Trends and determinants of breastfeeding within one hour in Ethiopia, further analysis of Ethiopian Demographic and Health Survey: multivariate decomposition analysis

Background Despite the substantial efforts to improve timely/early initiation of breastfeeding, avoidance of colostrum, and delayed initiation of breastfeeding remains a big challenge in developing countries. Therefore, this study aimed to analyze the trends of early breastfeeding rate over time based on the Ethiopian Demographic and Health Survey (EDHS). Methods Secondary data analysis was conducted based on the Ethiopian Demographic Health Surveys (EDHSs) conducted in 2005, 2011, and 2016. A total weighted sample of 9, 111, 10,106, and 8564 in 2005, 2011, and 2016 respectively were included for analysis. Trend and Logistic based decomposition analysis technique was used for analyzing the trends of early breastfeeding initiation over time and factors contributing to the change in early breastfeeding initiation rate. STATA 15 was employed for data management and analyses. All analyses presented in this paper were weighted for the sampling probabilities and non-response. Result Among children age less than 5 years the rate of early breastfeeding initiation rate overtime was increased from 70.5% in 2005 to 72.7% in 2016. The highest rate of improvement was seen in the second phase of the study (2011–2016) while it shows a decline in the first phase (2005–2011) from 70.5 to 55.1%. The decomposition analysis indicated that about half of the overall change in early breastfeeding initiation rate was due to the difference in women’s composition. Particularly, an increase in health facility delivery and vaginal delivery was a significant predictor of the increasing rate of early breastfeeding initiation over the surveys. Conclusion Early initiation of breastfeeding slightly increasing over the last 10 years in Ethiopia. Half of the overall increase in the early initiation of breastfeeding was due to the change in compositional characteristics of women over 10 years in Ethiopia. Change in the composition of women according to health facility delivery and vaginal delivery were the major source of the increase in early breastfeeding initiation over time. Public interventions including promoting health facility delivery of women for further improvements of early breastfeeding initiation should be needed.


Introduction
Early initiation of breastfeeding after birth is an essential intervention to reduce neonatal mortality and morbidity [1][2][3]. Despite the substantial efforts to improve timely/early initiation of breastfeeding, avoidance of colostrum, and delayed initiation of breastfeeding remains a big challenge in developing countries [1,4,5]. The World Health Organization (WHO) early initiation of breastfeeding within 1 hour as it confers many benefits to the child and prevents neonatal mortality that extends into adulthood [6,7]. Early initiation of breastfeeding stimulates the production of breast milk and ensures greater consumption of highly nutritious colostrum breast milk produced during the first few days after birth [1,8,9]. Early initiation of breastfeeding also an important bridge for the mother-to-child relationship and to increase the length of breastfeeding [10][11][12]. In addition to these advantages, early initiation of breastfeeding induces uterine contraction after pregnancy, decreasing the risk of postpartum haemorrhage and extending the duration of postpartum infertility, allowing women to return to their presentational weight, as well as reducing the risk of breast and ovarian cancer [9,11,13]. Globally, the prevalence of early initiation of breastfeeding ranged from 17.7 to 98.4% with an average of 57.6%, a slightly lower prevalence among mothers with complications during pregnancy and caesarean delivery [5].
Globally, an estimated 4 million newborns die each year from preventable infectious disease, which accounts 41% of mortality among under-five children [14]. In addition, most of the deaths occur within the first week of life, which is considered to be the early neonatal period [15]. The risk of newborn death during the first day of life was approximately 1 % globally, of which twothirds of the burden already exists in ten countries, including Ethiopia [2,16]. In recent decades, progress towards improving early initiation of breastfeeding has been sluggish, with the average rate of global initiation only increasing by 14%age points. However, less than half of all newborns are placed to the breast within an hour of birth globally. In the least developed region, the percentage of early initiation of breastfeeding was 53% [17], while the total pooled prevalence of early initiation of breastfeeding in Ethiopia was 61.4% [2]. Different studies on determinants of early initiation of breastfeeding showed that rural residence, educational status, mode of delivery, use of antenatal care (ANC), marital status, and place of delivery were important predictors of early initiation of breastfeeding [2,3,[18][19][20][21]. Only point survey data was analyze in previous studies [11,[20][21][22][23][24][25][26]; it was difficult to see the patterns and to identify potential factors that have been consistently influenced early breastfeeding initiation over time in Ethiopia.
Studying the change in early initiation breastfeeding using multivariate decomposition analysis to identify predictors associated with the change in early initiation breastfeeding overtime becomes relevant for targeting interventions to improve early initiation of breastfeeding rate over time and could critically inform policies and strategies aimed to rise early initiation of breastfeeding rate in Ethiopia. Therefore, this study aimed to evaluate trends and determinants of early initiation of breastfeeding rate in Ethiopia. Twenty-one sampling strata were established because the area of Addis Ababa was entirely urban. In the first phase, 540 Enumeration Areas (EAs) (145 in urban areas) were selected for EDHS 2005, 624 EAs (187 in urban areas) were selected for EDHS 2011, and 645 EAs (202 in urban areas) were selected for EDHS 2016 by probability proportional to the size of the enumeration region and independent selection of each sampling stratum. At the second stage, a complete household listing operation was carried out in all selected EAs prior to the start of the fieldwork and an average of 28 households were selected systematically. The detailed sampling strategies are fully presented in the EDHS report [27][28][29]. The source population is all under-five children in Ethiopia, whereas the study population was all underfive children in the selected enumeration region.

