Individual and community-level determinants, and spatial distribution of institutional delivery in Ethiopia, 2016: Spatial and multilevel analysis.

BACKGROUND: Institutional delivery is an important indicator in monitoring the progress towards Sustainable Development Goal 3.1 to reduce the global maternal mortality ratio to less than 70 per 100,000 live births. Despite the international focus on reducing maternal mortality, progress has been low, particularly in Sub-Saharan Africa (SSA), with more than 295,000 mothers still dying during pregnancy and childbirth every year. Institutional delivery has been varied across and within the country. Therefore, this study aimed to investigate the individual and community level determinants, and spatial distribution of institutional delivery in Ethiopia.
METHODS: A secondary data analysis was done based on the 2016 Ethiopian Demographic and Health Survey (EDHS) data. A total weighted sample of 11,022 women was included in this study. For spatial analysis, ArcGIS version 10.6 statistical software was used to explore the spatial distribution of institutional delivery, and SaTScan version 9.6 software was used to identify significant hotspot areas of institutional delivery. For the determinants, a multilevel binary logistic regression analysis was fitted to take to account the hierarchical nature of EDHS data. The Intra-class Correlation Coefficient (ICC), Median Odds Ratio (MOR), Proportional Change in Variance (PCV), and deviance (-2LL) were used for model comparison and for checking model fitness. Variables with p-values<0.2 in the bi-variable analysis were fitted in the multivariable multilevel model. Adjusted Odds Ratio (AOR) with a 95% Confidence Interval (CI) were used to declare significant determinant of institutional delivery.
RESULTS: The spatial analysis showed that the spatial distribution of institutional delivery was significantly varied across the country [global Moran's I = 0.04 (p<0.05)]. The SaTScan analysis identified significant hotspot areas of poor institutional delivery in Harari, south Oromia and most parts of Somali regions. In the multivariable multilevel analysis; having 2-4 births (AOR = 0.48; 95% CI: 0.34-0.68) and >4 births (AOR = 0.48; 95% CI: 0.32-0.74), preceding birth interval ≥ 48 months (AOR = 1.51; 95% CI: 1.03-2.20), being poorer (AOR = 1.59; 95% CI: 1.10-2.30) and richest wealth status (AOR = 2.44; 95% CI: 1.54-3.87), having primary education (AOR = 1.47; 95% CI: 1.16-1.87), secondary and higher education (AOR = 3.44; 95% CI: 2.19-5.42), having 1-3 ANC visits (AOR = 3.88; 95% CI: 2.77-5.43) and >4 ANC visits (AOR = 6.53; 95% CI: 4.69-9.10) were significant individual-level determinants of institutional delivery while being living in Addis Ababa city (AOR = 3.13; 95% CI: 1.77-5.55), higher community media exposure (AOR = 2.01; 95% CI: 1.44-2.79) and being living in urban area (AOR = 4.70; 95% CI: 2.70-8.01) were significant community-level determinants of institutional delivery.
CONCLUSIONS: Institutional delivery was low in Ethiopia. The spatial distribution of institutional delivery was significantly varied across the country. Residence, region, maternal education, wealth status, ANC visit, preceding birth interval, and community media exposure were found to be significant determinants of institutional delivery. Therefore, public health interventions should be designed in the hotspot areas where institutional delivery was low to reduce maternal and newborn mortality by enhancing maternal education, ANC visit, and community media exposure.

Background

Despite improvements over the last two decades, maternal mortality in developing countries, especially in sub-Saharan Africa (SSA), remains a significant public health concern [1, 2]. Globally, as a result of preventable causes of pregnancy and childbirth, about 358,000 maternal deaths occur annually, of which 99% occur in developing countries [3]. In high-income countries, maternal mortality has been decreased dramatically [4], but SSA continued to account for 66% of maternal deaths worldwide [5]. In SSA, like Ethiopia, pregnancy and childbirth-related complications such as Postpartum Hemorrhage (PPH), pregnancy-induced high blood pressure, fetal asphyxia, stillbirth, sepsis, obstructed labor, and unsafe abortions are unacceptably high, leading to the massive burden of maternal mortality [6-8].

The World Health Organization (WHO) recommends health facility delivery as a key strategy for reducing maternal and neonatal mortality [9]. Institutional delivery grants safe birth outcomes through the provisions of supportive facilities, clean delivery services with skilled experts, and early detection and management of maternal and neonatal complications [10]. Although institutional delivery are a key strategy for reducing pregnancy and birth risks, many women in developing countries give birth at home [11]. For example, the prevalence of institutional delivery in Asia and SSA is lower than 50% [12]. Thus, it varies from 26% in Ethiopia [13] to 67.3% in Tanzania [1].

