Spatial Distribution of Infant Mortality in Ethiopia: Using Demographic Health Survey, the Ethiopia


 Background

Mortality is one of the demographic variables that affect population trends. Among mortality of children, Infant mortality contributed to more than 75% of all under-five deaths globally. It disproportionately affects those living in the different regions of countries and within the region. Exploring the spatial distribution and identifying associated factors is important to design effective intervention programs to reduce infant mortality. Thus, this study aimed to assess the spatial distribution and associated factors of infant mortality in Ethiopia using the 2016 Ethiopian Demographic and Health Survey (EDHS).
Method

The Data this study were used Ethiopian Demographic and Health Survey in 2016. A total of 11,023 live births from the EDHS data were included in the analysis. Spatial analysis was done to explore spatial distribution of infant mortality using ArcGIS version 10.4.
Results

This study revealed that the spatial distribution of infant mortality was non-random in the country with Moran’s index 0.1546 (P-value=0.0185). The Afar and Somali regions of Ethiopia were identified in this study on the hot spot of infant mortality.
Conclusions

The spatial distribution of infant mortality varies across the country. ANC usage, sex of a child, birth interval, birth size, birth type, birth order, wealth index, residence, region, and the spatial variable (Si) were significant predictors of infant mortality. Therefore, it needs interventions in the hot spot areas. Focusing on maternal health care services, rural residences, multiple births, infants having a smaller birth size, and male infants deserves special attention.

Africa, the main risk factors related to a high number of infant deaths include lack of access to funds, and infrastructure, access to education, lack of medical professionals, poverty, and discrimination [5].
The Sub-Saharan African countries have achieved remarkable improvement in infant survival rates since the introduction of the Millennium Development Goals (MDGs), However, infant mortality was higher compared to others, an estimated 5.2 million children died in 2019 more than half of those deaths occurred in Sub-Saharan Africa [6]. Ethiopian Demographic and Health Survey (EDHS,2016) result indicated that the infant mortality rate for the 5 years preceding the survey was 48 deaths per 1000 live births [7]. It was a large share among under-ve mortality. About 72% of under-ve mortality in our country occurs before the rst birthday.Infant mortality varies in the country in space and time by changing its magnitude. Deaths occurring in children under one year of age measured by the infant mortality rate are nonrandom. Therefore, this study tries to address the spatial distribution of infant mortality and explore the major risk factors of infant death taking into consideration possible spatial correlations.

Materials And Methods
The source of data for this study was the 2016 EDHS. The data were downloaded from DHS website by this link www.meauredhs.com. The survey covered all nine regions and two city administrations of Ethiopia, and regions are divided into 68 zones, and zones, into administrative units called districts (817).
Each district is further subdivided into the 16,253 lowest administrative units, called kebele [7].

Study variables
Outcome variable.The outcome variable of this study is infant mortality that refers to the death of an infant before his or her rst birthday.
Explanatory variables. The predictor is characteristics by individual-level and community/clusterlevel( Figure 1).

Statistical Data Analysis
In this study, the data were analyzed using SAS version 9.4 with Proc Glimmix by using the method of LAPLACE approximation. In addition to SAS software to analyze the data, this study was used the ArcGIS version for spatial data analysis. In this study, infant mortality was considered as response variable.
Henceforth, from a statistical viewpoint, the outcome variable is given by a binary variable. The most protuberant logistic model for this condition is the binary logit model. Y ij be a dichotomous outcome random variable with categories 1 (infant who died) and 0 (infant who are alive) in the rst year of life before the survey. The X ( n × ( k + 1 ) ) denote the collection of k-predictor variables of the response.
Then, the conditional probability that the i th infant has died given the vector of predictor variables X i is denoted by P i = P(y i = 1 | X i ). The expression P i in logistic regression model written in linear combinations of predictors can be expressed in the form of [8,9]: Whereβ's are the regression coe cient for the explanatory variable Geospatial data processing In this study, we extended the binary logistic regression models by allowing random effects accommodating spatial correlations. The model was considered hierarchical logistic regression model for the infant mortality dataset was arranged in spatial form. Different spatial covariance structures were explored in datasets on mortality of infants giving due attention to the various implementation issues of the models. his study was used spatial models to discern spatial patterns of infant mortality. The study was explored different spatial covariance structures.
Geostatistical methods such as spatial autocorrelation, kriging and semivariograms were applied to create a prediction grid surface from a scattered set of points. Kriging assumes that the distance or direction between sample points re ects a spatial correlation that can be used to explain variation in the surface

