Prevalence and Determinants of Home Delivery in Eastern Africa: Further Analysis of Countries Recent Demographic and Health Survey Data

Background: Despite the high proportion of maternal mortality ratio in East African countries primarily attributed to home delivery, overall magnitude of home delivery and its determinants remains unclear. Therefore, the current study aimed to determine magnitude of home delivery and its determinant factors in East Africa using Demographic and Health Survey (DHS) data. Methods: We pooled the DHS survey data of the 11 East African countries, and a total weighted sample of 125,786 women were included in the study. Generalized Linear Mixed Models (GLMM) was tted to identify factors associated with home delivery. Variables with Adjusted Odds Ratio (AOR) with a 95% Condence Interval (CI), and p-value < 0.05 in the nal GLMM model were reported to declare signicantly associated factors with home delivery. Result: The weighted prevalence of home delivery was 23.79% [95% CI: 23.55 – 24.02] among women in East Africa countries. Home delivery was highest among Ethiopian women (72.5%) whereas, it was lowest among women from Mozambique (2.8%). In GMM, respondent’s age group, marital status, educational status, place of residence, living country, wealth index, media exposure and total children ever born were shown signicantly associated with the home delivery in the East Africa counties Conclusion: Home delivery was varied countries of East African zone. The home delivery was signicantly increased among women aged 20-34 years, higher of rural residence, never married or married participants. To the contrast home delivery was decreased with higher educational level, media and higher wealth index. Wide range interventions to decrease home delivery should maternal education, family the poor and rich families, and married and unmarried mothers.


Background
Maternal mortality remains a major public health problem globally. Sub-Saharan African (SSA) regions bear the highest burden with 85% of maternal deaths reported from the region. Studies indicated every year 529,000 maternal deaths and four million newborn deaths in the rst week of life occur worldwide [1,2]. The estimated maternal mortality ratio (MMR) in developing countries (239 per 100,000 live births) is 20 times higher compared to the developed regions (12 maternal deaths per 100,000 live births). Despite the great improvement over the last decades, the drop in maternal mortality is far from reaching a target decline of reaching less than 70 MMR by 2030 at the current pace [3,4].
Most of maternal deaths in sub-Saharan Africa are highly attributed to home delivery with most births occurring at home. In low and middle-income countries (LMIC), many deliveries still occur at home without the assistance of trained attendants [2,5]. Mothers deliver in an unhygienic environment, without a skilled birth attendant and lifesaving medications. Sub-Saharan Africa and South Asia together contribute over 85% of maternal deaths, and of which only half of deliveries are at home [6,7]. The negative impact of home delivery extends to child and responsible for neonatal morbidity and mortality.
Since home deliveries are attended by unskilled health professionals and occurs in an unsafe environment, they leads to adverse neonatal and maternal outcomes like the increased risk of infection, postpartum hemorrhage (PPH), and transmission of HIV/AIDS to relatives or traditional birth attendants, who conduct deliveries without protective equipment. Most of these maternal deaths are preventable if timely and appropriate interventions are applied [8,9].
Evidences showed that even though skilled birth attendance can save the lives of women, only 59% of births were attended by skilled birth attendants between 2012 and 2017 in sub-Saharan Africa. High load of home delivery in the region is a precipitating factor for the high maternal mortality ratio. The large proportion of direct cause of maternal death including obstetric complications such as hemorrhage, pregnancy-induced hypertension, sepsis, and obstructed labor which collectively accounts for 64% of maternal deaths could be prevented primarily by making the delivery attended by a skilled birth attendant at a health facility [10,11].
Despite the high proportion of maternal mortality ratio in East African countries primarily attributed to home delivery, overall magnitude of home delivery and its determinants remains unclear. Therefore, the current study aimed to determine magnitude of home delivery and its determinant factors in East Africa using Demographic and Health Survey (DHS) data. The nding of the current study provides evidence for health planners, decision makers, stakeholders, and health professionals in planning for further reduction of home delivery which is helpful in turn to decrease maternal mortality in low-income and middle-income countries.

