Socioecological predictors of breastfeeding practices in rural eastern Ethiopia

Background: Estimates by the World Health Organization indicate that over 800,000 global neonatal deaths each year are attributed to deviations from recommended best practices in infant feeding. Identifying factors promoting ideal breastfeeding practices may facilitate efforts to decrease neonatal and infant death rates and progress towards achieving the Sustainable Development Goals set for 2030. Though numerous studies have identified the benefits of breastfeeding in reducing the risk of childhood undernutrition, infection and illness, and mortality in low- and middle-income countries, no studies have explored predictors of breastfeeding practices in rural eastern Ethiopia, where undernutrition is widespread. The aim of this study is to examine predictors of infant feeding practices in Haramaya, Ethiopia, using a multi-level conceptual framework. Methods: This study uses data collected from household questionnaires during the Campylobacter Genomics and Environmental Enteric Dysfunction (CAGED) project among 102 households in the Haramaya woreda, Eastern Hararghe Zone, Eastern Ethiopia, and investigates factors influencing breastfeeding practices: early initiation, prelacteal feeding, and untimely complementary feeding. Results: Nearly half (47.9%) of infants in this study were non-exclusively breastfed (n = 96). Generalized liner mixed effects models of breastfeeding practices revealed that prelacteal feeding may be a common practice in the region (43.9%, n = 98) and characterized by gender differences (p =.03). No factors evaluated were statistically significantly predictive of early initiation and untimely complementary feeding (82% and 14%, respectively). Severely food insecure mothers had more than 72% lower odds of early breastfeeding initiation, and participants who self-reported as being illiterate had 1.53 times greater odds of untimely complementary feeding (95% CI, [0.30,7.69]) followed by male children having 1.45 greater odds of being untimely complementary fed compared to female (95% CI,[0.40,5.37]). Conclusions: This study found high rates of prelacteal feeding and low prevalence of exclusive breastfeeding, with girls more likely to be exclusively breastfed. While no predictors evaluated in this multi-level framework were associated with prevalence of early initiation or complementary feeding, rates may be clinically meaningful in a region burdened by undernutrition. Findings raise questions about gendered breastfeeding norms, the under-examined role of khat consumption on infant feeding, and the complex factors that affect breastfeeding practices in this region. This information may be used to guide future research questions and inform intervention strategies. Our breastfeeding practices model was developed as a conceptual framework that merges the socioecological model [41] with UNICEF’s conceptual framework on the causes of undernutrition [42]. As described in the introduction, the UNICEF model was adopted and modified to understand: 1) basic, 2) underlying, 3) immediate, and 4) physiological factors that impact breastfeeding practices in Haramaya. In the UNICEF model, suboptimal breastfeeding falls under “care practices” and is an underlying cause of undernutrition. Our model (Fig. 1) integrates these undernutrition factors with breastfeeding practices as our outcomes of interest. Further, our model incorporates factors at the physiological level which have been described in the literature to impact breastfeeding. While the model is certainly not exhaustive, it serves as a framework for conceptualizing how multilevel factors such as economic context, empowerment, and delayed breastfeeding initiation may affect breastfeeding practices in the study population. Breastfeeding practices in Haramaya are largely unknown and thus factors in each level of the model are derived from current literature on predictors of infant feeding in other regions, including in Ethiopia. Socioecological predictors of breastfeeding practices and infant and young child feeding At the center of our conceptual model is breastfeeding practices. These practices include early initiation, prelacteal feeding, colostrum avoidance, exclusive breastfeeding, and untimely complementary feeding. These practices are determined based on recommended and discouraged infant feeding practices reported in the literature. Optimum feeding practices include breastfeeding immediately after birth (early initiation), offering no other foods or liquids (exclusive breastfeeding), and initiating complementary foods no earlier than 6 months of age. The WHO also discourages introducing complementary foods later than 6–8 months of age, as breast milk is then insufficient for the nutritional needs of the infant. Prelacteal feeding is also a discouraged practice, defined as offering food or drinks (such as water, honey, formula milk, or fruit juice) to a newborn prior to establishing breastfeeding. CAGED formative research included rapid ethnography, full household enumeration within the study region, household questionnaires, anthropometric measurement of children, and collection of samples for laboratory analysis. The primary aims, research questions, and results have been published elsewhere [4, 5, 43]. Using data collected through household questionnaires during the formative research, this study examines breastfeeding practices in Haramaya woreda. Full details of data collection methodology have been previously published [5] but are presented briefly below for ease of reference. Haramaya woreda is a semi-arid region of East Hararghe Zone, Oromia Region, where most of the population practices mixed crop-livestock livelihoods. High rates of extreme poverty burden the region, and women are predominantly responsible for both infant and livestock housing, safeguarding, feeding, and healthcare [4]. It is a district characterized by large family sizes, high fertility rates, and very high childhood undernutrition rates [5]. The region is highly affected by climate change, increasing population pressure, and reducing land and water availability. Khat production is pervasive and has partially or fully replaced grain and vegetable production in many households across the region [12]. Within this context, households were targeted for inclusion in the study (see details below); while men, women, and children were participants in the overall study [5] data included in analysis here all come from surveys conducted with women in those households. In Ethiopia, kebeles are the smallest geographic administrative units recognized in census mapping. Five of the twelve Haramaya kebeles were selected for inclusion in the formative research based on maximized geographical distance between