N/A
Background Pre-lacteal feeding persists in low and middle-income countries as deep-rooted nutritional malpractice. It imposes significant negative consequences on neonatal health, including increased risk of illness and mortality. Different studies revealed that pre-lacteal feeding practice is decreased over time. Even though different studies are done on the prevalence and determinants of pre-lacteal feeding practice, up to our knowledge, the spatial distribution and the determinants of the change in pre-lacteal feeding practice over time are not researched. Objective To assess the spatial distribution and determinants of the change in pre-lacteal feeding practice over time in Ethiopia. Methods We used the Ethiopian demographic and health surveys (EDHSs) data. For this study, a total weighted sample of 14672 (5789 from EDHS 2005, 4510 from EDHS 2011, and 4373 from EDHS 2016) reproductive-age women who gave birth within two years preceding the respective surveys and whoever breastfeed were used. The logit-based multivariate decomposition analysis was used to identify factors that contributed to the decrease in pre-lacteal feeding practice over the last 10 years (from 2005 to 2016). Using the 2016 EDHS data, we also conducted a spatial analysis by using ArcGIS version 10.3 and SaTScan version 9.6 software to explore the spatial distribution and hotspot clusters of pre-lacteal feeding practice. Result Pre-lacteal feeding practice was decreased from 29% [95% Confidence interval (CI): 27.63–29.96%] in 2005 to 8% [95% CI: 7.72–8.83%] in 2016 with annual rate of reduction of 7.2%. The overall decomposition analysis showed that about 20.31% of the overall decrease in pre-lacteal feeding practice over the last 10 years was attributable to the difference in composition of women (endowment) across the surveys, while, the remaining 79.39% of the overall decrease was explained by the difference in the effect of characteristics (coefficient) across the surveys. In the endowment component, the difference in composition of residence, perception of distance from the health facility, maternal educational level, wealth status, occupation, ANC visit, place of delivery, the timing of breastfeeding initiation, and wanted last-child/pregnancy were found to be significant contributing factors for the decrease in pre-lacteal feeding practice. After controlling for the role of compositional changes, the difference in the effect of distance from the health facility, wealth status, occupation, antenatal care (ANC) visit, and wanted last-child/pregnancy across the surveys were significantly contributed to the observed decrease in pre-lacteal feeding practice. Regarding the spatial distribution, pre-lacteal feeding practice was non-random in Ethiopia in which the primary and secondary clusters’ of pre-lacteal feeding identified in Somalia and the Afar region respectively. Conclusion Pre-lacteal feeding practice has shown a significant decline over the 10-year period. Program interventions considering women with poor maternal health service utilization such as ANC visits, women with poor socioeconomic status, women with an unintended pregnancy, and women from remote areas especially at border areas such as Somali and Afar could decrease pre-lacteal feeding practice in Ethiopia.
We used the three Ethiopian demographic and health surveys (EDHSs) (2005, 2011, and 2016) data, which are the nationally representative surveys performed in Ethiopia. In each of the surveys, a two-stage cluster sampling was employed. In the first stage, 540 Enumeration Areas (EAs) for EDHS 2005, 624 EAs for EDHS 2011, and 645 EAs for EDHS 2016 were randomly selected proportional to the EA size and, on average, 27 to 32 households per EAs were selected in the second stage. A total weighted sample of 14672 (5789 from EDHS 2005, 4510 from EDHS 2011, and 4373 from EDHS 2016) reproductive-age women who gave birth within two years preceding the respective surveys and whoever breastfeed were used for this study. There is detailed and comprehensive information relating to the sampling process and other information in each survey report [27–29]. The outcome variable was feeding of the child other than breast milk within three days, which was a binary outcome variable coded as “1” if the mother gave anything other than breast milk and “0” if a mother gave nothing for her newborn child within three days. The independent variables included (after searching of literatures) for our study were region, place of residence, perception of distance from the health facility, age, educational level, wealth index, occupation, mass media exposure, parity, ANC visit, place of delivery, delivery by cesarean section, size of the child at birth, and timing of initiation of breastfeeding. Mass media exposure: Created by combining whether a respondent reads a newspaper, listen to the radio, and watch television and coded as yes (if a woman had exposed to at least one of these media) and no (if women were not exposed to at least one of the media). Size of the child at birth: It is defined as the size of the child during delivery, which is based on the mere report of mothers and categorized in the surveys as very small, small, average, large, and very large and recoded as average, small (includes very small and small), and large (includes large and very large) for this analysis. The other independent variable definitions are self-explanatory and more information about these variables can get from the EDHS 2016 report [28]. The data were extracted and recoded using Stata version 14. Throughout the analysis, the data were weighted to make it representative and to provide better statistical estimates. The trend and multivariate decomposition analyses were done using Stata version 14. The trend of pre-lacteal feeding practice was examined separately for the periods 2005–2011, 2011–2016, and 2005–2016. The trend of pre-lacteal feeding in each of the selected sociodemographic characteristics of respondents was also analyzed using descriptive analysis. The multivariate decomposition analysis technique was used to analyze the difference in pre-lacteal feeding practice between two points in time (2005 and 2016). It is widely practiced in public health studies to identify components of a change over time and identify contributing factors for the change [30,31]. The analysis decomposes the differences in pre-lacteal feeding practice over time into two components (the endowment part and coefficient part). For our study, the 2016 EDHS data was appended to the 2005 EDHS data using the “append” Stata command, and the logit based multivariate decomposition analysis (using mvdcmp STATA command) was used to identify factors that contributed to the decrease in pre-lacteal feeding practice over the last 10 years. Therefore, the observed decrease in pre-lacteal feeding practice was additively decomposed into differences due to endowment/characteristic and differences due to coefficient/effects of the characteristic component. In doing the decomposition analysis, the Logit or log-odd of pre-lacteal feeding practice is taken as [31]: In which, the “E” component is the part of the differential due to differences in characteristics while the “C” component refers to the part of the differential attributable due to differences in coefficients or effects of characteristics. We conducted a spatial analysis using ArcGIS version 10.3 and SaTScan version 9.6 software. To assess whether the spatial distribution of pre-lacteal feeding practice was random or non-random (spatial autocorrelation), Global Moran’s I statistic was used. Kriging spatial interpolation technique was used to predict pre-lacteal feeding practice in the un-sampled areas based on the values from sampled measurements. Besides, Getis Ord Gi* statistical hotspot analysis was done to identify the significant hot spot areas (areas with high rates of pre-lacteal feeding practice) and cold spot areas (areas with lower rates of pre-lacteal feeding practice). Moreover, we used Bernoulli based spatial scan statistical analysis to detect statistically significant clusters. To fit the model women who gave anything within three days for the newborn were taken as cases and those who gave nothing were taken as controls. The primary and secondary clusters were identified and p values were assigned and ranked using their log-likelihood ratio (LLR) test based on the 999 Monte Carlo replications. Areas with high LLR and significant p-value were considered as clusters with higher rates of pre-lacteal feeding practice and the spatial window with the highest significant LLR test statistic was defined as the most likely (primary) cluster. Since this is a secondary analysis of the Demographic and Health Survey (DHS) data, ethical approval was not necessary. However, we registered and requested the datasets from DHS on-line archive and received permission to access and download the data files. Moreover, for Geographic information system coordinates, the coordinates are only for the enumeration area (EA) as a whole and the measured coordinates were randomly displaced within a large geographic area so that no particular enumeration areas can be identified.