Mapping of mothers' suffering and child mortality in Sub-Saharan Africa

Child death and mothers who suffer from child death are a public health concern in Sub-Saharan Africa. The location and associated factors of child death and mothers who suffer child death were not identified. To monitor and prioritize effective interventions, it is important to identify hotspots areas and associated factors. Data from nationally representative demographic and health survey and Multiple Indicator Cluster administrated in 42 Sub-Sahara Africa countries, which comprised a total of 398,574 mothers with 1,521,312 children. Spatial heterogeneity conducted hotspot regions identified. A mixed-effect regression model was run, and the adjusted ratio with corresponding 95% confidence intervals was estimated. The prevalence of mothers who suffer child death 27% and 45–49 year of age mother 48%. In Niger, 47% of mothers were suffering child death. Women being without HIV knowledge, stunted, wasted, uneducated, not household head, poor, from rural, and from subtropical significantly increased the odds of the case (P < 0.05). The spatial analysis can support the design and prioritization of interventions. Multispectral interventions for mothers who suffer from child death are urgently needed, improve maternal health and it will reduce the future risk of cases.

Among women age 45-49 years. The average birth rate was 6 children per one mother and the highest birth rate countries are Niger and chad (each women in average birth 8 children), the lowest was South Africa which account 3 children in each mother. About half of (48%) mothers whose age 45-49 suffered child death and from 15 countries above half of women suffered child deaths. Out of five women in Niger four of them were suffer in child death, which is the highest proportion compared to other countries. The next highest proportion of mother who suffer in at least one child death in this age group were Burkina Faso (67%), Liberia (65%), Chad (63%), and Rwanda (62%) (Fig. 3; Table 1). www.nature.com/scientificreports/ SSA mothers who suffer from child death are a serious public health concern. Differences of up to 20 percentage points in prevalence within the same country were common. Mother who had lost their child due to death is high in most Western and some Central SSA regions (Fig. 4).
Hot/cold spot analysis. Hot spot analysis is performed by means of a statistical test. The red color (hot spot) shows a higher risk of mothers suffering from at least one child death. Out of the total regions in SSA, 68 of them of were the hot spot regions compared to its neighbour. Most of those regions were found in Western SSA. In the cold spot 65 regions were identified, Most of them obtained in Southern SSA (Fig. 5). Table 1. Spatial distribution of sample size case prevalence, birth rate, ratio, residence and older age in child death and mother suffering of child death. MD the number of mothers who are suffering child death, MT total number of mothers, CD total number of child death, CT total number of child birth, BR birth rate. www.nature.com/scientificreports/ Figure 2. The proportion of women who suffer at least one child death by residence: the red and the green color indicate the proportion of urban and rural, respectively mothers who suffer in at least one child with confidence interval.
High-value cluster using the Poisson model. Nineteen statistically significant clusters have been identified using SaTScan (P < 0.05). The primary circular window (cluster) of mothers who suffer from child death was found in Nigeria. In this cluster, 3824 cases expected but found 6406, 46% of mothers were suffered from child death with a Relative Risk (RR) of 1.7.The second major cluster was located in Chad. The expected number of cases in this cluster was 4351, but 6703 cases were identified. The relative risk was 1.6, with 42% of mothers suffering from child death. The third most likely significant clusters were found in Burkina Faso and Mali. In this cluster, 4328 cases were expected, but 6365 cases were identified, with RR 1.5 and 40% of mothers suffering from child death (Fig. 6).

