Determinants of Place Birth: A Multinomial Logistic Regression and Spatial Analysis of the 2019 Ethiopian Mini Demographic and Health Survey Data

Background: The presence of skilled attendants at birth and institutional delivery with quality serves signicantly improves maternal and neonatal health. However, in countries where a practice of home birth is common, maternal and neonatal mortality remained high. Thus, this study aimed to determine the spatial distribution of home birth and to identify determinants of place of birth in Ethiopia. Methods: Ethiopian mini-DHS-2019 data was used in this analysis. A survey multinomial logistic regression model was used to analyze determinants of place of birth. An adjusted relative risk ratio and its 95% condence interval with a p-value of < 0.05 and marginal effect and its 95% condence interval with a p-value of < 0.05 were used to declare statistical signicance. The Global Moran’s I analysis was done by using ArcMap 10.8 to evaluate the clustering of home birth. The magnitude of home birth was predicted by ordinary kriging interpolation. Then, scanning was done by SaTScan V.9.6 software to detect scanning windows with low or high rates of home birth. Result: Prevalence of home birth in Ethiopia was 52.19% (95% CI: 46.49 – 57.83). Whereas, only 2.99% (95% CI: 1.68 – 5.25) of mothers gave birth in the health posts. Bigger family size, family wealth, multiparity, none and fewer antenatal visits, and low cluster level coverage of 4+ antenatal visits were predictors of home birth. Homebirth was clustered across enumeration areas and it was over 40% in most parts of the country with >75% in the Somali region. SaTScan analysis detected most likely clusters in the Somali region, eastern and southern zones of Oromia region, central zones of Amhara region, and eastern zones of the South Nations Nationalities and People’s region. Conclusion: Home birth is a common practice in Ethiopia. Among public health facilities, health posts are the least utilized institutions for labor and delivery care. Nationally implementing the 2016 WHO’s recommendations on antenatal care for a positive pregnancy experience and providing quality antenatal and delivery care in public facilities through qualied providers with midwifery skills and systems of back-up in place could be supportive.

Several interventions are in place to combat maternal mortality. From the three risk periods of maternal mortality namely antepartum, intrapartum and postpartum; antenatal coverage was signi cantly reduced antepartum mortality, and the presence of skilled attendants at childbirth dropped intrapartum and early postpartum mortality [7]. Amongst several planned interventions, the Ethiopian government proposed to achieve over 90% coverage of 4+ antenatal visits and delivery attended by skilled providers by 2019/20 [8]. However, the 2019 national report indicated that 43% of mothers received 4+ antenatal care and 48% gave birth in the health facilities [9].
A recent quantitative study done in Ethiopia revealed that giving birth at home is a common practice.
Rural residence, low education, economic status, not planning for place birth and unknown due date, distance to a health facility, not attending antenatal care, and low coverage of antenatal care at the community level [10][11][12] were signi cant predictors. While, socio-cultural factors such as assuming labor and delivery as a natural process, presence of enjoyable rituals during and after delivery, perceived friendly care by traditional attendants, and unavailability, inaccessibility, and perceived poor quality of modern services were qualitatively extracted factors [13][14][15] for home delivery in Ethiopia.
In most previous studies conducted in Ethiopia, the place of birth was categorized binary and health posts, the one in the primary health care systems [16] in the country, were considered as health facilities that provide basic and comprehensive obstetrics care. However, compared to other health facilities, the private and non-governmental facilities included, health posts are supposed to be less equipped with basic facilities and services to provide skilled and quality labor and delivery care. Hence, the multinomial approach could yield better estimates than binary. Also, in spatial analysis, the exclusion of visitors could result in a robust estimate. So that, the ndings of this study would inform policymakers to consider all public health institutions in plans in order to achieve local and global targets of attended births by skilled providers.