Exploratory Data Analysis
Among 645 clusters, 622 were included in this study, 21 clusters were excluded due to zero GPS longitude and latitude coordinate readings for spatial analysis while the rest two of them were not included initially from the EDHS coordinate le. A total weighted 11,023 live births within ve years preceding the 2016 EDHS was included in the analysis, and the infant mortality rate (IMR) in Ethiopia was 48 per 1000 live births in 2016. The majority of 9,807(89.0%) in this study participants were from rural dwellers. Among the respondents, 4,851(44.0%) were from the Oromia region and 2,296(20.8%) were from SNNPR. The infant mortality rate (IMR) varies across the regions of the country. The highest IMR was observed in Harari (76.9 per 1000 live birth) and Afar (70.2 per 1000 live birth). The lowest was observed in Tigray   The estimated Global Moran's Index in this study was 0.1546; indicates that the spatial distribution of infant mortality signi cantly clustered by in Ethiopia. In addition to looking at the Moran's I to identify whether there is a spatial correlation, the P-value (<0.0185) was found to be less than 0.05, suggesting signi cant evidence of unexplained spatial autocorrelation in the risk of infant mortality ( Figure 4). Similarly, the incremental spatial autocorrelation graph identi ed the maximum peak distance value (30000 meters), which indicates distances where spatial processes promoting clustering are most pronounced. The color of each point on the graph corresponds to the statistical signi cance of the zscore values ( Figure 5).

Cluster and Outlier Analysis of Infant Mortality
Cluster and outlier analysis was conducted to identify the nature of clustering by using Anselin local Moran's I. The red color (cluster-low) indicates that the low rate of infant mortality is surrounded by a low rate of infant mortality, and the dark green color (cluster-high) indicates a high rate of infant mortality surrounded by the high rate of infant mortality. Whereas the green color (high outlier) indicates a high rate of infant mortality surrounded by the low rate of infant mortality and the yellow color (low outlier) shows a low rate of infant mortality surrounded by the high rate of infant mortality. Signi cant clusters were found in Afar, Addis Ababa, border of Benishangul Gumuz and Oromia. High outliers were observed on Harari, Dire Dawa, South Tigray, border of Amhara and Oromia, border of Oromia and SNNPR while the low outliers were found in the Somali region ( Figure 7).

Hot Spot Analysis of Infant Mortality
The Local Getis-Ord Gi* statistics identi ed signi cant hot spot and cold spot areas of infant mortality. The red color indicates that signi cant hot spot (high-risk) areas for infant mortality and found in Afar and Somali regions. The blue color indicates the cold spot (low risk) areas of infant mortality. These cold spot areas were observed in Addis Ababa, the central part of the Oromia and Amhara regions ( Figure 8).

Spatial Interpolation of Infant Mortality
The spatial kriging interpolation analysis was used to predict infant mortality for unsampled areas of the country based on sampled enumeration area measurements. Prediction of the high-risk areas was indicated by red predictions. Afar, Somali, Harari, Southwest SNNPR, East Benishangul Gumuz, South and central part of Oromia were predicted as more risky areas compared to other regions. Whereas, Tigray, Amhara, Addis Ababa, Northwest Oromia, North Somali, Dire Dawa, Northwest Benishangul Gumuz, Northeast SNNPR, border of Gambela and SNNPR, border of Gambela and Oromia were predicted as having less risk for infant mortality ( Figure 9). Spatial distribution of the spatial autocorrelation term: autocovariate variable Figure 10 shows the spatial distribution of the autocovariate variable in equation , which represents the spatial autocorrelation term in the GLMM. The autocovariate variable has the same unit as the dependent variable, which also represents the incidence of death occurrence, but it is just a macro spatial trend. Figure 10 show that the spatial distribution of incidence of death occurrence has a strong spatial tendency and heterogeneity, which presents a transitional and gradual change throughout the country.
The incidence is very high in the Northern and central part of Afar, the entire part of Harari, the central part of Somali, Southern and Western SNNPR, Eastern Oromia, and Western Benishangul Gumuz regions that is consistent with what we observed in Figure 6 that these areas were characterized by a high proportion of infant mortality. Figure 4 show that the spatial distribution of infant mortality exhibit spatial correlations and this study investigate various possible spatial covariance structures. Based on statistical software SAS offers several possible spatial covariance structures: Exponential, Gaussian, Linear, Spherical, etc. each represents a particular pattern of changes in spatial covariance among residuals as observations grow in distance from one another. Below We utilized the DIC and AIC t statistics to examine Spherical and Gaussian spatial covariance structures. Our rst observation is that the t statistics drop in value, indicating a better t to the data( Table 1).

