Data source

The data was obtained from the measure DHS program at after preparing concept notes on the project. Demographic and Health Survey (DHS) data were aggregated from the 32 countries of Sub-Saharan Africa (SSA) from 2010 to 2020. The Sub-Saharan African continent consists of 54 recognized countries. Geographically, Sub-Saharan Africa is a region located south of the Sahara Desert on the African continent. Sub-Saharan Africa, according to the United Nations (UN), includes all African countries that are wholly or partially located south of the Sahara. As part of sub-Saharan Africa, the United Nations Development Program recognizes 46 of the 54 African countries, while the World Bank mentions Somalia and Sudan. The recent DHS of the country-specific dataset was extracted during the specified period.

In this study, 34 countries in the sub-region met our selection criteria (countries in sub-Saharan Africa that had DHS datasets between 2010 and 2020) available in the public domain. The countries were Angola, Benin, Burkina Faso, Burundi, Cameroon, Ivory Coast, Comoros, Congo Brazzaville, Democratic Republic of Congo, Ethiopia, Gabon, Gambia, Ghana, Guinea, Kenya, Lesotho, Liberia, Malawi, Mali, Namibia. , Niger, Nigeria, Rwanda, Senegal, Sierra Leone, Tanzania, Uganda, Zambia and Zimbabwe.

The DHS program adopts a standardized method involving uniform questionnaires, manuals and field procedures to collect comparable information between countries around the world. DHS are nationally representative household surveys that provide data from a wide range of population, health and nutrition impact monitoring and evaluation indicators with face-to-face interviews with women aged 15-49. , random sampling plan. Information was obtained from eligible women aged 15-49 in each country. The detailed survey methodology and process used to collect the data has been recorded elsewhere. [28].


Result variable

The outcome variable, early/rapid initiation of breastfeeding, was determined by asking mothers for details of when their babies were placed on their breasts after birth. The ratio of the number of children put to the breast within one hour of birth to the total number of children was used to calculate the prevalence of early initiation of breastfeeding.

Independent variables

Socio-demographic and economic variables (residence, region, maternal age, marital status, religion, maternal education, paternal education, wealth index, maternal occupation/mother’s activity status), pregnancy and pregnancy-related factors (consultation prenatal, parity, previous interbirth interval, use of contraceptives, place of delivery, birth order, mode of delivery, size of the child at birth). Behavioral factors.

(smoking, media exposure) were included in this study.

Community level variables

Non-aggregated community variables were place of residence and region. The place of residence was recorded as rural and urban. The region has been described as the province from which a child comes. By aggregating from an individual level, another group of community-level variables was developed using average approaches to conceptualize the neighborhood effect on EIBF implementation. Education of neighborhood women, community poverty, community visit to the ANC, community birthing place.

Data management and analysis

The research for this thesis was performed using STATA version 15 (STATA Corporation. IC., TX, USA). For the calculation of descriptive statistics such as proportions, sampling weights were used to account for the non-proportional distribution of the sample to the strata. In the case of standard regression models, research participants are considered independent of the outcome variable. Nevertheless, the units of the same category are rarely independent when the data are hierarchical [29]. Units in the same frame (cluster) are more similar to each other in relation to other units, or in relation to the outcome of interest, than units in another frame. This can then lead to a violation of the independence assumption which could have the effect of underestimating standard errors and increasing Type I error rates (increases the false positivity rate of our results). In such circumstances, multilevel modeling can simultaneously take into account individual and community level variables and provide a more comprehensive understanding of the factors of early breastfeeding initiation. [30].

Multi-level analysis

Multilevel models are therefore developed to overcome the analytical problems that arise when the data is organized hierarchically and the sampled data is a sample of several stages of this hierarchical population, such as the DHS, in which children are nested in households and households are nested in clusters. , and there is an intra-group correlation. In order to estimate both the independent (fixed) effects of the explanatory variables and the random effects at the community level on the onset of prelacteal feeding, a two-level mixed-effects logistic regression model was fitted. The person (children) is the first level and the cluster is the second level (community). In the bivariate multilevel logistic regression model, individual and community variables associated with early initiation of breastfeeding were tested independently and variables that were statistically significant at p– the value 0.20 was considered for the final adjustments at individual and community level. In multivariate multilevel analysis, variables with p-value

Therefore, using the two-level multilevel model, the record of the probability of implementation of pre-milk feeding was modeled as follows:

$$mathrm{log}left(frac{{pi }_{ij}}{1-{pi }_{ij}}right)={beta }_{0}+{beta }_{1}{X}_{ij}+{beta }_{2}{Z}_{ij}+{mu }_{j}$$

where, i and j are the level 1 (individual) and level 2 (population) units respectively; X and Z apply to individual and community level variables, respectively; ({pi}_{ij}) is the probability of having prelactated sucklings in the je community for the ie mother; the βs are the fixed coefficients – hence there is a corresponding efficiency for each unit increase of X/Z (a set of predictor variables). While in the absence of control of the predictors, ({beta}_{0}) is the interception—the effect on the mother’s likelihood of prelacteal feeding; and MI indicates the random effect for the je community (effect of community on mother’s decision to provide pre-dairy feeding). The existence of clustered data and variations within and between communities were taken into account by assuming that each community has a different intercept (({beta}_{0})) and fixed coefficient (β).

Model making

A total of four models were installed. The first was a null model without exposure variables, which was used to determine population-level random effects and assess heterogeneity in the community. Then, Model I was the fit of the multivariate model for the individual-level variables and Model II which was fitted for the community-level factors. In Model III, the outcome variable was equipped with potential candidate variables at both the individual level and the community level.

Parameter estimation method

Fixed effects (a measure of association) were used to estimate the relationship between the odds of IBFE and explanatory variables at the population and individual level, and the results were expressed as odds ratios with a 95% confidence interval. Community-level variance with standard deviation, intracluster correlation coefficient (ICC), proportional change in community variance (PCV), and median odds ratio (MOR) were used as indicators of heterogeneity (random effects). The median odds ratio (MOR) is used to transform the area-level variance into the commonly used odds ratio (OR) scale, which has a consistent and intuitive interpretation. During the random selection of two zones, the MOR is defined as the median value of the odds ratio between the zone with the highest risk and the zone with the lowest risk. The MOR can be conceptualized as the increased risk that (in median) would have if he moved to another area with higher risk. It is determined by (MOR={e}^{sqrt{(2times VA)}times 0.6745}) [31]. Where; VA is the variance of the region norm and 0.6745 is the 75th percentile of the cumulative distribution function of the normal distribution with mean 0 and variance 1, see detailed definition [28]. While the proportional variance lag is determined as [29] (PCV=[(VA-VB)/VA]*100%), where; VA = variance of the original model and VB = variance of the model with more terms.