According to prior studies conducted in Ethiopia, the prevalence of institutional delivery varied across the country. Studies conducted on the prevalence and associated factors of institutional delivery in different regions of Ethiopia showed that 18.2% of the mothers in the Oromia region [14], 4.1% in the Tigray region [15], 78.8% in the Amhara region [16], 62.2% in southern Ethiopia [17], 31% in the Gurage zone [18], 14.4% in West Shewa [19], 18.3% in Northwest Ethiopia [20] gave birth at the health facility.

Previous literature revealed that sex of household head, maternal age, maternal occupation, parity, birth order [21-25], number of Antenatal Care (ANC) visits [1, 12, 26–28], knowledge towards danger signs of pregnancy and childbirth [1, 21, 22], household wealth index [1, 25, 29], media exposure [30], maternal and parental education [1, 26, 29], previous history of prolonged labour [31], number of children [31, 32], birth preparedness/complication readiness [32, 33], and decision making on health care [8, 9, 12, 16] were the individual-level predictors significantly associated with institutional delivery. Studies also documented that community-level factors such as region, residence [25, 26, 32], distance to the nearest health facility, and community media exposure [2, 34] were significantly associated with institutional delivery.

Despite Ethiopia having made a large scale investment to reduce maternal and neonatal mortality through free of charge maternal health care services such as ANC, institutional delivery, and PNC [35, 36], still maternal and newborn mortality is highest in Ethiopia [9]. It is common in rural parts of the county, where it is more challenging to get access to health facilities, and home delivery is highly practiced [13].

In Ethiopia, previous studies were done on the prevalence and associated factors of institutional delivery [28, 30, 37] and reported that the prevalence had been varied across the country [15–19, 38] but none of these studies have tried to explore the spatial distribution of institutional delivery in Ethiopia. Besides, there are two studies on institutional delivery based on the nationally representative Ethiopian Demographic and Health Survey (EDHS) data [28, 30]. These studies were failed to capture the spatial distribution of institutional delivery in Ethiopia, and the data they used were not weighted data. Therefore, we aimed to investigate the individual and community-level determinants, and spatial distribution of institutional delivery in Ethiopia based on weighted 2016 EDHS data. Thus, the identifications of significant hotspot areas with a low prevalence of institutional delivery have become indispensable to design targeted effective public health interventions to enhance institutional delivery and reduce maternal and newborn mortality in Ethiopia. Furthermore, this study’s findings could guide policymakers to work on individual and community-level determinants to improve institutional delivery in the country to strengthen maternal and child health.

Methods

Study design, setting, and period

A secondary data analysis was done based on the 2016 EDHS data. The EDHS was a nationally representative study conducted every five years in Ethiopia. Ethiopia is situated in the Horn of Africa. It has 9 Regional states (Afar, Amhara, Benishangul-Gumuz, Gambela, Harari, Oromia, Somali, Southern Nations, Nationalities, and People’s Region (SNNP) and Tigray regions) and two city Administrations (Addis Ababa and Dire-Dawa) (Fig 1). About 84% of the population lives in rural areas [39]. In EDHS 2016, a two-stage stratified cluster sampling technique was employed using the 2007 Population and Housing Census (PHC) as a sampling frame. In the first stage, 645 EAs (202 in the urban area) were selected, and in the second stage, on average 28 households were systematically selected. A total of 18,008 households and 16,583 eligible women were included. The detailed sampling procedure was presented in the full EDHS 2016 report [40]. The source population was all women of reproductive age who gave birth in Ethiopia within five years before the survey, while the sample population was all women of reproductive age who gave birth in the selected EAs within five years before the survey. A total weighted sample of 11,022 reproductive-age women who gave birth within five years preceding the survey was included in this study.

Fig 1

Map of the study area (Source, CSA: 2013).

Study variables

Outcome variable

The dependent variable was whether a woman who gave birth within five years preceding the survey was delivered at a health facility or at home. We used the "place of delivery" as the outcome variable and recoded as home delivery (when the birth took place at home) and institutional delivery (when the birth took at the hospital, health center, or health post).

Independent variables

Consistent with the study’s objective and given the hierarchical structure of EDHS data where women were nested within the cluster, two levels of independent variables were considered. At level-1 contained individual-level variables such as age, maternal education, husband education, media exposure, wealth index, sex of household head, ANC visit, parity, preceding birth interval, multiple gestations, religion, ever had of a terminated pregnancy, and birth order was included. At level-2 the community-level variables considered in this study were region, residence, community media exposure, and distance to get health facility.