Spatial Covariance Structures
Results of this study indicated that the Spherical and Gaussian spatial covariance structures were t to the residuals from the GLMM. Focusing on the AIC t statistics, we observe a drop from 4146.12 to 4039.27 in Gaussian covariance structure compared to a model with an unstructured covariance structure. This is an improvement, but as part of our model building process, we consider the two spatial covariance structures. When comparing the Gaussian and Spherical models, the t statistics do now show a meaningful drop, suggesting that perhaps a Gaussian covariance structure is not appropriate. As a result, the Spherical spatial covariance structure better ts the data. We used this covariance structure for modeling in this study, since it has smaller DIC and AIC t statistics as compared to the other (Table  1). The intercept-only model without explanatory variable was constructed to measure the effect of community variation on infant mortality.
The variance of random effects at the cluster is ( =3.6413, p-value<0.0001) which was statistically signi cant and re ects there is statistically signi cant variation in the infant mortality among infants across the community (see Table 2).
The estimated intra-class correlation computed(ICC)was 52.5%. This indicates about 52.5% of the total variation for infant mortality was due to the difference between communities, leaving 47.5% of the variability to be accounted for the infants or other unknown factors.
The covariance parameter estimates in Table 2 indicates that the estimated higher-level error variance goes down from 3.6413 to 2.7646. The proportion of explained variance at infant-level computed as   2). In this study the random coe cients model does not converge; as a result, we exclude the random coe cients model.

Factors Of Infant Mortality
The result of the selected model Table 3 shows that individual-level factors such as birth size, birth type, sex of a child, breastfeeding status, birth order, birth interval, ANC usage, and wealth index were found to be signi cantly associated with the odds of infant mortality. On the other hand, community-level factors such as place of residence and region had signi cant effects (P-value<0.05) on the log-odds of the i th infant in the j th cluster experiences death. In addition to the individual and community level factors, the spatial autocovariate variable (Si) was also signi cantly associated with infant mortality ( Table 3). Sex of infant: The ods ratio for male infants was 1.59 found to be (95% CI:1.29,1.95) meaning that the risk of males dying was about 59% higer than that for female infants. The con dence interval indicated that the risk of death for female infants could be as low as 29% and as high as 95%.
Birth order: The reference group, in this case, was taken as a single birth. Infants belonging to the 6 and more birth order category were about 13% more likely to die relative to the reference group (OR= 1.1331 95%,CI:0.73,1.53).
Birth size :The estimated odds of infant death among infants born with birth size perceived by their mothers as small was 1.27 (OR=1.27, 95% CI:0.79,1.60) times more likely than infants born with birth size perceived by their mothers as large.  Table 3).Furthermore, the spatial autocorrelation between clusters. In Table 3 the P=0.0028 also proves that it was true, in a sense that there was a spatial correlation of infant mortality between clusters. The spatial variable correlation with infant death was a negative value, -0.5778, which indicates that clusters with a low incidence of infant mortality were usually surrounded by clusters with a high incidence of infant mortality. Bear in mind that during the interpretation of one variable so far it is assuming that the other variables are held constant (Table 3).