Outcome variable
The outcome variable was early initiation of breastfeeding practice (EIBF) among mothers with children younger than 24 months of age. The initiation of breastfeeding within 1 h of birth is an early initiation of breastfeeding. Initiating breastfeeding was coded as '1' within 1 h of birth, while beginning breastfeeding was coded as '0' for logistic regression analysis after 1 h of birth.

Independent variables
Socio-demographic and economic variables (residence, region, maternal age, marital status, religion, maternal education, paternal education, wealth index, maternal occupation, maternal working Status), Pregnancy and pregnancy-related factors (ANC visit, Parity, Preceding birth interval, contraceptive use, Place of delivery, Birth order, Mode of delivery, wanted pregnancy). Behavioural factors.
(Smoking, media exposure) were included in this study.

Data collection procedure
The study was conducted based on EDHS data by accessing from the DHS program official database www. measuredhs.com after permission was granted through an online request by explaining the objective of our study. The raw data was collected from all parts of the country on childbearing aged women using a structured and pre-tested questionnaire. We used the kid Record (KR file) data set and extracted the outcome and independent variables.

Data management and analysis
Before any statistical analysis, the data were weighted using sampling weight, primary sampling unit and strata to restore the representativeness of the survey and to inform STATA to take account of the sampling design when calculating standard errors to obtain reliable parameter estimates. Since the response variable have binary outcome, binary logistic regression model was employed. Sampling was applied as part of a complex survey design using primary sampling unit, strata, and women's individual weight (V005). Data from 2005, 2011, and 2016 were appended together after extracting relevant variables for trend and decomposition analysis. Cross tabulation and summary statistics were performed to describe the study population using STATA 14 software.

Trend and decomposition analysis
The trend was assessed using descriptive statistics stratified by selected background characteristics of respondents and was assessed separately in the periods 2005-2011, 2011-2016, and 2005-2016. A multivariate analysis of the shift in the early initiation of breastfeeding rate was employed to address possible factors contributing to the difference in the percentage of early initiation of breastfeeding over the study era. This approach is used for many purposes in economics, demography, medicine, and other specialties. This research focused on how early initiation breastfeeding rate reacts to the variation in women's characteristics and how these variables form the difference across surveys conducted at different times. The study was a regression analysis of the difference in the percentage of early initiation of breastfeeding rate between EDHS 2005 and 2016. Decomposition analysis aimed to determine the cause of the discrepancy in the percentage of EIBF in the last 10 years. Both the difference in composition (Endowment) of the population and difference in the effect of characteristics (Coefficients) between the surveys is necessary to know the factors contributing to the increase/decrease in the EIBF rate over time. The multivariate decomposition analysis for the nonlinear response model utilizes the output from the logistic regression model since it is "a binary outcome" to parcel out the observed difference in the EIBF rate between the surveys into components. The difference in the rate of EIBF between the surveys can be attributed to the compositional difference in population (difference in characteristics or endowment) and the difference in the effect of explanatory variables (difference in coefficients) between the surveys.
Logit based decomposition analysis technique was used for the analysis of factors contributing to the change in EIBF over time to identify factors contributing to the change in the EIBF rate in the last 10 years. The variations of EIBF over time can be due to the compositional disparity between surveys and discrepancies in the results of the chosen explanatory. Hence, the observed variation in EIBF between surveys is additively decomposed into a characteristics (or endowments) component and a coefficient (or effects of characteristics) component. For logistic regression, the Logit or log-odd of EIBF is taken as: The E component refers to the part of the differential owing to differences in endowments or characteristics. The C component refers to that part of the differential attributable to differences in coefficients or effects [30].
The equation can be presented as: The recently developed multivariate decomposition for the non-linear model was used for the decomposition analysis of early initiation of breastfeeding using the mvdcmp STATA command [30]. In this study, variables with p-value < 0.2 in the bivariable multivariate decomposition analysis were considered for the multivariable multivariate decomposition analysis. In the multivariable multivariate decomposition analysis variables with p-value < 0.05 in the endowments and coefficients were considered as a significant contributing factor for the change in early initiation of breastfeeding over time.