In EDHS data, there was no variable collected at the community level except region (recoded as pastoralist region (Benishangul, Somali, Gambella, and Afar), Semi-pastoralist (Oromia, SNNPR), Agrarian (Amhara and Tigray) and City administration (Addis Ababa, Dire Dawa, and Harari)), distance to get health facility (recorded as a big problem and not a big problem), and residence (recoded as urban and rural). Therefore, we generated community media exposure by aggregating listening radio, watching television, and reading newspapers at the cluster level. These were categorized as higher community media exposure and lower media exposure based on the national median value of media exposure since it was not normally distributed [41].

Data management and analysis

The data were weighted using sampling weight, primary sampling unit, and strata before any statistical analysis to restore the representativeness of the survey and to tell the STATA to take into account the sampling design when calculating standard errors, to get reliable statistical estimates. Descriptive and summary statistics were conducted using STATA version 14 software.

Spatial analysis

Spatial autocorrelation analysis. ArcGIS version 10.6 software was used to explore the spatial distribution of institutional delivery. The global spatial autocorrelation (Global Moran’s I) was done to assess whether institutional delivery patterns were dispersed, clustered, or randomly distributed in the study area [42]. Moran’s I is a spatial statistic used to measure spatial autocorrelation by taking the entire data set and producing a single output value ranging from -1 to +1. Moran’s I value close to −1 indicates the spatial distribution of institutional delivery is dispersed, whereas Moran’s I close to +1 indicate spatial distribution of institutional delivery is clustered. The Moran I value close to 0 means the spatial distribution of institutional delivery is random. A statistically significant Moran’s I (p < 0.05) indicates the spatial clustering of institutional delivery.

Spatial interpolation. The spatial interpolation was done to predict institutional delivery on the un-sampled areas in the country based on sampled measurements. Ordinary Kriging (OK) and Empirical Bayesian Kriging (EBK) were done since it statistically optimizes the weight [43], to predict the prevalence of institutional delivery on the unobserved areas based on the observed measurement. The ordinary Kriging spatial interpolation method was selected for this study for predictions of institutional delivery since it had a smaller residual and Root Mean Square Error (RMSE) than EBK.

Spatial scan statistical analysis. In the spatial scan statistical analysis, Bernoulli based model was employed to identify statistically significant spatial clusters of institutional delivery using Kuldorff’s SaTScan version 9.6 statistical software. For this study, we used a circular scanning window that moves across the study area since the elliptical window is inactive in the SaTScan software. Women with home delivery were taken as cases and those who had institutional delivery were considered as controls to fit the Bernoulli model. The numbers of cases in each location had Bernoulli distribution and the model required data for cases, controls, and geographic coordinates. The default maximum spatial cluster size of <50% of the population was used as an upper limit, which allowed both small and large clusters to be detected and ignored clusters that contained more than the maximum limit.

For each potential cluster, a likelihood ratio test statistic and the p-value were used to determine if the number of observed home delivery within the potential cluster was significantly higher than expected or not. The scanning window with maximum likelihood was the most likely performing cluster, and the p-value was assigned to each cluster using Monte Carlo hypothesis testing by comparing the rank of the maximum likelihood from the real data with the maximum likelihood from the random datasets. The primary and secondary clusters were identified and assigned p-values and ranked based on their likelihood ratio test based on 999 Monte Carlo replications [44].

Multilevel analysis

There is a hierarchical nature of the EDHS data; therefore, women have been nested within a cluster, and we assume that women in the same cluster may share similar characteristics to women in another cluster. These violate the usual hypothesis of the logistic regression model, which is the independence of observations and equal variance between clusters. This implies the need to take into account the heterogeneity between clusters by using an advanced model. Therefore, a multilevel binary logistic regression model was performed.

The ith mother’s response variable is represented by a random variable Yi with two possible values coded as 1 and 0. So, the ith mother Yi’s response variable was measured as a dichotomous variable with possible values Yi = 1, if ith mother gave birth in the institution and Yi = 0 if a mother gave birth in their home. We will fit the multilevel model by

Where:

IIij: the probability of having institutional delivery

1 − πij: the probability of having home delivery

β0: the intercept

β1/Bn: regression coefficient of individual and community level factors

u0j: random errors at cluster levels

e0ij: random error at the individual level

Model comparison was made based on deviance (-2LL) since the models were nested models, and a model with the lowest deviance was the best-fitted model for the data. Likelihood Ratio (LR) test, Intra-class Correlation Coefficient (ICC), Median Odds Ratio (MOR), and Proportional Change in Variance (PCV) were computed to measure the variation of institutional delivery between clusters. The ICC quantifies the degree of heterogeneity of institutional delivery between clusters (the proportion of the total observed variation in institutional delivery that is attributable to between cluster variations).