Study area
The mini-Ethiopian demographic and health survey (EDHS) was conducted in Ethiopia, a country in the Horn of East Africa. The survey is a nationwide mini-survey and included all administration areas. Since May 1991, the country is arranged in nine regional administrative states and two city administrations.
The country is further subdivided into 68 zones, 817 districts, and 16,253 kebeles (the lowest level of administration) administrative structures [9].
Data source and sampling procedure The sampling frame used in the survey was the census enumeration areas (EAs) created for the upcoming Ethiopian Population and Housing Census (PHC). The EDHS is a nationally representative two-stage cluster cross-sectional survey. As described in detail in the EDHS 2019 report [9], in the rst stage, 305 EAs (93 urban and 212 rural) were selected with probability proportional to EAs size and with independent selection of each sampling stratum (urban and rural). Then, in the second stage, 30 xed households per cluster were selected with an equal probability systematic selection. In the current analysis, as shown in the gure (Figure 1), a weighted total of 5423.31 mothers were included.

Study variables
In this study, the outcome variable was the places of birth of the most recent child and places were categorized as (1= home, 2=health post, and 3=health institution).
Health institutions are public hospitals and health centers, private hospitals and clinics, and nongovernmental organization (NGO) health facilities. These institutions are generally providing basic and comprehensive health services and delivery service is usually provided by trained health care providers with midwifery skills.
Health posts: according to the three-tier health care delivery system of Ethiopia, are among the primary health care units and are satellite sites for health centers. Each health post is expected to serve a population of 3,000 -5,000 and is a functional unit of health extension workers in rural areas [16]. Whereas, home in this study was the respondent's home or other homes where a recent child birth took place.
The independent variables used in this analysis were both individual and community-level variables. Maternal age, media access, family size, maternal education attainment, family wealth index, parity, and antenatal care utilization were among individual-level variables. Whereas, place of residence, poverty level of the community, media accessibility of the community, literacy level of the community, and ≥ 4 antenatal care visits coverage at the community/cluster level were community-level variables included in the analysis.
Community-level variables such as poverty, media access, literacy, and cluster-level ≥ 4 antenatal care visits coverage were generated by aggregating individual-level variables at the community (cluster) level. Poorest and poorer family income categories were re-categorized as 'poor'; maternal education category of no education was categorized as 'illiteracy'; and family who didn't access television or radio or both television and radio was categorized as 'no'. Then, the prevalence of these variables was divided by the cluster size, and the generated value was further categorized as 'low' and 'high' based on the median value. Four and more antenatal care visits coverage was computed the same way but nally categorized in percentages as 'below 25%', '25-50%', '51 -74%', and '≥75%'.

Statistical analysis
Sociodemographic and reproductive characteristics of the study participants and the outcome variable were described in frequency and percentage.
A survey multinomial logistic regression model was used to analyze the association between the outcome and independent variables. Individual independent variables that had an association with place of birth at a p-value of < 0.2 were considered for the nal multivariable model. The nal survey multinomial multivariable model was selected based on the log likely (LL) ratio and the one with the highest LL ratio was selected. In the nal model, an adjusted relative risk ratio (aRRR), its 95% con dence interval, and a p-value of < 0.05 and marginal effect and its 95% con dence interval, and a p-value of < 0.05 were interpreted and used to declare statistical signi cance [17].

Spatial analysis
The Global Moran's I analysis was done by using ArcMap 10.8 to evaluate whether home birth is clustered, random, or dispersed across the study areas. Since home birth was clustered, spatial interpolation by using ArcMap 10.8 and scan statistics by using a SaTScan V.9.6 were carried out to predict the magnitude and to detect clusters and a scanning window with low or high rates of home birth.
Nearly two-thirds (62.08%) of the mothers who didn't expose themselves to media access in their household had given birth at home. Similarly, two-third and more mothers, whose family size was greater than six members were delivered at home. Giving birth at home showed a decreasing prevalence as mothers' level of education and the family wealth index increases.
While most (86.34%) of the mothers who hadn't get antenatal care, gave birth at home. The majority (66.35%) of grand multiparous mothers similarly delivered at home. Almost one-third of urban and twothird of rural residents gave home birth for their most recent delivery. From the regions in Ethiopia, Afar and Somali were the most common home birth regions in the country (Table 1).