Study setting, design, and period
We conducted Pooled analysis based on Demographic and Health Surveys (DHS) conducted in the 11 East African countries (including Burundi, Comoros, Ethiopia, Kenya, Malawi, Mozambique, Rwanda, Tanzania, Uganda, Zambia and Zimbabwe) from 2012 to 2017. The DHS is considered as the main data source as it was designed to provide population and health indicators at the national and regional levels.
The data collection period was varying but includes the data of ve years prior to the survey. This further data analysis was conducted from January to February, 2021.
Based on Updated country income classi cations for the World Bank's 2020 scal year Burundi, Ethiopia, Malawi, Mozambique, Rwanda, Tanzania and Uganda are low income countries while Comoros, Kenya, Zambia and Zimbabwe are low middle income countries [12].

Data source and sampling
Data was obtained from the measure DHS program at www.measuredhs.com website after we submitted concept notes about the project. We pooled the most recent DHS data from the 11 East Africa Countries.
There are 20 countries in World Health Organization (WHO) regions of East Africa. In history, only 13 of these countries had DHS data. For this study, 11 countries were included [13] (Fig. 1).
The DHS used two stages of strati ed sampling technique to select the study participants. In the rst stage, Enumeration Areas (EAs) were randomly selected. In the second stage households were selected. We pooled the DHS survey data of the 11 East African countries, and a total weighted sample of 125,786 women who had history of bearing child in the last 5 years preceding the survey day was included in the study.

Data collection methods
The DHS program adopts standardized methods involving uniform questionnaires, manuals, and eld procedures to gather the information that is comparable across countries in the world. It is the representative household surveys that capture data from a wide range of monitoring and impact evaluation indicators in the area of population, health, and nutrition with face to face interviews of women age 15 to 49. Each country's survey consists of different datasets including men, women, children, birth, and household datasets. Detailed survey methodology and sampling methods used in gathering the data have been reported elsewhere [14]. For this study, we used the Individual record dataset (IR le) contained information of eligible women aged 15 to 49 years in each country.

Outcome variable
The outcome variable of this study was a home delivery. The response variable was generated from the question asked to the women who gave birth within 5 years preceding the survey question. The response was dichotomized as a home delivery and institutional delivery (if delivered at any type of health institutions). Home delivery includes the option given in the survey question termed as respondents home, and others home. Health institutions include government hospitals, health centers, health posts, private clinics or private hospitals. If women delivery at home, we coded as "1", otherwise coded as "0".

Independent variables
Country, age, marital status, educational level, place of residence, wealth index, sex of head of household, age of head of household, media exposure, total children ever born were included as independent variables in this study

Statistical analysis
The variables were extracted using the IR le. We use STATA software version 16.0 to clean, recode, and analysis the pooled data. After joining the extracted data from the 11 East African countries, we weighted data using women's individual sample weight (v005), and strata (v021). The proportion of home delivery was described and presented using pie chart. The DHS data had a hierarchical structure as women were nested within a cluster, and clusters with in the country. Hence the data violates the independency of the observation as the women may share similar characteristics within the cluster (and /or country). This implies that there is a need to consider the between cluster variability by using Generalized Linear Mixed Models. Likelihood Ratio test (LR), Intra-cluster Correlation Coe cient (ICC), Median Odds Ratio (MOR), and Proportional Change in Variance (PCV) were computed to measure the variation between clusters. The ICC quanti es the proportion of the total observed difference in home delivery attributable to cluster variations (degree of heterogeneity). On the other hand, MOR was used to quantify the variation or heterogeneity in home delivery between clusters. Hence, MOR is de ned as the median value of the odds ratio between the cluster high odds of home delivery and cluster at lower odds of home delivery when randomly picking out two clusters /EAs. Finally, PCV measures the total variation of home delivery attributed to individual-level and community-level factors in the nal model compared to the null model.
The detail description and formulas for ICC [15], MOR [16] and PCV [16] are described elsewhere. The null model, individual level, clusters level and both cluster and individual level factors were tted. Model comparison was made based on deviance (− 2LLR) since the models were nested. Finally, model with both at individual and cluster level factors were selected.
Variables with p-value ≤ 0.2 in the bi-variable analysis for both individual and community-level factors were tted in the multivariable model. Variables with Adjusted Odds Ratio (AOR) with a 95% Con dence Interval (CI), and p-value < 0.05 in the nal GLMM model were reported to declare signi cantly associated factors with home delivery.