ICC = ϭ2/ (ϭ2+π2/3) [45], but MOR quantifies the variation or heterogeneity in institutional delivery between clusters in terms of odds ratio scale and is defined as the median value of the odds ratio between the cluster at high likelihood of institutional delivery and cluster at lower risk when randomly picking out individuals from two clusters (EAs). [46].

∂2 indicates that cluster-level variance

PCV measures the total variation attributed to the final multilevel model as compared to the null model. We calculated the percentage of the Proportional Change in Variance (PCV) as follows

Where; var (null model) = variance of the initial model, and var (final model) = variance of the final model. PCV measures the variation in institutional delivery explained by the full model (a model with both individual and community level variables simultaneously). Total variance was calculated by adding individual level variance (π2/3) and community level variance, as individual level variable binary model is π2/3 (3.29).

A two-level multilevel binary logistic regression model was used to analyze factors associated with institutional delivery. Four models were constructed for the multilevel logistic regression analysis. The first model was a null model without explanatory variables to determine the extent of cluster variation in institutional delivery. The second model was adjusted with individual-level variables; the third model was adjusted for community-level variables while the fourth was fitted with both individual and community level variables simultaneously.

Variables with p-value <0.2 in the bi-variable analysis for both individual and community-level factors were fitted in the multivariable model. We used 0.2 because incorporating variables with p-value up to 0.2 is important since these variables might have a good contribution in the multivariable analysis. Adjusted Odds Ratio (AOR) with a 95% Confidence Interval (CI) in the multivariable model were used to declare statistically significant determinants of institutional delivery. Multi-collinearity was also checked using the variance inflation factor (VIF) by doing pseudo linear regression analysis and indicates that there was no multi-collinearity since all variables have VIF <5 and tolerance greater than 0.1.

Ethical consideration

Since the study was a secondary data analysis of publicly available survey data from the MEASURE DHS program, ethical approval and participant consent were not necessary for this particular study. We requested DHS Program and permission was granted to download and use the data for this study from http://www.dhsprogram.com. There were no names of individuals or household addresses in the data file. The geographic identifiers only go down to the regional level (where regions are typically very large geographical areas encompassing several states/provinces). Each enumeration area (Primary Sampling Unit) has a PSU number in the data file, but the PSU numbers do not have any labels to indicate their names or locations. In surveys that collect GIS coordinates in the field, the coordinates are only for the enumeration area (EA) as a whole, not for individual households. The measured coordinates are randomly displaced within a large geographic area so that specific enumeration areas cannot be identified.

Results

Socio-demographic and economic characteristics of participants

A total of 11,022 reproductive-age women who gave birth within five years preceding the survey were included in this study. Of these, 4,851 (44.0%) were from Oromia region and 26 (0.25%) were from Harari region. About 9,807 (89.0%) of the women were living in rural areas, and the majority (41.4%) of the respondents were Muslim followers. Nearly two-thirds (66.1%) of the women and a half (47.8%) of their husbands didn’t have formal education. Regarding the age of the women, 7,910 (71.8%) were in the age group of 20–34 years (Table 1).

Table 1

Socio-demographic and economic characteristics of respondents in Ethiopia, 2016.

Variable	Frequency (N = 11,022)	Percentage
Region
Tigray	716	6.5
Afar	114	1.0
Amhara	2,072	18.8
Oromia	4,851	44.0
Somali	508	4.6
Benishangul	122	1.1
SNNPs	2,296	20.8
Gambella	27	0.2
Harari	26	0.2
Addis Ababa	244	2.2
Dire Dawa	47	0.4
Residence
Urban	1,215	11.0
Rural	9,807	89.0
Religion
Orthodox	3,772	34.2
Muslim	4,561	41.4
Catholic	103	0.9
Protestant	2,329	21.1
Traditional	257	2.3
Maternal age (in years)
< 20	378	3.4
20–34	7,910	71.8
≥ 35	2,734	24.8
Maternal education
No education	7,284	66.1
Primary education	2,951	26.8
Secondary education	514	4.7
Higher education	274	2.5
Husband education
No education	5,003	47.8
Primary	4,115	39.3
Secondary	797	7.6
Higher	544	5.3
Wealth status
Poorest	2,636	23.9
Poorer	2,520	22.9
Middle	2,280	20.7
Rich	1,998	18.1
Richest	1,588	14.4
Media exposure
No	7,375	66.9
Yes	3,647	33.1
Sex of household head
Male	9,494	86.1
Female	1,528	13.9

Obstetric and maternal service-related characteristics of respondents

Nearly half (43.9%) of women had 2–4 births, and about 91.2% had no prior pregnancy termination history. Of the total, 4,738 (43.0%) of women had a preceding birth interval of 24 to 48 months, and the majority (60.6%) of respondents reported as the distance to reach a health facility was a big problem (Table 2).