Characteristics of the study population
In the age group of fewer than 20 years, nearly threefourths of respondents were found in all three surveys. Based on ANC visits, the number of ANC visits in the first quarter increased from 29% in 2005 to 37% in 2016. With regard to maternal education, 76% of women were not taught in 2005, compared to 65% in 2016. In addition, the percentage of primary education for husbands and spouses grew from 26% in 2005 to 36% in 2011, but decreased marginally to 33% in 2016 ( Table 1).
The media representation of respondents increased from 37% in 2005 to 68% in 2016, and the number of women without an ANC visit fell dramatically from 72% in 2005 to 24% in 2016. The proportion of respondents' institutional delivery was increased from 5.  (Table 2).

Decomposition analysis Decomposition analysis of early breastfeeding initiation, 2005-2016
Overall, there have been increases in breastfeeding rates within 1 hour in Ethiopia from 2005 to 2016. The overall decomposition result suggested that the increases in breastfeeding rate within 1 hour has been explained by both the disparity in women's characteristics and effects between the surveys. About half percentage of the increments in early initiation of breastfeeding rate was attributed to the differences in the composition of the respondents while half of the change in early initiation of breastfeeding change was due to the difference in the effect of the selected explanatory variable (Table 3).
In the detailed decomposition analysis, the overall increase in early initiation of breastfeeding rate from 2005 to 2016 was attributed to the difference in characteristics (endowment) of women between the survey. Covariates that provides a significant contribution to the improvements of EIBF were the mode of delivery, place of delivery, birth order, and size of child at birth. The increase in the composition of women with health facility delivery between 2005 and 2016 significantly contributed to the increase in early initiation of breastfeeding. Also, the change in composition of women with vaginal delivery significantly contributed to the increase in early initiation of breastfeeding rate over time. Similarly, an increased composition of women with first and 2nd-3rd birth order significantly contributed to the increase in early initiation of breastfeeding rate over the study period (2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016). In addition, the increase in women with small birth size between 2005 and 2016 significantly contributed to the increments of delayed initiation of breastfeeding (Table 4). Difference due to effects of coefficient (C) Half of the rise in early initiation of breastfeeding was attributed to behavioural changes to breastfeeding within 1 hour, controlling the effect of change in compositional characteristics. Covariates that have a significant on coefficient contribution to the observed change in breastfeeding within 1 hour are husband/partner education, mode of delivery, and trimester of antenatal care visit. Controlling all compositional change factors, 14% of breastfeeding within 1 hour increment was attributed to behaviour of husband/partner primary education over the last decade. Keeping the compositional factors constant, women who have attend late ANC was negatively correlated with the early initiation of breastfeeding over the last 10 years.

