the index child will be randomly selected manually or by the computer-assisted personal interview software.

A baseline survey and an endline survey will be conducted in recruited vendors and households prior to the intervention (baseline) and 12 months after baseline (endline). The questionnaire to vendors will gather information on the following three elements: (i) milk hygiene and handling practices of daily operations, including sourcing of milk, equipment available and use of quality testing; (ii) business performance, recording milk volumes purchased/sold, expenditures, revenues and cost of equipment; (iii) women's empowerment in their business and related household gender dynamics. In addition, two un-announced visits to the participating vendors will be conducted between the intervention and the enline survey, where observable features on milk hygiene and handling will be recorded through a short questionnaire. These two mid-term visits will allow for collection of information on business performance indicators and any unexpected negative consequences from participation in the study. In households, questionnaires will record the following information (i) household food expenditure, with focus on milk products; (b) milk consumption, handling and hygiene practices; (c) a 24hr dietary recall for the selected child 12-48 months at baseline; (d) presence of diarrhea in the past 7 days (diarrhea understood as one or more days having at least 3 loose stools/day).

At baseline, endline, and at the 2 unannounced visits, up to 100ml of milk sold by participating vendors will be bought. Samples will be tested for total bacterial counts (TBC), enumeration of enterobacteriaceae and presence and enumeration of coagulase positive staphylococci. For vendors in the intervention clusters, results of the laboratory analysis and feedback on milk hygiene practices will be communicated after each testing.

The "multiple-pass" 24hr recall method will be used. This approach uses a special method to help the primary caretaker, usually the mother, of a child remember what was consumed and ensure that enumerators do not miss important information. Using the 24hr form, enumerators review the day's food and drink several times with the primary caretaker of the child. Each time, or "pass" more detailed information is added to the recall form, thus reducing the probability that foods are omitted. A detailed 24hr recall protocol has been developed. Length/height and weight of the selected child and his/her mother will be recorded.

The process evaluation will examine the primary inputs, processes, outputs, and outcomes along the program impact pathway to determine the "how" and "why" of program impact. The scope of the process evaluation is limited to providing a qualitative assessment of issues related to program implementation and delivery. A combination of random and purposive sampling techniques will be used for the process evaluation. Quantitative data will be collected through precoded questionnaires. Qualitative data will be collected through semi-structured continuous observations of program activities, semi-structured individual qualitative interviews with implementation staff and beneficiaries, and group free listing with beneficiaries.

A scientific advisory committee has been set up, formed of 3 individuals external to the research team and to the project partner institutions. The Committee will revise and provide feedback on the study protocol and the study research tools (questionnaires, consent forms, …), as well as conduct data monitoring for selected indicators. The Committee will select a chair for the duration of the study, who will convene meetings as required based on the outcomes of the monitoring.