Discussion
This study aimed to assess the spatial distribution and associated factors of infant mortality in Ethiopia using 2016 EDHS data. The spatial analysis in different methods consistently veri ed hot and cold spot areas of infant mortality among infants in Ethiopia. The spatial analysis indicated that Afar and Somali regional states were statistically signi cant hot spot areas for infant mortality. The possible justi cation could be is the variation in ANC utilization across regions. The lowest ANC utilization rate was reported in the hot spot areas (Afar and Somali regions) as compared to cold spot areas [7]. This could be attributed to the discrepancy in the distribution of maternal health services, and environmental factors across the area [10]. This implies that identifying regions with high infant mortality is important for prioritizing areas for analysis of cause and planning of remedial actions. In contrast, the cold spot areas were observed in the entire part of Addis Ababa, the central part of the Oromia and Amhara regions. The possible justi cation could be these regions are urban as compared to the hot spot areas. They have good access to health facilities, mothers may have awareness about ANC utilization, and its bene ts compared to other regions.
The ndings of this study show that ANC usage, preceding birth interval, birth order, breastfeeding status, sex of a child, birth size, type of birth, wealth index, region, residence, and the spatial autocovariate variable (Si) were determinants of infant mortality.
ANC usage has a signi cant association with infant mortality. The odds of infant death among infants from mothers who did not use ANC more likely as compared to infants born from mothers who did during their pregnancy in the same clusters. This nding is consistent with studies conducted in Ethiopia [11], Pakistan [12], and Brazil [13].
The birth order of a child is another factor of infant death. The odds of dying among infants with sixth or higher birth was more likely than rst-born infants in the same clusters. This nding is consistent with the studies conducted in Ethiopia [11], and South Africa [14].
The sex of a child signi cantly in uenced the occurrence of infant mortality. The risk of death among male infants was more likely as compared to female infants in the same clusters. This result is in line with studies conducted in Ethiopia [15], and Bangladesh [16].
Concerning birth interval, infants born with a preceding birth interval of between 25-36 months and less than or equal to 24 months was more likely to die as compared to infants born with a birth interval of more than 36 months in the same clusters. This result is similar to studies done in Ethiopia [17], and Nepal [18].
This study also demonstrated that the odds of death among multiple births were 6.82 times more likely than singletons in the same clusters. This nding is consistent with studies conducted in Ethiopia [15,19], Kenya [20] and Brazil [21]. Multiple births are at high risk for numerous negative birth outcomes, and these outcomes contribute to a higher rate of mortality during the infancy period [22].
This study indicated that breastfeeding status has a signi cant association with infant mortality. Among mothers who did not breastfeed their child, the odds of infant death were 3.93 times more likely as compared with mothers who did in the same clusters. This nding is supported by previous studies done in Ethiopia [23], and Kenya [24].
Our nding shows that the wealth index has a signi cant association with infant mortality. Among mothers who belong to the poor wealth index, the odds of infant death were 1.36 times more likely as compared to their counterpart rich wealth index categories in the same clusters. This result is in line with studies done in Bangladesh [16].
This study revealed that place of residence is signi cantly related to infant mortality. The odds of infant death among rural residents was more likely as compared to their counterpart urban residents. This nding is consistent with studies conducted in Ethiopia [15], Nigeria [25], and Pakistan [12]. The reason may be that mothers living in rural areas lack access to health institutions to have ANC follow-ups or lack of media exposure which in turn affects their knowledge and practice of care for the infant.
This study revealed that region has a signi cant association with infant mortality. The odds of dying during the infancy period in the Afar, Somali, and Harari regions were more likely as compared to infants from the Tigray region. This nding is in line with studies conducted in Ethiopia [19]. The difference in mortality between regions may be due to variations in service accessibility and coverage. The This study reveals that the spatial variable has a negative signi cant effect where clusters with a low incidence of infant mortality were usually surrounded by clusters with a high incidence of infant mortality. Whereas, clusters with a high incidence of infant mortality were usually surrounded by clusters with a low incidence of infant mortality, which is in line with studies in Brazil [28].

