Extended Pilot Project Community Based Production of Complementary Food in Ethiopia

Overview

The purpose of the study is to evaluate the impact of a package of interventions including production and distribution of locally produced complementary foods (via so-called "Grain-banks") and Micronutrient Powders (MNPs) supplementation, and optimized food based Complementary Feeding Recommendation, on growth and micronutrient status of infants and young children. The impact of the intervention package will be evaluated in a quasi-experimental matched-control cluster design in infants and young children between 6 - 29 months of age. The impact, outcome, and output indicators of infants/children will be assessed in cross-sectional samples at baseline, after 9 and 18 months. A total of60 pair clusters are selected in which 15 households per cluster will be identified from the eligible population. A total of 1800 children from intervention and match-controlled clusters, will be sampled in the target age groups. Qualitative and quantitative data will be collected to gather information on Knowledge, Attitude and Practice (KAP), Infant and Young Child Feeding practices, anthropometry and anemia, following the impact pathway developed for the study. Research questions to be answered in this program evaluation are: 1. What impact does the extended pilot (implementation of local complementary food production and MNP) have on: 1. Growth in children 17-29 months of age 2. Infant and Young Child Feeding practices in children (6-23m), and 3. Anemia status in children 11-23 m? 2. What are determining factors for the impact/no impact related to: 1. Immediate outcomes: skills and capacity; knowledge, attitude and practices; and improved access 2. Intermediate outcomes: utilization; provision, and ensuring enabling environment 3. Program performance as measured by program monitoring data on output and activities?

The aim of this study is to assess the impact/effectiveness of the improved interventions on infant growth and micronutrient status so as to measure what would have happened in the absence of interventions. The study will be conducted in the same four regions in Ethiopia where the program intervention is taking place namely, Amhara, Tigray, Oromiya, and South Nations Nationalities and Peoples Region (SNNPR) regions. * Sample size calculations assume an expected difference of 0.2 Standard Deviation (SD) in HAZ, and 5% difference in feeding practices, 80% power, and α-error of 5%; and a cluster effect of approximately 2. * A total of 1800 children will be sampled in both intervention and non-intervention (control) villages. A total of 120 clusters are chosen for both groups. Each intervention cluster will be matched with a control cluster selected to be similar in geographical and ecological conditions, access to a health-care centre, status of food security and the existence of Community Based Nutrition (CBN) programme. Within each pair of clusters, households will be randomly selected after the complete listing of eligible households. A total of 15 households will be identified in each 120 clusters. * Demographics and socio-economic status indicators will be collected using Demographic Health Surveys (DHS) methods. * Infant and Young Child feeding indicators will be assessed in all children 6-23 months at baseline, mid- and endline * Knowledge attitude and practices concerning the intervention exposure indicators including perceptions and utilization of complementary food, MNP, and the use and perceptions on the grain banks will be assessed at baseline, mid- and endline in caregivers. Data analysis include: * For demographic and socio- economic characteristics of the study participants descriptive statistics will be used. * To calculate the nutritional status of children 6-23 months, Epi-Info/Emergency Nutrition Action (ENA) for SMART software will be used. * The median (min, max) of the weight or number of MNP consumed per day during the intervention period will be calculated as measure of compliance; however analyses will be done based on intention-to-treat. * Qualitative data will be recorded in digital recorders which will be transcribed first to local language then to English. Then specific themes will be identified and matrices will be prepared and data will be analysed. * The qualitative data will be analysed separately and triangulation will be made to compare the results. * All analysis will be done on an intention-to-treat basis, taking into account the matched-controlled design. For continuous variables we will use linear mixed models that include cluster, household and child as random effects to account for clustered observations. Fixed effects to be included in the model are covariates such as child's sex, age, household socio-economic status and relevant baseline values. Normality will be examined by creating a Quantile-Quantile (QQ) plot and subsequent visual inspection. Equal variances will be assured with Levene's test. * For categorical variables mixed-effects logistic regression models will be used with random effects for cluster and households.