Table 2

Obstetric and maternal service-related characteristics of respondents in Ethiopia, 2016.

Variable	Frequency	Percentage
Parity
1	1,434	13.0
2–4	4,836	43.9
5+	4,752	43.1
Multiple gestation
No	10,730	97.4
Yes	292	2.6
Preceding birth interval
< 24 month	1,942	17.6
24–48 month	4,738	43.0
>48 month	4,343	39.4
Distance to health facility
Big problem	6,676	60.6
Not a big problem	4,346	39.4
Ever had of a terminated pregnancy
No	10,056	91.2
Yes	966	8.8
Birth order
1	2,058	18.7
2	1,784	16.2
≥3	7,180	65.1
Number of ANC visit
None	2,818	37.1
1–3 visit	2,342	30.9
≥ 4 visits	2,429	32.0

Regional prevalence of institutional delivery in Ethiopia, 2016

The overall prevalence of institutional delivery in Ethiopia was 26.2% [95 CI: 25.4%, 27.1%], which was significantly varied across regions ranging from 14.7% in the Afar region to 96.6% in Addis Ababa (Fig 2).

Fig 2

Regional prevalence of institutional delivery in Ethiopia, 2016.

Spatial analysis

Spatial autocorrelation analysis

The global spatial autocorrelation analysis revealed that the spatial distribution of institutional delivery was significantly varied across the country with Global Moran’s Index value of 0.04 (p<0.05) (Fig 3). In this study, areas with a low prevalence of institutional delivery were identified in Addis Ababa, Dire-Dawa, and Tigray regions. In contrast, areas with a high prevalence of institutional delivery were detected in Amhara, Afar, Somali, and Gambella regions (Fig 4).

Fig 3

Global autocorrelation of institutional delivery in Ethiopia, 2016.

Fig 4

Spatial distribution of institutional delivery in Ethiopia, 2016 (Source, CSA: 2013).

Spatial interpolation

In the Kriging interpolation analysis, the highest prevalence of institutional delivery was detected in Addis Ababa, Dire Dawa, Harari, central Gambella, and Tigray regions. In contrast, the predicted low prevalence of institutional delivery was identified in Afar, east Somali, southwest Oromia, Benishangul, and Amhara regions (Fig 5).

Fig 5

Kriging interpolation of institutional delivery in Ethiopia, 2016 (Source, CSA: 2013).

Spatial scan statistical analysis

A spatial scan statistical analysis identified a total of 331 significant clusters, of which 104 were most likely (primary) clusters, and 227 were secondary clusters. The primary clusters were located in Harari, south Oromia, and most parts of Somali regions centered at 4.180558 N, 42.052871 E with 567.56 km radius, a Relative Risk (RR) of 1.24 and Log-Likelihood Ratio (LLR) of 106.5, at p < 0.0001 (Fig 6, Table 3). It showed that women inside the spatial window had a 1.24 times higher likelihood of having home delivery than women outside the spatial window.

Fig 6

SaTScan analysis of hotspot areas of poor institutional delivery (home delivery) in Ethiopia, 2016 (Source, CSA: 2013).

Table 3

SaTScan analysis result of home delivery.

