Study area and data source

The PMA-ET 2019 Ethiopia study employs a two-stage cluster design to gather data on pre-lacteal feeding practices in Ethiopia. The study categorizes residential areas as urban or rural and sub-regions as strata, covering all 11 geographic regions in Ethiopia. The target population for the study is women aged 15–49 years, with approximately 95% of the population residing in five regions: Addis Ababa, Amhara, Oromia, SNNP, and Tigray. The remaining regions are grouped as a sixth synthetic region called “other” due to population distribution disparities and limited resources.

To estimate pre-lacteal feeding rates with a margin of error below 2% at the national level, the study selects 238 EAs for the fourth round sample. The margin of error is set below 3% for urban and rural estimates and below 5% for each of the five regional levels. The study finds that individual-level factors such as educational status and household wealth, as well as community-level factors such as wealth status, are associated with pre-lacteal feeding practices in Ethiopia.

Given these findings, targeted interventions that address these factors should be developed to reduce the prevalence of pre-lacteal feeding. Programs that aim to improve maternal education and financial stability may be effective, as well as community-level interventions that address poverty and inequality. Further research is needed to identify the most effective strategies for improving infant and young child feeding practices in Ethiopia. The secondary data for this analysis were obtained from PMA-ET of 2019 which was found in the PMA portal (https://www.pmadata.org/_2019).

Variable measurement

The dependent variable for pre-lacteal feeding was divided into two categories: “Yes/No”. Mothers who intended to use pre-lacteal feeding during the interview were labeled as “Yes”, while those who did not use it during the interview were labeled as “No”.

Variables at the individual level

include maternal age, educational status, wealth status, Anti-natal care, and desired delivery place.

At the community level, variables

include Region, place of residence, community education, and community wealth status.

Data processing and analysis

Before recording, labeling, and exploratory analysis using Stata/SE version 17.0, data cleaning was conducted to ensure consistency and completeness. Descriptive statistics were used to present frequency distributions in tables and text. To account for potential disparities in geographical strata selection and non-responses, a sample weight was applied.

After confirming the eligibility of the data for multilevel analysis (i.e., Intra-cluster Correlation Coefficient (ICC) greater than 10% (ICC = 63.4%)), a multilevel analysis was performed. Since the PMA-ET data had a hierarchical structure with individuals (level 1) nested within communities (level 2), a two-level mixed-effects logistic regression model was utilized. This model estimated both the independent (fixed) effects of the explanatory variables and the community-level random effects on the plan for Pre-lacteal feeding. The log of the probability of plan to Pre-lacteal feeding was modeled using a two-level multilevel model.

A preliminary bivariable multilevel logistic regression was conducted, selecting variables with a p-value less than 0.25 before creating three models (models 1–3). Following this, four stages were carried out: Model 0 (an empty or null model without explanatory variables), Model 1 (solely individual-level factors), Model 2 (solely community factors), and Model 3 (both individual and community-level factors). The measures of association (fixed-effects) estimate the connections between the likelihood of women planning Pre-lacteal feeding and various explanatory variables, expressed as Adjusted Odds Ratio (AOR) with their corresponding 95% confidence interval. A variable with a p-value less than 0.05 was considered statistically significant. The measures of variation (random effects) were reported using ICC, Median Odds Ratio (MOR), and proportional change in variance (PCV) to assess the variation between clusters.

The ICC demonstrates how community characteristics affect the variation in plans for Pre-lacteal feeding for mothers. A higher ICC indicates that community characteristics are more important in understanding individual variation in these plans. MOR is the median value of the odds ratio between the highest and lowest risk areas, showing how much residential area determines the probability of Pre-lacteal feeding plans for mothers. PCV measures the total variation attributed to individual and area-level factors in the multilevel model. Multicollinearity was checked among independent variables using a standard error cutoff point of ± 2, and no multicollinearity was found. The log-likelihood test was used to assess the goodness of fit of the adjusted final model compared to previous models (individual and community-level adjustments).