• No results found

Chapter 2. Study design, setting and methods

2.8 Methodology and research tools

Details of research methods relevant to each chapter are given in Chapters 3-7. Chapter 3 outlines the databases and other search methods used to identify relevant literature for review.

Chapter 5 describes inclusion criteria for national and regional food composition databases in sub-Saharan Africa, and the approach taken to collate and analyse data. Chapters 4, 6 and 7 provide information on the project sites in Tanzania, the two-stage selection of study participants, the research tools used, and the processes of data collection and analysis. Some repetition exists, as a result of the publication-based nature of this thesis. Additional information provided in this chapter includes an overview of the processes and people involved in data collection, and each of the research tools employed.

Amongst the challenges of a study involving repeated household visits over multiple rounds of data collection is the challenge of respondent fatigue. Researchers working in resource-poor settings commonly face ethical dilemmas about the nature and distribution of benefits to study participants and the wider community (Molyneux, Mulupi, Mbaabu, & Marsh, 2012), summarised by Ballantyne as: “offer participants too little and they are exploited, offer them too much and their participation may be unduly induced” (2008, p. 179). In the Nkuku4U study, small gifts such as soap, fabric or hats for children were provided to acknowledge the involvement of enrolled households, and refreshments were provided during focus group discussions. This appeared to contribute to maintaining good relationships with study participants, particularly during a period when poor and abnormally timed rainfall had adversely impacted on harvests.

However, it has been suggested that such an approach may introduce bias, create expectations in follow-up rounds of data collection, alienate non-participating households, and set a precedent which other research projects may be unable to follow (Malleson et al., 2008).

2.8.1 Quantitative data

Questionnaire-based data were collected using face-to-face interviews with respondents, conducted by male and female enumerators recruited and trained from within each ward. These people were selected in consultation with local leaders, with the requirements of having adequate levels of literacy and numeracy, being well-respected within the community, and being available to work for periods of one to three weeks of data collection for a casual wage.

Efforts were made to ensure approximately even numbers of male and female enumerators, and to accommodate the involvement of breastfeeding mothers. Many enumerators continued to work with the project across multiple data collection periods in each ward.

The decision to employ community members as enumerators contrasts with the approach of many large-scale research programs, which rely on external teams with prior data collection experience. Data for the most recent Demographic and Health Survey in Tanzania, for example, were collected by 16 male and 64 female interviewers, all trained nurses (MoHCDGEC [Tanzania Mainland] et al., 2016). Benefits of this approach include enumerators’ likely familiarity with the research process, questionnaire design and interview technique and, for a multi-site survey, avoiding the need to train a new team in each location. By contrast, employing local people may reduce the perceived hierarchy between interviewer and interviewee (Molyneux, Wassenaar, Peshu, & Marsh, 2005) and has the distinct advantage of an awareness of the context in which the research is being conducted. In this project, enumerators contributed an understanding of local foods (including wild foods), livelihood strategies, health facilities, and the seasonality of rainfall, livestock diseases and agricultural activities.

Training sessions were conducted prior to each questionnaire application, in most cases led by the thesis candidate with translation assistance from Tanzanian colleagues. Questionnaires and training were in Swahili, but enumerators were encouraged to use local languages (Kigogo and Kisukuma) where appropriate to aid in communication with interviewees. Training sessions included opportunities to discuss the local translations for certain key words, including the food items listed in a dietary recall section. Time was also spent discussing unlisted food items which might be consumed within the study setting, including non-domestic animals and uncultivated plants, and in which category these items should be recorded. At the end of each training session, role play was used to provide opportunities for enumerators to rehearse interview skills and clarify areas of uncertainty. Training programs for community health workers in South Africa (Evangeli et al., 2009) and Vietnam (Fisher et al., 2014) have identified role play as an effective activity for building communication skills and increasing adherence to protocols.

Enumerators travelled by foot, or sometimes by motorcycle or vehicle, to locate the relevant respondents and complete the questionnaires, most commonly at their home or agricultural plot. Field supervisors, including the thesis candidate and members of the Tanzanian research team, reviewed completed questionnaires throughout each day to identify recording errors, sections which had been mistakenly skipped, issues of potential misunderstanding (by the enumerator or the respondent), and information suspected to be inaccurate. Enrolled households were revisited to clarify any outstanding points.