the kebeles. Household surveys were conducted with men and women in randomly targeted households across the five Haramaya kebeles. Eligibility criteria of the household included the presence of at least three chickens in the homestead (defined as a collection of households, often extended family, that share common areas), non-participation in any other animal husbandry project, and having lived in Haramaya woreda for at least three months. Eligibility criteria of the child included the absence of any visible congenital anomalies, absence of an extended hospital stay for mother or child following the child’s birth, and child age of 11–13 months. A total of 102 households were randomly selected and completed the household survey. Data were collected between September and December 2018. The sample size was based on the primary research objectives of the formative research; thus, this study is not powered to measure maternal and infant health or breastfeeding outcomes. Research presented here is covered under the CAGED study protocol that underwent ethical review and received approval from the Haramaya University Institutional Health Ethics Research Review Committee (IHRERC/152/2018), the Ethiopia National Research Ethics Review Committee (MoST/3–10/168/2018), and the Institutional Review Board at the University of Florida (201,703,252). The household survey was developed in collaboration with the University of Florida (UF) and Haramaya University (HU) teams. A draft was developed initially by the UF team, then shared electronically with the HU team, who made substantive revisions in early 2018. The modified version was entered in April/May of 2018 into REDCap electronic data capture tools hosted at University of Florida Clinical and Translational Science Institute [44]. REDCap draft survey was loaded onto Samsung Galaxy Tablets, where all data entry was conducted. Substantial revisions came from a three-step piloting and revision process that occurred during a two-week training workshop at Haramaya University in August 2018. First, three Ethiopian data collectors recommended a round of revisions after reading the tablet-based survey independently and with each other. A second round of revisions was provided by an extended team of social science and public health experts from UF and HU who went over the survey together, line by line. The team took two full days to review the survey, validate local examples, identify appropriate food choices, brand names, and vitamins available, and reword or remove questions that were not culturally appropriate or relevant. The team also discussed at length how best to ask each question, including what translation into Afan Oromo would be most appropriate to preserve the meaning of questions. Finally, the three data collectors and supervisor field-tested the survey within local non-eligible communities near Haramaya University. Revisions were made to the final survey. The primary outcome variables for this study were 1) early breastfeeding initiation, 2) prelacteal feeding, and 3) untimely complementary feeding. Early initiation was coded dichotomously as whether the infant was put to the breast within one hour of birth. Prelacteal feeding was coded as a dichotomous variable, defined by whether the child was given anything to drink other than breast milk during the first 3 days after delivery. Untimely complementary feeding was measured by identifying the child’s age when first given something to eat other than breast milk (either early – before 6 months or late – after 8 months). Exclusive breastfeeding was analyzed only descriptively, as this practice is a composite of other breastfeeding practices in the analysis but is defined as the infant receiving breast milk and nothing else for the first 6 months of life. Several standard predictor variables used in this study (ie. maternal age, and number of antenatal care visits) were generated using single questions from the household survey. However, composite variables were also used; further detail on methods used to generate those variables is included below. Time poverty was monitored using the Women’s Empowerment in Agriculture (WEIA) Index. Participants are asked to define number of hours (sum amount of time) of work-related tasks (ie. employment, own business work, farming, cooking) performed 24 h prior to a household interview [45]. Individuals who worked more than 10.5 h in the day were identified as time impoverished [45]. Whether the participant’s primary source of income was from khat was measured by a single question in which the participant was asked to identify their primary source of livelihood (animal production, crop production, khat production, petty trade, remittances, or other). Severe food insecurity was measured using the Household Food Insecurity Access Scale (HFIAS) [46] with items which asked whether the participant experienced at least one of these often (more than ten times in the past month): running out of food, going to bed hungry, or going a whole day and night without eating. Food secure, mildly food insecure, and moderately insecure households were not identifiable as our questionnaire included a subset of questions for severe food insecurity from the HFIAS. Statistical analysis was conducted using R version 4.0.2. Descriptive statistics were conducted to characterize breastfeeding practices, potential covariates, and additional factors such as food insecurity. Bivariate logistic regression was conducted on our breastfeeding outcomes. Untimely complementary feeding was disaggregated to early and late complementary feeding in our bivariate analysis and aggregated for multivariate models. Predictors that met the established cut-off (p-value < 0.20) (Table ​(Table2)2) were included in the multivariate analysis. Given the sample size, a backward selection approach was taken to conduct generalized linear mixed effects modeling (GLMM) to account for any dependency between clustered covariates and response variables. The lme4 package in R was utilized for GLMM analysis. Models were created to understand the effect of potential confounders (ie. mother’s age, literacy, child sex, kebele) on each outcome. To select the best fitting model and assign covariates as a fixed or random effect, variance, Akaike information criterion, Bayesian information criterion and log likelihood scores were evaluated. Mother’s age, literacy, and child sex were considered as fixed effects while kebele was included as a random effect in all models. Odds ratios were calculated with a 95% confidence interval to determine the relative odds of targeted breastfeeding practices occurring given exposure to variables of interest. Early initiation of breastfeeding mixed model regression summary (n = 94) 1Kebele included as a random effect