This study is part of the World Vegetable Centre project entitled “Deploying Improved Vegetable Technologies to Overcome Malnutrition and Poverty” funded by the United States Agency for International Development (USAID) in the regions of Sikasso and Mopti. This project is a community-based intervention, which aims to contribute to reduce malnutrition, especially of children and their mothers, through diet diversification by promoting the production and consumption of vegetables as affordable sources of essential vitamins and micronutrients.
Study settings and target population
The rural regions of Sikasso and Mopti were selected as the study sites. The Sikasso inhabitants are mainly from the Bambara, Senufo, Miniankas, Fulani, and Samogo ethnic groups. Due to its favorable agricultural conditions, Sikasso is Mali’s biggest vegetable producer and considered one of the country’s most important bread baskets. The population of Mopti consists of the Dogon, Songhai, Bozo, Peulh, Bambara, and Tamaschek ethnic groups. The region is well watered in some parts and very dry in others, with agriculture and fishing being important economic activities. Mopti is a commercial crossroad between northern and southern Mali and its neighboring countries.
Study design and study population
This study was a community-based cross-sectional study conducted between January and March 2016 using a multistage sampling approach. Three districts in Sikasso (Bougouni, Koutiala, and Sikasso) and two districts in Mopti (Koro and Bankass, which are in the dry zone) were purposively selected for the study and in each district, four municipalities were randomly selected. The final stage was the purposive selection of villages that had an existing health center. Additionally, the selected villages needed to be at least 50 km apart from each other.
A census was performed in all the villages to establish a sampling frame. The households were the Primary Sample Unit (PSU). Therefore, all the children in a household fulfilling the age criteria were potential participants. During the census, information on the households (marital status and educational level of parents, the total number of children, and the number of children between 6 and 24 months, and so on) was collected. A specific number was attributed to each household to facilitate the identification of those included in the sample collection. Other exclusion criteria were children with severe congenital malformations, children with serious illnesses or complications requiring hospitalization, as well as children whose parents declined to participate in the study. A random sampling per village, based on the households, was performed to constitute the final sample size of the study.
Sample size calculation
The calculation of the sample size was based on the main objectives of the project, that is, improvement in the growth in status and reduced prevalence of diarrhea among the children of the intervention group compared with the control. The cluster effect was set at 1.5. We obtained a sample size of 2000 participants. This sample size was large enough to estimate the more prevalent malnutrition trait (stunting) at the national level at a 95% confidence interval and a 3% error precision.
Data measurements
A pilot study was conducted to test the survey forms and procedures and these were adapted as necessary. The data was collected by three-member survey teams, comprising of at least one female worker. All enumerators were either medical doctors or nurses with survey experience and at least one enumerator in each team had participated in the census. They were all trained and certified for this study. The data were collected using structured questionnaires via face-to-face interviews and anthropometric measurements.
Outcome variable
The nutritional status of children aged 6–24 months, expressed as the prevalence of underweight, stunting, and wasting, was assessed using anthropometric variables such as height, Mid Upper Arm Circumference (MUAC), and weight. The weight was recorded on children wearing minimal clothing and bare feet using a standard calibrated weighing Uniscale (Seca®, Hamburg, Germany) in kilograms, to the nearest 0.1 kg. The height was taken using a stadiometer (Schorr®, UNICEF) in centimeters to the nearest 1 mm. The height and weight were taken twice and a difference of 0.1 cm in height and 100 g in weight was accepted as normal. The MUAC was measured thrice using non-stretchable tape on the left mid-upper arm to the nearest 1 mm.
Based on the recorded weight, the nutritional status was graded as per the 2006 WHO child growth standards using WHO Anthro version 3.2.2 software [18]. The children were considered stunted, wasted, or underweight if the height-for-age Z-score, the weight-for-age Z-score, or the weight-for-age Z-score was less than − 2 SDs (Standard Deviation) using the new WHO child growth standards, while those children with a score equal to or greater than − 2 SD were considered normal [18, 19]. During data processing, the exclusion criteria were applied to the anthropometric data of the children based on WHO recommendations to remove data that are most likely to be erroneous (HAZ and WAZ were excluded if the child value was <− 6.00 or > 6.00. The WHZ was excluded if the value was <− 4.00 or > 6.00). A MUAC below 12.5 cm indicated acute undernutrition and a value below 11.5 cm indicated severe acute undernutrition [20].
Explanatory variables
Three main types of questionnaires (for households, for mothers, and for children) were designed to record the data on any indicators of household socioeconomic and socio-demographic status, household food security, and care practices for children and their mothers.