Cluster	Enumeration area(cluster)identified	Coordinate/radius	Population	Case	RR	LLR	p-value
1 (104)	520, 208, 556, 394, 278, 164, 187, 480, 377, 318, 7, 358, 85, 138, 82, 289, 492, 286, 146, 422, 92, 543, 472, 490, 601, 452, 171, 198, 34, 95, 398, 316, 497, 518, 405, 21, 468, 313, 232, 600, 576, 445, 182, 26, 521, 574, 588, 562, 32, 123, 553, 458, 634, 365, 619, 213, 12, 319, 589, 215, 216, 308, 391, 408, 50, 148, 214, 578, 529, 251, 573, 245, 77, 239, 524, 503, 522, 116, 372, 22, 342, 347, 438, 609, 476, 122, 505, 20, 420, 162, 568, 412, 277, 86, 53, 513, 454, 373, 180, 580, 68, 506, 450, 501	(4.180558 N, 42.052871 E) / 567.56 km	2359	1877	1.24	106.5	<0.0001
2 (91)	520, 208, 556, 394, 278, 164, 187, 480, 377, 318, 7, 358, 85, 138, 82, 289, 492, 286, 146, 422, 92, 543, 472, 490, 601, 452, 171, 198, 34, 95, 398, 316, 497, 518, 405, 21, 468, 313, 232, 600, 576, 445, 182, 26, 521, 574, 588, 562, 32, 123, 553, 458, 634, 365, 619, 213, 12, 319, 589, 215, 216, 308, 391, 408, 50, 148, 214, 578, 529, 251, 573, 245, 77, 239, 524, 503, 522, 116, 372, 22, 342, 347, 438, 609, 476, 122, 505, 20, 420, 162, 568	(4.180558 N, 42.052871 E) / 558.39 km	2035	1630	1.24	96.96	<0.001
3 (36)	4, 632, 75, 596, 440, 366, 178, 499, 205, 427, 334, 570, 348, 599, 544, 389, 368, 241, 55, 547, 191, 571, 344, 276, 332, 189, 254, 37, 249, 620, 488, 307, 135, 611, 345, 283	(11.845228 N, 41.915793 E) / 242.50 km	697	617	1.34	88.44	<0.001
4 (91)	109, 3, 361, 498, 515, 382, 516, 615, 429, 541, 375, 548, 431, 167, 602, 246, 533, 494, 474, 403, 559, 386, 259, 73, 24, 169, 415, 36, 150, 184, 456, 158, 183, 120, 531, 218, 137, 512, 244, 292, 364, 132, 482, 206, 35, 229, 350, 320, 163, 38, 161, 176, 88, 627, 294, 399, 279, 10, 280, 70, 545, 640, 327, 256, 510, 124, 52, 621, 517, 65, 349, 267, 460, 234, 569, 152, 312, 199, 638, 335, 485, 304, 457, 423, 118, 209, 572, 324, 23, 563, 628	(10.934452 N, 36.945496 E) / 252.94 km	1434	1135	1.34	54.32	<0.01
5 (9)	566, 1, 186, 622, 8, 436, 210, 212, 419	(9.455401 N, 42.455144 E) / 33.24 km	240	225	1.40	49.81	<0.01
6 (49)	207, 154, 477, 489, 76, 338, 586, 177, 325, 437, 376, 168, 552, 459, 243, 299, 465, 371, 554, 470, 486, 526, 432, 197, 119, 46, 447, 555, 306, 227, 326, 62, 113, 411, 406, 141, 337, 126, 502, 434, 558, 565, 448, 180, 142, 331, 41, 360, 450	(7.220845 N, 36.133859 E) / 180.87 km	903	731	1.22	43.96	0.06
7 (12)	266, 618, 309, 435, 536, 370, 507, 592, 104, 260, 233, 69	(8.389747 N, 33.258557 E) / 71.61 km	203	186	1.37	34.11	0.08
8 (3)	130, 511, 172	(13.169308 N, 39.987117 E) / 10.69 km	82	82	1.49	32.27	0.1

Individual and community-level determinants of institutional delivery

The random effect analysis result

In the null model, the ICC indicated that 57% of the total variability for institutional delivery was due to differences between clusters while the remaining unexplained 43% of the total variability of institutional delivery was attributable to the individual differences. Moreover, the MOR was 7.01 (95% CI: 6.02, 9.07) in the null model, which indicated that there was variation in institutional delivery between clusters. If we randomly select two women from different clusters, if we transfer women from low institutional delivery clusters to higher institutional delivery clusters, she could have 7.01 times higher odds of having institutional delivery. The PCV in the final model was 73%, it showed that about 73% of the variability in institutional delivery was explained by the full model (a model with individual and community level variables). Deviance was used to compare the fitted models and model 3 with the lowest deviance value was the best-fitted model (Table 4).

Table 4

Random effect analysis result.

Parameter	Null model	Model1	Model2	Model3
Community level variance (SE)	4.41 (0.47)	1.37 (0.16)	1.65 (0.19)	1.21(0.153)
Log likelihood	-4737.52	-3015.16	-4491.17	-2952.70
Deviance	9475.04	6,030.32	8982.34	5905.40
MOR	7.01 [6.02, 9.17]	3.05[2.70, 3.49]	3.39[2.97, 3.92]	2.84 [2.52, 3.27]
PCV	Ref	0.69	0.62	0.73
ICC	0.57	0.29	0.33	0.27

The fixed effect analysis result

In the multivariable multilevel logistic regression analysis parity, preceding birth interval, the number of ANC visits, wealth status, residence, community media exposure, region, and maternal education were significantly associated with institutional delivery.