Spatial interpolation.
Our spatial interpolation revealed that mothers who have suffer from child deaths have been a serious public health problem in most areas of SSA, with the exception of southern SSA countries. Most areas of Western SSA, Democratic Republic of Congo, Ethiopia, Uganda, and Mozambique and some areas of Angola and Tanzania had extremely high burden of mothers who lost at least one child (Fig. 7).
Model comparison. Without any covariant, the ICC value for child death and mothers who suffer from child death were 8.6% and 10.5%, respectively. This illustrated that the multilevel analysis was more appropriate for both cases because ICC is > 5%. The Poisson regression variance (0.843) was greater than the mean (0.447), indicating that the data were over dispersed. Thus, Poisson regression with over dispersion is treated as a negative binomial regression. According to Log's likely hood results, the best fitness was the negative binomial regression ( Table 2).
Mixed effect logistic regression. After controlling for confounding factors at the individual (maternal) and community levels, HIV knowledge, stunting and wasting, age, education, household size, relationship to household head, wealth, age at first birth, number of unions, residence, source of drinking water, and ecology were statistically significant factors for both child death and mothers who suffer in child death (Table 3).
Mothers who suffer from child death. Stunted, wasted, and overlapping (both stunting and wasting in women) mothers who were 21%, 25%, and 43%, respectively, are more likely to suffer by children death than mothers who were neither stunted nor wasted. Repeatedly married mothers are 31% more at risk than once married moth- www.nature.com/scientificreports/ ers in child death. Mothers living in rural areas are 17% more likely to suffer child deaths compared to urban mothers. Mothers using an unimproved water source had a 7% higher risk of child death than mothers using an improved water source (Table 3).
Child death. The incidence rate for child death from mothers without HIV knowledge were 44% more risk than the incidence rate of the number of child death whose mother have HIV knowledge. The incidence rate of the number of child deaths from mothers aged 20-24, 25-29, 30-34, 35-39, 40-44 and 45-49 were 1.85, 3.8, 5.9, 8.6, 11.6 and 14.5, respectively times higher than the incidence rate of the number of child death from mothers age 15-19 (Table 3).

Discussion
Child death and mothers who suffer from child death are high in SSA. This finding is in line with other finding 6 . Most parts of western Africa, in some parts of Central and Eastern SSA were high compared to others. Health care services are a key proximate determinant of maternal and child health 7 . In addition, timely and appropriate maternal and child care can provide an opportunity to prevent or manage the causes of child mortality. Maternal health status is directly linked to child mortality 8 . ANC can improve children's and maternal health by identifying, managing, and referring to potential complications. ANC is also related to the prevention, identification and treatment of multiple health problems associated with child mortality 7,9 . The other possible reasons will maternal health, much is known about the consequences of anemia during pregnancy, including the increased risks of low birth weight, preterm birth, and neonatal mortality 10 . The other possible reason for contributing to child death is child anemia 11,12 . It has serious consequences, including child morbidity and mortality 13 . The consequence of malnutrition is high in SSA 14 . In addition, inadequate dietary diversity in children is one of the potential risks of child mortality. A lack of complementary feeding practices is the main cause of under-nutrition, which is a direct cause of child mortality 15,16 .
Maternal HIV knowledge has negative association with child mortality. This finding is in line with other findings [17][18][19] . The possible reason for this strong correlation is that increased maternal knowledge reduces child death in HIV by protecting mother-to-child transmission during pregnancy, childbirth and post-birth 20 . www.nature.com/scientificreports/ Figure 4. Subnational prevalence of mothers who lost at least one child due to death: the color from green to red shows an increasing prevalence of women who have lost at least one child. This analysis was carried out by QGIS 3.16.