Spatial distribution of home birth in Ethiopia
A clustering pattern of home birth was revealed in the global spatial autocorrelation across the EAs (Moran's index = 0.667563, z-score = 14.541580, p-value < 0.001) ( Figure 2). In addition, the ordinary kriging interpolation analysis predicted that home birth was relatively about 40% and higher in most parts of the country and more than 75% of home delivery was widely distributed in the Somali region ( Figure 3).
Also, the SaTScan analysis detected a total of seven statistically signi cant cluster areas with a high magnitude of home birth. The most likely primary cluster areas with the highest home birth were detected in the Somali region, eastern and southern zones of Oromia region with a relative risk (RR)=1.72, and a pvalue of <0.001. In addition, the most likely secondary cluster areas with a high magnitude of home birth were spotted in the central zones of Amhara and eastern zones of South Nations, Nationalities, and People's Region (SNNPR) (Figure 4, Table 2). The survey multinomial multivariable analysis identi ed that the relative probability of giving birth at home rather than health facility was about one and half times higher for mothers who had a family size of six to ten members than less than six members (aRRR = 1.46 (95% CI: 1.10, 1.93)). The marginal effect analysis also indicated that the probability of giving birth at home was on average ve percentage (0.05 (0.01, 0.10)) points higher for mothers who had a family size of six to ten members. Whereas, the relative probability of giving birth at health post rather than health institution was 0.02 (aRRR = 0.02 (95% CI: 0.003, 0.20)) for mothers who had a family size greater than ten than less than six members implies that the probability of giving birth at health post on average three percentage (-0.03 (95% CI: -0.04, -0.009)) points lower for mothers who had largest family size than lowest family size.
The family wealth index was also found to be a predictor for a home birth. As compared to the richest family, the relative probability of giving birth at home rather than health facility among mothers was about two times higher for richer aRRR 2.13 (95% CI: 133., 3.43), more than four times higher for the middle (aRRR = 4.29 (95% CI: 2.68, 6.89)) and poorer (aRRR = 4.60 (95% CI: 2.70, 7.85)), and ten times higher for poorest (aRRR = 10.08 (95% CI: 5.66, 17.98)) family. As shown by the marginal effect analysis, the probability of giving birth at home among mothers was higher at 12 percentage points for richer, around 25 percentage points for middle and poorer families. Whereas, home birth was 38 percentage points higher for the poorest family.
In addition, the relative probability of giving birth at home rather than health facility was nearly twice higher for multiparous (aRRR = 1.95 (95% CI: 1.20, 3.15)) and grand multiparous (aRRR = 1.93 (95% CI: 1.08, 3.43)) mothers than primiparous mothers. The marginal effect analysis also revealed that the probability of giving birth at home was about ten percentage points higher among multiparous and grand multiparous than primiparous mothers.
Moreover, as compared to mothers who attended four and more antenatal care visits, the relative probability of giving birth at home rather than at a health facility was more than one times (aRRR = 1.59 (95% CI: 1.22, 2.07)) higher for mothers who attended less than four antenatal care visits and over six times (aRRR = 6.31 (95% CI: 4.27, 9.32)) higher for mothers who didn't attend antenatal care during the index pregnancy. In the marginal effect analysis, the probability of giving birth at home was eight percentage points higher among mothers who attended less antenatal care visits and 30 percentage points higher among mothers who didn't attend antenatal care than who attended four and more antenatal care visits. Likewise, the probability of giving birth at health posts on average two percentage (-0.02 (95% CI: -0.04, -0.001)) points lower for mothers who didn't attend antenatal care.
Similarly, the relative probability of giving birth at home rather than health facility was more than two times (aRRR = 2.25 (95% CI: 1.20, 4.20)) higher among mothers who residing in the clusters in which 51 -74% of mothers attended 4+ antenatal care visits than those residing in the clusters of ≥ 75% 4+ antenatal care visits coverage. And it was about ve times (aRRR = 4.88 (95% CI: 2.40, 9.9)) higher among mothers who residing in the clusters in which 25 -50% of mothers attended 4+ antenatal care visits and about seven times (aRRR = 749 (95% CI: 3.54, 15.85)) higher among mothers who residing in the clusters in which < 25% of mothers attended 4+ antenatal care visits as compared to those who residing in the clusters of ≥ 75% 4+ antenatal care visits coverage. The marginal analysis also showed that the probability of giving birth at home was 37 percentage points higher among mothers who residing in the clusters in which < 25% of mothers attended 4+ antenatal care visits, 30 percentage points higher among mothers who residing in the clusters in which 25 -50% of mothers attended 4+ antenatal care vists, and 14 percentage points higher among mothers who residing in the clusters in which 51 -74% of mothers attended 4+ antenatal care visits as compared to mothers who residing in the clusters in which ≥ 75% of mothers attended a coverage of 4+ antenatal care visits (Table 3).  [12]. As explored by qualitative ndings, home birth is common due to cultural reasons.
Most society and women believe that labor and delivery is a natural process and the ritual processes during labor and delivery at home are pleasant [13]. Study participants further pointed out that mothers lack such joyful customs at health facilities [14].
Mothers from a family size above ve members and those who were multiparous inclined to give birth at home than smaller family size and nulliparous mothers. Family size is directly related to birth order and similar studies also identi ed that higher birth order and multiparity were found to be signi cant factors for home birth [12]. Scienti c explanations for the relation of parity and home birth are de cient. It could be explained by birthing experience, experience from previous health facility birth, and cultural reasons as revealed by qualitative ndings [13,14,18,19].
This study identi ed that antenatal care attendance at an individual level and its frequency, as well as community-level coverage of 4+ antenatal care, played a signi cant role in determining the place of birth.
Several small-and large-scale studies also revealed that receiving no antenatal care [12] and delay in receiving antenatal care [20] were signi cantly associated with home delivery. Homebirth also signi cantly contributed by late entry to antenatal care [21] and receiving fewer than four antenatal care [22,23] among antenatal care booked mothers. In countries like Ethiopia in which preconception care is not in place, antenatal care is an important entry for the continuum of maternity care. Pieces of evidence revealed that antenatal care when provided with a minimum recommended quality, it found to increase the likelihood of institutional delivery in developing countries [24,25].