Socio-demographic characteristics
In this study, a total of 125,696 women who gave birth in the 5 years preceding each country's DHS survey were included. Majority (26.93%) of women were in age group 25-29 years followed (24.05%) by 20-24 years. The median age of women was 28.3 with IQR of 23.5-33.9 years. Currently married women accounted for the large majority (85.35%) of the study participants. Half (52.99%) of women were attend primary education. More than three-fourth (77.22%) of the study participants were living in rural residents. Similarly, males were the head of the household in three out four (76.42%) of the study participants.
Country with the highest proportion of study participants was Kenya (15.48%) followed by Malawi (13.82%) and Uganda (12.14%). Whereas Comoros (2.54%) and Zimbabwe (5.10%) were countries with the smallest number of women included in the study. About two third (65.41%) of the participants reported exposure to media. The total number of children ever born from the women was ranging from 1 to 18 with mean (± SD) of 3.81 and 2.38. Majority (54.46%) of women gave 1-3 children in their life time (Table 1).  Factors associated with home delivery From tted four models (null model, individual level, cluster level and both cluster and individual level factors), model with both cluster and individual level factors was found optimal model (variance = 1.33, p < 0.001). Accordingly, respondent's age group, marital status, educational status, place of residence, living country, wealth index, media exposure and total children ever born were shown signi cantly associated with the home delivery in the East Africa counties.
In this nal best t model, about 29% of the variability among communities in the odds of a home delivery was due to the community-level factors (ICC = 28.90%) and about 24% of the variance in the odds of home delivery (PCV = 24.0%) across clusters was attributed to both individual and community-level factors. The MOR (4.96) showed that the unexplained heterogeneity between clusters (EAs) was of greater relevance than the individual variables considered in the analysis for understanding the pattern of home delivery.

