Program context and study rationale

This research took place within the context of a program promoting production, marketing, and consumption of traditional African vegetables among small farmers in Kiambu, Kenya and Arusha, Tanzania. The Traditional Foods Project was designed and implemented as an income-generating program within the context of the dual burden of malnutrition . The goal of the program was to increase production and marketing of traditional African vegetables (TAVs) among small farmers, and to improve household income and nutrition. It was a three-year program implemented in both sites by the International Potato Center (CIP), the World Vegetable Center Regional Center for Africa (AVRDC-RCA), and Farm Concern International, a Kenyan non-profit organization. The program had four main components: (1) agronomic training and seed supply; (2) farmer cooperative formation and links to buyers; (3) a nutrition-focused marketing campaign, which

used posters, flyers, demonstrations, and television to promote the image of TAVs among consumers, based on their nutritional content; (4) farmer training about the nutritional content of the vegetables.

There are hundreds of traditional African vegetables [32], but five were cho-sen for promotion in the TF Project based on their nutritional profile, their ex-isting abundance in the sites’ agroecosystems, and their popularity relative to other TAVs, which gave them a higher potential for marketability. All five of the TAVs were dark green leafy vegetables: amaranth (Amaranthus spp.), nightshade (Solanum nigrum, S. scabrum, S. americanum), African spider plant (Cleome gy-nandra), sweet potato leaves (Ipomoea batatas), and cowpea leaves (Vigna unguic-ulata).

The program used a nutrition-focused marketing strategy to boost esteem and demand for TAVs among both consumers and producers. The strategy focused on spreading information about the micronutrient contents of the crops. It did not specifically promote the crops based on medicinal properties, although in explana-tions of the function of iron and vitamin A in the body (iron improves blood and gives energy; vitamin A is important for immunity and eyesight), medicinal use could easily have been interpreted. Iron and vitamin A were the main micronutri-ents discussed in nutrition-focused marketing activities, because the TAVs stand out in content of those two nutrients. Calcium was also sometimes discussed. See Table 4.1 for nutrient values in comparison with kale and cabbage, which are also commonly consumed in the study regions.

The program was designed based on the idea that if consumers and farmers understand more information about the nutrient value of TAVs, they will be more motivated to consume them for health reasons. While this idea may be correct, it also is based on a somewhat Westernized view of “nutrition knowledge.”

Knowl-Table 4.1: Nutrient Composition per 100 grams of African Traditional Leafy

Sources: All data for TAVs are from the FAO Food Composition Table for Use in Africa [151]

except where noted. All data for kale and cabbage are from U.S. Department of Agriculture (USDA) Agricultural Research Service, National Nutrient Database for Standard Reference, Re-lease 20 [277]. All values are for raw vegetables.

RAE = Retinol Activity Equivalent. Conversion rate of 12 units beta-carotene for 1 unit retinol used.

DFE = Dietary Folate Equivalent

†The value listed is an AI (Adequate Intake) rather than an RDA.

*Bioversity (IPGRI) data [33]

**Bioversity (IPGRI) data [60]

***AVRDC data [288]

edge about biochemical contents of vegetables may or may not be relevant, or fit within traditional knowledge frameworks, among the societies of rural farmers targeted. It is possible that exploration of traditional knowledge and knowledge frameworks about the health value of TAVs may be instrumental in understanding the type of knowledge which motivates eating behavior. Such traditional knowl-edge is likely to include medicinal uses of the plants targeted for promotion.

4.1.3 Objectives

The objectives of this study are:

1. To test whether medicinal knowledge is related to consumption in these sites;

2. To explore the sources of and influences on such knowledge, and whether medicinal knowledge has any relevance to nutrition behavior change inter-ventions.

The principle hypothesis is that if people value TAVs for health-promoting properties, which may be expressed as medicinal properties, they will consume more of them.

Understanding how this kind of knowledge is relevant to nutrition interven-tions is important to the success of future interveninterven-tions for both implementing organizations and communities targeted.

4.2 Methods

4.2.1 Study design and survey

This study utilizes two years of survey data collected from the same households. A baseline survey was done in each site immediately following recruitment of house-holds into the program, and a follow-up survey was administered to the same households one year after baseline. The survey was administered to 358 farmer households at baseline (October, 2007); 181 households in Kiambu, Kenya, and 177 households in Arusha, Tanzania. In the one-year follow-up survey (October, 2008), five percent of households were lost to follow-up , mostly due to moving away. A total of 338 households were interviewed twice, 169 in each site.