The odds of having institutional delivery among women who had 2–4 births and more than four births were decreased by 62% (AOR = 0.48; 95% CI: 0.34–0.68) and 62% (AOR = 048; 95% CI: 032–0.74) as compared to primiparous women respectively. Women who had preceding birth interval ≥ 48 months had 1.51 (AOR = 1.51; 95% CI: 1.03–2.20) times higher odds of giving birth at health institutions compared to women who had preceding birth interval less than 24 months. The odds of having institutional delivery for women who had primary education, and secondary and above education were 1.47 (AOR = 1.47; 95% CI: 1.16–1.87) and 3.44 (AOR = 3.44; 95% CI: 2.19–5.42) times more likely to have institutional delivery than women who had no formal education respectively.

Women in the poor and richest households had 1.57 (AOR = 1.57; 95% CI: 1.10–2.30) and 2.44 (AOR = 2.44; 95% CI: 1.54–3.87) times higher odds of having institutional delivery than women in the poorest household, respectively. Mother who had 1–3 ANC visit and ≥ 4 ANC visit for the index pregnancy was 3.88 (AOR = 3.88; 95% CI: 2.77–5.43) and 6.53 (AOR = 6.53; 95% CI: 4.69–9.10) times higher odds of having institutional delivery as compared to mother who had no ANC visit.

Regarding regions, women residing in city administrations (Addis Ababa and Dire Dawa) had 3.13 (AOR = 3.13; 95% CI: 1.77–5.55) times higher odds of institutional delivery as compared to women residing in pastoral regions. Women from communities with high media exposure had 2.01(AOR = 2.01; 95% CI: 1.44–2.79) times higher odds of institutional delivery as compared to women from a community with low media exposure. Besides, urban women had 4.70 (AOR = 4.70; 95% CI: 2.70–8.01) times higher odds of having institutional delivery as compared to rural residents (Table 5).

Table 5

Multivariable multilevel logistic regression analysis of individual and community level determinants of institutional delivery in Ethiopia, 2016.

Variable	Null model	Model 1	Model 2	Model 3
Individual level factors
Parity
1		1		1
2–4		0.49 [0.39, 0.62]**		0.48 [0.34, 0.68]**
>4		0.46 [0.34, 0.61]**		0.48 [0.32, 0.74]**
Women age (in years)
<20		1		1
20–34		0.87 [0.62, 1.21]		0.81 [0.52, 1.26]
≥35		0.88 [0.60, 1.30]		0.78 [0.45, 1.35]
Preceding birth interval (in months)
< 24		1		1
24–47		1.25 [0.99, 1.58]		1.26 [0.88, 1.80]
≥ 48		1.52 [1.18, 1.95]**		1.51 [1.03, 2.20]*
Maternal education
No education		1		1
Primary		1.52 [1.28, 1.80]**		1.47 [1.16, 1.87]**
Secondary and higher		3.94 [2.82, 5.51]**		3.44[2.19, 5.42]**
Husband education
No education		1		1
Primary		1.07 [0.90, 1.27]		1.06 [0.81, 1.40]
Secondary and above		1.41 [1.13, 1.76]**		1.36 [0.99, 1.86]
Number of ANC visit
None		1		1
1–3		4.04 [3.31, 4.94]**		3.88 [2.77, 5.43]**
≥ 4		7.15 [5.85, 8.73]**		6.53 [4.69, 9.10]**
Wealth status
Poorest		1		1
Poorer		1.67 [1.31, 2.13]**		1.59 [1.10,2.30]*
Middle		1.54 [1.20, 1.97]**		1.44 [0.99, 2.08]
Richer		1.64 [1.26, 2.13]**		1.46 [0.99, 2.16]
Richest		5.20 [3.82, 7.09]**		2.44 [1.54, 3.87]**
Sex of household head
Female		1.06 [0.86, 1.30]		1.01 [0.74, 1.38]
Male		1		1
Media exposure
No		1		1
Yes		1.12 [0.95, 1.32]		1.00 [0.78, 1.29]
Covered by health insurance
No		1		1
Yes		1.58 [1.12, 2.24]**		1.59 [0.93, 2.74]
Multiple gestation
Single		1		1
Twin		2.45 [1.49, 4.01]**		2.35 [0.96, 5.76]
Community level factors
Distance to health facility
Not a big concern			1	1
Big concern			1.34[1.10, 1.63]**	1.20 [0.96, 1.51]
Residence
Rural			1	1
Urban			15.26[9.89, 23.56]	4.70 [2.76, 8.01]**
Community media exposure
Lower			1	1
Higher			2.91 [2.11, 4.01]	2.01 [1.44, 2.79]**
Region
Pastoral			1	1
Semi pastoral			3.17 [1.97, 5.10]	1.37 [0.80, 2.33]
City administration			12.91[7.6321.83]	3.13 [1.77, 5.55]**
Agrarian			3.33 [2.26, 4.92]	1.39 [0.89, 2.17]
Constant	0.49[0.40,0.60]	0.08 [0.05, 0.12]	0.04 [0.03, 0.06]	0.04 [0.02, 0.08]