Discussion
Delayed initiation of breastfeeding is a major public health issue in resource-limited settings like Ethiopia, the potential risk factor for neonatal mortality and morbidity [17,23,31,32]. The incidence of breastfeeding within 1 hour in a community is a reflection of time to antenatal care, residence, and delivery service utilization [8,11,16,18]. In this study, we note that the timely initiation of breastfeeding from 2005 to 2016 was slightly lower than the WHO and UNICEFF Children's Fund recommendations [33]. This may be due to Ethiopia's government provide sufficient emphasis to key messages since 2004 on the timely introduction of breastfeeding [34] and establishing guidelines for feeding infants and young children.
From the decomposition analysis, the breastfeeding rate within an hour suggests an improvement over the survey period in Ethiopia. Hence, understanding the source of change has critical public health relevance to uncover what are contributing factors to the change in breastfeeding within an hour as well as understanding which factors are making progress in   growing breastfeeding with an hour to assess current implementation strategies. The disparity in the composition of women over time was responsible for 50.02% of the rise in the breastfeeding rate within an hour over the entire sample survey period (2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016). An improvement in the composition of health facility distribution over the study showed a major impact on the increase in breastfeeding within 1 hour in accordance with other research performed in Tanzania [35], Bangladesh [36], South Asia [37], and Ethiopia [2,19,38]. This agreement confirms that prioritization should be given to encouraging and facilitating the use of health facility delivery services in order to make progress towards early initiation of breastfeeding. Hence delivering at a health facility reduces traditional beliefs about breastfeeding practice such as prelacteal feeding and misconception about colostrum because the health professionals empowered the mother to resist the role of external pressure and interference that promotes delayed initiation of breastfeeding. However, there is a suggestion that mothers who were delivered at home were highly exposed to practicing prelacteal feeding in Ethiopia [39][40][41]. Similarly vaginal delivery has a significant positive contribution to the rising of breastfeeding within 1 hour consistent with other studies conducted in Brazil [42], and Ethiopia [19,21,26,43]. Since mothers who deliver vaginally have immediate skin-to-skin contact and infant treatment, as well as the onset of lactation is rapid that inhibits the likelihood of prelacteal feeding supplementation. However, a mother delivered by caesarean section typically associated with reducing timely initiation of breastfeeding hence the effect of anaesthesia delays the onset of lactation.
Also, increasing the composition of women with a child of small birth size has a significant contribution in reducing the timely initiation of breastfeeding within 1 hour of delivery. Since small babies may have an illness and stay away from their mother, resulting in the early supplementation of prelacteal feeding that inhibits the ability to initiate breastfeeding within 1 hour of delivery due to poor breastfeeding reflex and sucking inability similar findings were published in West Africa [20] and Zimbabwe [24], and prospective study in low and middle-income countries [1]. However, the large-size child can be considered as healthy and completely matured who possesses good reflection to initiate timely breastfeeding as viewed by their mothers and health care workers. This result indicates that the initiation of breastfeeding involves additional care and encouragement for mothers with small-sized infants.
Approximately 50% of early initiation breastfeeding progress was attributed to the shift in conduct towards breastfeeding initiation within 1 hour in Ethiopia, monitoring the effects of endowment characteristics. Factors substantially contributing to the positive and negative changes in early breastfeeding initiation were the achievement of husband/partner primary education and ANC visit at the third trimester respectively.
Increasing behavioural improvement in primary education for husbands/partners leads significantly to increasing early initiation of breastfeeding in Ethiopia, i.e. a 14 percent point change. Since husband education has a greater understanding regarding child care practice that receives information from health extension staff as well as daily mass media messages that can pressure their wife to start early breastfeeding even though there is no supporting evidence for this finding. Late initiation of ANC visits i.e. at the 3rd trimester significantly contributed to the reduction of early initiation of breastfeeding by 16 percentage points. Since a mother attends ANC visit at the 3rd trimester after feeling uncomfortable and illness, they may not well aware of the importance of timely initiation of breastfeeding and colostrum to their child hence their first aim of attending ANC might be receiving treatment their health similar findings are published in low and middle-income countries [1] prospective study. Currently, focussed antenatal care guideline in Ethiopia recommends intensive health education about breastfeeding at the time of antenatal care visits for mothers.

Strengths and limitations
One of the strengths of this analysis is that we used data from a national survey, the 2005-2016 EDHS. Therefore, the results of the study have thoughtful consequences at the personal, group and policy level.
However, there were limited sample sizes in some local areas represented in the survey, and the findings should therefore be viewed with caution. Since this study is a secondary data review of a national sample, other primary variables are not included, such as traditional attitudes, psycho-social influences, the preference of spouses for breastfeeding, and in-depth qualitative views of mothers. This research is focused on cross-sectional data and it is therefore difficult to show the cause and effect relationships of the covariates on the timely initiation of breastfeeding, and a recall bias may be susceptible to the survey responses.

Conclusion
In this study, we find that the trends of early initiation of breastfeeding have increased marginally over the last 10 years in Ethiopia. With regarding to multivariate decomposition analysis, half of the overall rise in early initiation of breastfeeding was attributed to the change in compositional characteristics of women over 10 years in Ethiopia. Change in the composition of reproductive ages women's characteristics according to place of delivery and mode of delivery were the major source of the increase in early breastfeeding initiation rate overtime while the size of the child at birth was contributed to delay timely breastfeeding over the study period.
Also half of the increase in early breastfeeding initiation rate was due to the changed behaviour of women towards early breastfeeding initiation over the last 10 years in Ethiopia. The change in behaviour of husband/ partner education has a positive influence on the early initiation of breastfeeding while late initiation of antenatal care visits has been negatively correlated with early initiation of breastfeeding over time. Public action should be required, including encouraging the provision of women's health facilities for more changes in the initiation of early breastfeeding. The ministry of health and other stakeholders should continue to enrich the coverage of early ANC visits and promoting community education regarding the need of early breastfeeding initiation to their newborns. Further interventions are needed among women deliver by caesarean section, one of the main barriers in delayed initiation of breastfeeding over the last 10 years in Ethiopia.