Conclusions
The spatial distribution of infant mortality was found signi cantly clustered in Ethiopia. This study investigated that high-risk areas of infant mortality were found in Afar and Somali regions of the country.
In contrast, the cold spot (low risk) areas were identi ed in Addis Ababa, the central part of the Oromia and Amhara regions.
The result of multilevel GLMM that adjusted for spatial effects better accounted for geographical variability and provided more accurate information on the spatial distribution of infant mortality than non-spatial multilevel models. Therefore, the model that adjusted for spatial effects performed better in this study. The study showed that there was a variation in infant mortality among the regions of Ethiopia.
About 57.2% of the variability at the community level for infant mortality was explained by the individuallevel and community-level variables.
This study identi ed the risk factors of infant mortality at the individual level and community level. At individual-level variables are: being male, multiple birth type, having a smaller birth size, born from mothers who did not attend ANC, born from mothers with poor wealth index, and having a higher birth order were signi cant factors that increase the risk of death during infancy period. Whereas, having a large birth interval and initiating breastfeeding early were factors that decrease the risk of infant death. At the community level, being a rural residence, born in Afar, Harari, and Somali regions were factors that increase the risk of infant mortality. There was a signi cant spatial correlation of infant mortality between clusters in which clusters with a low incidence of infant mortality were usually surrounded by clusters with a high incidence of infant mortality and vice versa.

Declarations Ethical Considerations
Permission for data access was obtained from a major demographic and health survey through an online request from http://www.dhsprogram.com to download and use the raw data for this study. The data used in this study were available without individual identi ers. The Institutional Review Board approved procedures for DHS public-use datasets do not allow speci c households or sample clusters to be identi ed. The geographic identi ers are available only for the enumeration areas (EAs) as a whole, not for particular household addresses. The measured GIS coordinates are randomly displaced in a large geographic area so that particular EAs cannot be identi ed. Each enumeration area or primary sampling unit has a number in the data le, but these numbers do not have any labels to indicate their names or locations.

Consent for Publication
Not applicable.

Competing interests
We the authors declared that we have no competing interest.

Funding
No fund was obtained.

Availability of Data and Materials
The datasets used and analyzed during this study are available from the corresponding author on reasonable request. The contact person is Ashena Abate ashu.abate@gmail.com.The statement to con rm that all methods were carried out in accordance with relevant guidelines and regulations.

Author contributions
All authors made substantial contributions to conception and design, acquisition of data, or analysis and interpretation of data; took part in drafting the article or revising it critically for important intellectual content; agreed on the journal to which the article will be submitted; gave nal approval of the version to be published; and agreed to be accountable for all aspects of the work. Figure 1 The predictor is characteristics by individual-level and community/cluster-level The proportion of infant mortality varies from cluster to cluster. That means the enumeration area is considered as a random effect for this study Figure 5 The incremental spatial autocorrelation graph identi ed the maximum peak distance value (30000 meters), which indicates distances where spatial processes promoting clustering are most pronounced.

Figures
The color of each point on the graph corresponds to the statistical signi cance of the z-score values Cluster and outlier analysis was conducted to identify the nature of clustering by using Anselin local Moran's I. The red color (cluster-low) indicates that the low rate of infant mortality is surrounded by a low rate of infant mortality, and the dark green color (cluster-high) indicates a high rate of infant mortality surrounded by the high rate of infant mortality. Whereas the green color (high outlier) indicates a high rate of infant mortality surrounded by the low rate of infant mortality and the yellow color (low outlier) shows a low rate of infant mortality surrounded by the high rate of infant mortality. Signi cant clusters were found in Afar, Addis Ababa, border of Benishangul Gumuz and Oromia. High outliers were observed on Harari, Dire Dawa, South Tigray, border of Amhara and Oromia, border of Oromia and SNNPR while the low outliers were found in the Somali region Figure 8 Page 23/23 The Local Getis-Ord Gi* statistics identi ed signi cant hot spot and cold spot areas of infant mortality.
The red color indicates that signi cant hot spot (high-risk) areas for infant mortality and found in Afar and Somali regions. The blue color indicates the cold spot (low risk) areas of infant mortality. These cold spot areas were observed in Addis Ababa, the central part of the Oromia and Amhara regions Figure 9 Afar, Somali, Harari, Southwest SNNPR, East Benishangul Gumuz, South and central part of Oromia were predicted as more risky areas compared to other regions. Whereas, Tigray, Amhara, Addis Ababa, Northwest Oromia, North Somali, Dire Dawa, Northwest Benishangul Gumuz, Northeast SNNPR, border of Gambela and SNNPR, border of Gambela and Oromia were predicted as having less risk for infant mortality Figure 10 The spatial distribution of the autocovariate variable in equation , which represents the spatial autocorrelation term in the GLMM. The autocovariate variable has the same unit as the dependent variable, which also represents the incidence of death occurrence, but it is just a macro spatial trend