Sampling frame

In each country, four districts were included in the study. Prior to the start of the study, program staff had selected four villages for the TAV program implementa-tion, which were located in four different districts. Program staff reported that the main selection criteria for villages where the program would operate were prior ex-perience producing horticultural crops, agronomic suitability for production, and proximity or access to markets. In keeping with the overall evaluation design, in each district, another village was selected based on similarity to the program village in size, wealth, proximity to markets, average household size, and public services availability. The survey was therefore administered in eight villages: four program villages, and four matched comparison villages. In each program commu-nity 25 households were interviewed, and in each comparison commucommu-nity, 20 were interviewed.

Respondent selection

In program villages, households to be interviewed were selected randomly from lists of program participants. In comparison villages, households were randomly se-lected from village census lists. The initial intent was to interview only households with children age 2-5 years, but that inclusion criterion was subsequently dropped due to many households who were recorded as having young children actually not having children in that age group. Therefore, households with preschool-age chil-dren were somewhat oversampled relative to the general population. Respondents within the household included three distinct roles: the person mainly responsible for agriculture, the person mainly responsible for TAV production, and the person mainly responsible for taking care of the reference child age 2-5 (or if no child, the person responsible for the majority of the cooking for the household). If a required

respondent for a particular questionnaire section was temporarily unavailable, the interviewer made an appointment to return later in the day or week to complete the questionnaire with the appropriate respondent.

Household survey administration

The survey was administered over a period of four to five weeks simultaneously in both sites, in order to ensure comparable seasonality. The follow-up survey was administered exactly one year after the baseline also so that changes seen in agricultural production and TAV use were not due to differences in seasonality.

Teams of interviewers administered the questionnaire in the local language of re-spondents, which was Kikuyu in Kenya, and Kiswahili, or occasionally Maasai, in Tanzania. During interviewer training, the importance of communicating consis-tent meaning of each question was emphasized, and role playing, discussion, and practice interviewing ensured that interviewers fully understood each question and agreed on its exact translation. The survey took one to two hours to complete for each household. Each day of the survey, research assistants and the PI reviewed the completed questionnaires, held a de-briefing session with enumerators, and returned any questionnaires with inadvertently missing information for follow-up.

Construct measurement

The questionnaire gathered information on household demographics, assets, in-come sources, agricultural production (particularly focused on the TAVs), diet, nutrition knowledge and attitudes about TAVs, medicinal knowledge and use of TAVs, and child age, weight, and morbidity. The questionnaire was pre-tested with 22 households and revised iteratively so that respondents’ understanding and responses to questions were clear and consistent.

The main indicators analyzed are medicinal use of the vegetables, and TAV consumption.

Medicinal use was measured using two survey questions. The first, designed to assess medicinal knowledge, asked, “Can you use any traditional leafy vegetables to treat/prevent any illness?” If the respondent answered yes, data were then collected on which illness was treated with which of the five TAVs. The second question, asked only in the follow-up survey, was, “Has consuming TAVs helped you or someone in your household with any particular illness personally?” Again if the respondent answered affirmatively, data were collected on the illnesses treated and the specific plants used, out of the five TAVs. An indicator of “number of unique illnesses treated with TAVs” was created by counting the number of unique illnesses a caretaker listed; if the respondent said “Amaranth and nightshade both are useful for anemia,” the response would be coded as only one unique illness. This was done because it was observed that a number of respondents said “all TAVS treat (illness),” a response which did not indicate in-depth medicinal knowledge. It was desirable for the purpose of this study, which sought to classify households based on extent of medicinal knowledge, to differentiate households with generalized knowledge from those who listed specific uses for each of the five TAVs. Binary (yes/no) indicators were also created for whether the household listed a TAV as a treatment for various specific classes of illness (iron-related, malnutrition-related, non-communicable/chronic disease related).

TAV consumption was measured using a 7-day food frequency questionnaire (FFQ) validated by HKI [219]. Respondents were asked how many days per week anyone in the household consumed each of the five TAVs, and if anyone consumed them, how many days the reference child age 2-5 (if present) consumed them. The caretaker was also asked to estimate the amount of each TAV the child consumed

on a typical eating occasion, using a standard bowl provided in the survey. From these data, average grams per day that the child consumed were calculated.

Wealth, an important socioeconomic control variable, was measured as a sum of real market values (by site) of 23 durable goods and livestock. Each household was asked how many they owned of each item.

Data entry and cleaning

Data were double-entered in CSPro data entry software [42], using electronic forms that looked identical to the printed survey questionnaire. Frequencies, histograms, and cross-tabulations were used to detect outliers, and implausible data were re-coded or removed. In some cases missing data could be found during the data cleaning process; for example if the number of TAVs grown was missing, but in the list of crops grown for the same household, a certain number of TAVs were listed individually as having been grown. Data consistency across years was improved by printing individualized follow-up questionnaires for each household, which included most demographic information obtained at baseline.