The Household Food Insecurity Access Scale (HFIAS) was used to determine the household food insecurity including all nine generic questions that require recollection about the worry of food availability and accessibility in the previous months [21, 22]. The responses were summed to create a total score between 0 (the most food secure household) and 27 (the most food insecure household), which was determined using the Household Food Insecurity Access Prevalence (HFIAP) status indicator as a proxy of the household food insecurity prevalence [21]. Each household was then classified as either food secure, or mildly, moderately, or severely food insecure, based on the Food and Nutrition Technical Assistance (FANTA)'s recommended cut-offs. Information on durable assets (cupboard, hurricane lamp, radio, bicycle, boat, telephone, refrigerator, motorcycle, car, and so forth) and the materials of the dwelling structure were used to construct a relative index of the household wealth (asset) status using principal components analysis [23]. For each amenity available in the household, a score based on the Health/Nutrition/Population/Poverty Thematic Group of the World Bank for Mali was given and their sum was used as the household amenities score [24]. From the total household score, quartiles were computed and four wealth classes were defined, from the poorest (first quartile) to the richest (fourth quartile). The household dietary diversity scores (HDDS) were assessed using standard tools which collected information on the number of different food groups consumed over a given reference period among the 12 different food groups (cereals; roots and tubers; vegetables; fruits; meat, poultry, offal; eggs; fish and seafood; pulses/legumes/nuts; milk and milk products; oil/fat; sugar/honey; and spices/condiments/beverages) [25, 26]. The number of months during which the household was unable to meet its food needs during the last 12 months was collected using a structured questionnaire developed by FANTA [27]. Other household indicators included household possession of a latrine (available or not available), the presence of a fence around the household, the educational level of the household’s head, livestock, and the disposal of garbage.
The characteristics of the mother included the maternal age in years, educational level, current occupation, number of children, and vegetable intake. The educational status was measured according to the education levels in Mali: no formal education, having a primary level education, or secondary level and above.
The characteristics of the children included gender, the term of delivery (< 37 weeks or ≥ 37 weeks) and age in months; immunization against BCG and Penta 3 + Polio; deworming status in the past 6 months (yes or no); and a history of illness episodes. To assess childhood illnesses, the mothers were asked whether their children had been affected by diarrhea, fever, or coughs in the past 2 weeks. Diarrhea was defined as having three or more loose or watery stools in a 24-h period in the 2 weeks prior to the survey [28]. Assessment of feeding covered breastfeeding (yes or no) and dietary diversity of the children. The quality of complementary feeding was assessed using the individual dietary diversity score adapted for children, with 8 items instead of 12 as was the case for the household [26].
Statistical analyses
Data were recorded using the Epi-Data® version 3.1 software and subjected to statistical analysis using the Stata® 11.1 software. Statistical procedures were adapted to the sampling methods (multi-stage random sampling with clusters being the primary sampling unit). For each participant included in the study, a weighting of the inverse multiple stage probability of being selected in the study was computed and the survey data analyses procedures of the Stata software were used.
Descriptive statistics were computed and the results were presented as means (95% confidence intervals) or medians (25th–75th percentiles) for the quantitative variables or as percentages for the qualitative variables. Comparison between the means or percentages was performed using a Student’s t-test or a chi-squared test, taking into account the complex sampling frame. Bivariate analyses were done for the four outcome variables: stunting, underweight, and wasting (using both WHZWHO and MUAC) separately. The independent variables with a p-value less than 0.20 for at least one of the nutritional parameters during bivariate analyses were selected as the candidates for multivariable analyses. Multivariable binary logistic regressions were fitted to identify the determinants of underweight, stunting, and wasting, separately. Associations between the dependent and independent variables were assessed using an Odds Ratio (OR) and a 95% confidence interval (CI). The tests were two-sided, and the statistical associations were declared significant if the p-values were less than 0.05.
Ethical considerations
The protocol of the present study was approved by the Ethical Board of the University of Bamako, Faculty of Medicine, Pharmacy, and Odonto-stomatology of Mali (N°2016/44/CE/FMPOS). The objective of the study, the confidentiality, and the right to withdraw at any time without facing any consequences were explained to each participant in the local language at the time of recruitment. Written informed consent was obtained from each participant before any study enrollment. Every mother-infant pair meeting the inclusion criteria was considered for the study after obtaining the informed consent of the mother. Informed consent was acquired from a legal guardian for participants under 16 years of age at the time of the study. Participants with diarrhea, respiratory tract infections, and undernutrition were referred to health institutions and organizations working on nutrition.