Discussion
This study included a total of 125,696 women who gave birth in the 5 years preceding each resent survey conducted in East African Countries. The prevalence of Home delivery was wide-ranging among countries in Easter Africa (ranged from 2.85% in Mozambique to 72.5% in Ethiopia). The prevalence of home delivery was associated with respondent's age group, marital status, educational status, place of residence, living Country, wealth index, media exposure and total children ever born.
The pooled prevalence of home delivery from East African Region is consistent with report of India DHS report (22%) [17], and lower than studies conducted in Nigeria [18]. When compared to the Kenyan, the likelihood of home delivery was 2.24 times higher among Ethiopian. But the likelihood of home delivery was decreased by 98% for Mozambique, 96% for Malawian, 94% for Rwandan, 92 for Burundian, 85% for Zambian, 75% for Comoros, 69% for Zimbabwean, 66% for Ugandan. The geographical locations of studies varied widely with populations with differing background characteristics and social customs. In addition to social determinants, the health service coverage, quality of maternal healthcare services, economical and health policy of the country might have a role in reducing the home delivery [19].
The likelihood of home delivery was lowered by 16% among women aged 15-19 years and 35-49 years compared to middle age group women (20-35 years). This association is similar with the result of previous study [2]. Qualitative study conducted in Zambia suggested young women and those without experience in childbirth consulted their parents for the place of delivery. Other studies also reported home preference to give birth is more likely among multiparous mothers. This might they believe that they should not be delivered by either a young nurse or a male staff at the clinic [20]. Cross sectional study conducted in Ethiopia [21] also suggested preference for a health facility delivery was largely due to the understanding that if complications occurred and this was the only place where they could be managed.
Hence self-perception, women attitude and fear of complication might be the factors reduce home delivery during the early (young age) and late (advanced age) pregnancies.
Currently unmarried (never married and formerly married) women were more likely to delivery at home as compared to currently married women. Theories linking marital status, pregnancy and birth preparedness indicated that unmarried women faced lack or reduced level of psychosocial support and relationship stability [22]. Pregnant women without marriage might be unplanned and/ or unwanted. On other hand, low social acceptance of unmarried status in that there is still social stigmatism surrounding illegitimate births in many countries. Hence unmarried women may be intrinsically different from married women who may be less empowered, self-isolated or lack of motivation to access the health service [23][24][25]. All these factors might be increasing the likelihood of home delivery among currently unmarried mothers.
Educated women had less likelihood of home deliver as compared to uneducated women. The result was in line with the individual studies conducted in rural Ethiopia [26][27][28], Zimbabwe [29], Nigeria [30], Ghana [31,32], Guinea [23] and Nepal [33]. The reason for this might be due the fact that when mothers are educated it is more is likely to enhance female self-determination, positive attitude, and nancial freedom [34,35]. Furthermore, it is more likely that educated women demand higher quality service and be alert of di culties during pregnancy as well as child birth. As a result, they are more probable to use maternal health care services unlike that the illiterate one [36]. These could collectively in uence mothers' awareness to seek better maternal health services, including delivering in health facilities.
Women living rural area had as higher odds of home delivery, which is similar to ndings in previous studies [23,28,31,37]. Rural residents in Sub Saharan countries have poor access to health care facilities. Moreover, lack of privacy and con dentiality, and negligence in the provision of care during childbirth by skilled birth attendants are the fear of the women [38]. With rural healthcare provider shortages, greater travel distances and very limited access to obstetric care, it could be likely that there would be high risk of home delivery in rural areas.
In compared with women with poorest wealth status, the odds of home delivery were decreased by 26%, 40%, 55% and 73% among women with poorer, middle, richer and richest wealth status, respectively. This result was in agreement with the previous studies [26][27][28]31]. Financial capability of families and costs related to transportation may determine the place of delivery. Moreover, women from higher wealth status might be more empowered, participate in decision and seek maternal health service.
Consistent with previous studies [23,28,39,40] our nding suggests that women who exposed to media were 20% lower chances of home delivery in compared with women without media exposure. Media promotion of institutional delivery and danger of home delivery may in uence mothers to develop positive behavior towards delivering in a health facility.
The other most signi cant determinants of home delivery this study was number of children. The likelihood of home delivery was increased by 16% as the number of total children in the household is increased by one child. This nding was in line with previous studies conducted in Ethiopia [23,28]. Since the women normalize childbirth, they might be less likely to seek care during labor [41]. Literature also indicated women's previous interactions with health facilities have an in uence and poor experiences during past deliveries may discourage women from returning for the next birth [42,43]. Hence, the less fear of complication, adverse experience of care for women during childbirth could discourage them to use health facility services in subsequent pregnancies. In addition to these, Multiparous mothers who had done their previous deliveries at home might be more likely to deliver at home in their recent pregnancy [44,45].
This further analysis of data obtained from nationally representative in the East Africa DHS dataset was population-based with high response rate. The sample size used is large enough to detect the association factors with high power of the study. Hence this study is bene cial to inform policymakers and planners for their intervention to line up.
Limitations of the study include as in any cross-sectional nature of the study design the nding from this study may not found a true causal association between the home delivery and covariate. The data was collected based on self-report from mothers within 5 years prior the survey and this could be a potential source of recall bias. There was no information on numerous other important factors of maternal health service use during childbirth including the existence of emergencies during home deliveries that prompt seeking professional assistance and outcomes from past healthcare service use. Additionally, some important factors like antenatal care, obstetric histories are not included in the analysis as there was no similar de nition or classi cation among the included countries. Since some countries have no DHS program, some countries have limited data access, and some excluded due to old survey (more than ten years) the result of this study may not representative of the entire East Africa zone.

Conclusion
Home delivery was varied between countries of East African zone. The home delivery was signi cantly increased among women with middle aged, high number of ever born children, rural residence, never married or formerly married participants. To the contrast home delivery was decreased with higher educational level, media exposure, and higher wealth index.
Wide range interventions to decrease home delivery should focus on addressing inequities associated with maternal education, family wealth, increased access to media, as well as narrowing the gap between the rural and the urban areas, poor and rich families, and married and unmarried mothers.