Conclusion
In Ethiopia, home birth is a common practice. In contrast, health posts, which are community-level governmental health units, are the least utilized facilities for labor and delivery service. Individual-level none and fewer visits of antenatal care and lower cluster level coverage of 4+ antenatal care played a signi cant role in predicting home birth in Ethiopia. Nationally adapting the 2016 WHO's recommendations on antenatal care for a positive pregnancy experience and providing quality antenatal and delivery care in public facilities by quali ed providers with midwifery skills and systems of back-up in place could be helpful. Also, piloting the bene ts of planned home birth with quali ed professionals for low-risk pregnancies could be worth more. Formal ethical approval was not required in this secondary data analysis. The DHS program was communicated for the data set used in this analysis and permission was granted to download and use the data from https://dhsprogram.com/Data/terms-of-use.cfm. The geographic identi ers were limited at the regional and EAs level that was a large geographical area. And individual-level variables and the aggregated community variables were not included any personal identi ers like names, house numbers, and phone numbers.

Consent for publication
Not applicable Availability of data and materials The datasets generated and/or analyzed during the current study are available in the [the DHS program] repository, [https://dhsprogram.com/]

Competing interests
The author declares that he has no competing interests

Funding
The author declares that he has no source of funding involved in this secondary data analysis Author's contributions TWG obtained permission to use the dataset, identi ed research question, reviewed available works of literature, analyzed and interpreted results, wrote up the manuscript, and reviewed and approved the manuscript.