Figure 5.
Hot-spots and cold-spots of women who suffer child death: each polygon on the map represents a single zone area with mothers who suffer child death. High (red colour) means high (hot spot) of mothers who suffer from child death. Low (blue colour) shows a low (cold spot) of mothers who lost at least one child. To perform this analysis was used GeoDa GIS version 1.14. www.nature.com/scientificreports/ Child mortality in educated mothers is lower than in uneducated mothers. This finding is similar to that of the other findings [21][22][23][24] . Therefore improving maternal education will improve the health of their children and the community 25,26 . A mother's age at first birth is negatively correlated with child death, with a decrease in mother's age at first birth increases the risk of child mortality. Which is similar to the previous studies 22 . Rural children are more likely to die than their counterparts. Potential causes of rural child deaths include access to health centre, reproductive health education, sanitation, quality of drinking water, and so on 27 . Children in relatively poorer households were more likely to died 28 . This is due to inadequate household income, inadequate sanitation, Figure 6. High-value clusters: the red circular shape shows the windows of the hotspots of mothers who have experienced at least one child's death. Cl cluster number on the map, O observed cases in the clusters, E expected cases in the cluster, RR relative risk, Pre prevalence, P P value. To conduct this analysis was used SaTScan v9.6 and QGIS. www.nature.com/scientificreports/ malnutrition, poor access to health care among families. Women who have been repeatedly married higher exposure to the risk of child death. This finding is in line with other findings 29 . This is due to women who have been repeatedly married may have had more children. The number of child increase, the child death increase too 29 . Table 3. Multilevel regression associated characteristics with child death and mothers suffering child death. a AOR: adjusted odds ratio for mixed effect logistic regression. b AIRR: adjusted incidence rate ratio for mixed effect negative binomial regression.   www.nature.com/scientificreports/ Finally we recommended that policymakers and other concerned bodies who are working on this serious public health issues in SSA countries should work in collaboration and understanding in order to mitigate the underline problems. Though the child death and mothers who suffer in child death are serious public health issues in each SSA country, priority has to be given in most areas of Western, some areas of Central and Eastern SSA countries.
As the problems of child death and mothers who suffer from child death are relatively serious in women who do not know about HIV, stunted, wasted, not house wife, repeatedly married and older age; and in households of large family size, unimproved water and sanitation; from communities who indwells in rural and low land areas, effective intervention measure should be designed and followed. Inclusion/exclusion criteria. Women at fertile age  included in the study, whereas women who have not give birth exclude from the study and the child included in this study were from those women who fulfilled this criteria. www.nature.com/scientificreports/ Data processing and analysis. Hot spot analysis. Based on 95% confidence interval (CI), hot spot is defined as clustering of high value of child death and number of mothers who lost their child with a Z score ≥ 2 and a P value < 0.05; Cold spot is a cluster of fewer child deaths and fewer mothers who have lost their child due to death with a Z score ≤ − 2 and a P value < 0.05; and a Z score close to zero means no spatial clustering. Mapping hot spot analysis is applied to local G* statistics and is used to identify and display clusters of high prevalence (hot spot) regions and low prevalence (cold spot) regions 30 .

Materials and methods
Spatial scan statistical analysis. Kulldoruff 's Scan Statistic was used to analyse the spatial distribution of the prevalence of child death and mothers who had lost their child. A purely spatial scan statistic used to identify areas with higher number of child deaths and mothers lost their child. Spatially significant higher and lower aggregate concentrations were identified and circular windows were observed [31][32][33][34] . With the discreet Poisson model, the number of cases in each cluster (enumeration area) has been estimated 30 .
Spatial interpolation. Spatial interpolation using ordinary kriging has been used to predict the prevalence of non-mediated areas from measured areas 35 .
Factor analysis. Due to the nature of the data, the risk factor of child death and mothers who suffer child death were only maternal/household and community characteristics. Because in both cases, the starting point of the problem was child death. Improved maternal health at the same time reduces the number of child deaths and mothers suffering from child deaths 36 .
Mixed effect model. Multi-level regression was conducted to assess factors related to maternal and communitylevel characteristics of child death and mothers suffering from child death. Multilevel analysis was considered appropriate due to the hierarchical nature of the DHS data as well as the estimation of individual and community level effects 37 . Data organized in two ways, the first to analyse the burden of child death on mothers, the outcome variable is child loss or not, which has binary response and run logistic regression analysis. The second outcome is the number of child deaths. The response is a count (0, 1, 2, 3, …) that fits Poisson's regression. Mixed effect multi-level Poisson and logistic regression model use and consisted of two levels (individual (women) and community levels). The appropriateness of the mixed model is checked using Inter-Class Correlation. The Poisson regression has been checked over/under dispersion.