Anthropometry of enrolled children and their mothers was conducted at a central location within each village (a health facility, village office or primary school), with a level concrete surface on which digital bathroom scales could be used. Weight, length or height and mid-upper arm circumference measurements were conducted and recorded by trained representatives from the health sector (from the Ministry of Health, Community Development, Gender, Elderly and Children, the Tanzania Food and Nutrition Centre, or Manyoni District Hospital), together with the thesis candidate. Further details on the process of anthropometric data collection are given in Chapter 7.

Between major fieldwork sessions, conducted at six-monthly intervals between May 2014 and May 2016, regular data were collected by two representatives from each village (one male, one female) who were employed on a part-time basis by the Nkuku4U project as “Community Assistants”. As for enumerators, the selection of Community Assistants was guided by leaders’

recommendations, with similar requirements of literacy, numeracy and being well-respected within their village. This role included visiting households on a twice-monthly basis to record the number of chickens owned, the breastfeeding status of enrolled children, and maternal reports of child illness, as well as providing a point of contact between study participants and the research team.

2.8.2 Qualitative data

This thesis has predominantly utilised quantitative methodologies, however a targeted qualitative component was employed within a mixed methods study in Chapter 6. More broadly, time spent by the thesis candidate living and working in the study villages, over a cumulative period of approximately five months between May 2014 and May 2017, provided valuable opportunities to observe and engage with communities through daily interactions. This was instrumental in developing an understanding of the context in which this research has been conducted. Although not reported in this thesis, a series of focus group discussions held in March 2015, conducted separately for men and women, and for different language groups, also contributed to a formative understanding of food-related knowledge, attitudes and practices in the study setting.

A series of in-depth interviews with a subset of mothers of enrolled children (n = 39) was conducted in October 2016. Stratified purposive sampling was used to identify four to six women in each of the eight villages. Eligibility criteria were that women were available on the intended day of interview and willing to engage in discussions for approximately one hour. With

the aim of achieving diverse representation of households, selection of mothers for interviews was also guided by children's HAZ, timing of introduction of complementary foods, chicken ownership and language group, as determined by prior analysis of questionnaire and anthropometric data.

The majority of interviews were conducted at women's homes, with a smaller number held in a central location in the village at the time that women and their children attended the local health facility. Distances to be travelled to reach women at their home were not a consideration in selection of interviewees. Discussions were conducted predominantly in Swahili, with occasional use of Kigogo, and were led by the thesis candidate using a semi-structured guide with open‐ended questions and facilitated by a translator familiar with the study setting. For each interview, a Community Assistant was also present to lead introductions and provide additional translation assistance where required. Further details about selection of interviewees and the interview process are given in Chapter 6.

2.8.3 Research tools

An outline of research tools, including the scope of information covered, target respondents and periodicity, is given in Table 1. Two major semi-structured questionnaires were applied to all enrolled households: the Maternal and Child Health and Nutrition Questionnaire (Appendix A), for mothers of enrolled children at six-monthly intervals, and the Livelihood Questionnaire (Appendix B), for a male or female member of each enrolled household at twelve-monthly intervals. Information on chickens and children within enrolled households was collected by the Community Assistants during twice-monthly visits to enrolled households (Appendix C).

Chicken vaccination records were the only component of data extending beyond the enrolled households, to the wider community. For each vaccination campaign, Community Vaccinators were responsible for documenting the number of chickens intended for vaccination, ordering an appropriate quantity of the I-2 ND vaccine, recording the number of chickens vaccinated and payment received from each chicken-keeper (Appendix D).

The “Visual Diary” was a novel approach used to document the consumption of chicken meat and eggs by enrolled children and any pregnant or breastfeeding woman within their household over a period of four consecutive weeks, at four-monthly intervals (Appendix E). This household-level research tool was developed by Dr. Brigitte Bagnol, adapted from an approach used in reproductive health research in Tanzania and Uganda (Francis et al., 2013; Francis et al.,

2012). Further details about the design and application of this research tool are given in Chapter 6.

Appendix F provides a set of questions used as a guide for semi-structured in-depth interviews with a subset of mothers of enrolled children, which also contributed to the mixed methods evaluation of infant and young child feeding practices in Chapter 6.