Note;

* = p-value<0.05,

** = p-value<0.01.

Discussion

The most effective intervention to prevent maternal mortality from the major causes of maternal death, such as bleeding, sepsis, eclampsia, and obstructed labor is institutional delivery [47]. The current study was aimed to investigate the individual and community level determinants, and spatial distribution of institutional delivery in Ethiopia based on the nationally representative EDHS data.

The prevalence of institutional delivery in Ethiopia was found to be 26.2% in this analysis. It was lower than the prevalence in Nepal [10] and Tanzania [1], it could be due to the differences in accessibility and availability of maternal health care services across countries as Nepal and Tanzania have comparatively better socio-economic status compared to Ethiopia. While it was higher than previous studies reported in Ethiopia [30], and Bangladesh [50]. This may be attributed to the strengthened political commitment of Ethiopia for enhancing maternal health care services availability and accessibility by establishing health extension programs, expansion of health facilities, increased qualified health professionals, and improved quality of service [48-50].

The spatial analysis found that the spatial distribution of institutional delivery across the country was substantially varied. In the Harari, southern Oromia, and most parts of the Somali regions, significant hotspot areas with a low prevalence of institutional delivery (high home delivery) were established. The possible explanation might be due to the disparity in the unavailability of maternal health services, and the inaccessibility of infrastructure such as road for transportation in the border regions of those regions [51]. Besides, these areas are more pastoral areas where individuals have no permanent residents, as a result, compared to other areas, comparatively health facilities are not open and accessible [52]. This finding suggests that public health planners and programmers should design effective public health interventions to enhance institutional delivery in these significant hotspot areas where institutional delivery was low.

In the multilevel logistic regression analysis; preceding birth interval, the number of ANC visits, wealth status, residence, community media exposure, region, and maternal education were significantly associated with institutional delivery. Among individual-level factors, maternal education was found to be a significant predictor of institutional delivery. Women who attained primary and secondary education had a higher likelihood of institutional delivery than women who didn’t attain formal education. It is consistent with previous study findings [10, 53–57]. It might be because education is the key to adapting positive behaviors like utilizing maternal health care services and educated mothers might be well informed about the benefits of institutional delivery [58]. Furthermore, maternal education could lead to the corresponding improvement in the mothers’ health-seeking behavior compared to un-educated women.

The odds of institutional delivery among women who had ANC visits during pregnancy were higher than those who didn’t have ANC visits. It was consistent with studies reported in Ethiopia [59, 60], and Bangladesh [56]. This is due to the assumption that ANC visits during pregnancy may increase the awareness of women about the risks of pregnancy and childbirth, as well as helping the mother to have an effective birth preparedness plan, which may increase the chance of their delivery at health facilities [61]. Besides, health education, counseling, and treatment services offered by the health professional during ANC visits can result in women’s behavioral changes and increased perceived benefits of seeking institutional delivery services [59].

Consistent with previous studies [10, 53, 56, 57], this study revealed that household wealth status was a significant predictor of institutional delivery. The likelihood of having institutional delivery was higher among mothers in the richest household wealth index than the poorest. These may be because better economic status may increase healthcare-seeking behavior and autonomy of healthcare decision-making as they are capable of paying the required medical and transport costs. [62]. While maternity and ambulance services are free in Ethiopia, it is still well known that drug and transportation services are still out of pocket charge, as many of the drugs are not accessible in public health facilities and there is a small number of ambulances.

In our study, multiparty was significantly associated with decreased odds of institutional delivery compared to primiparous women and this was consistent with previous study findings [10, 53]. This may be because primiparous women are afraid that they are more vulnerable to complications during childbirth and seek early maternity care services, which makes them more likely to give birth at the delivery of health facilities [63]. Also, multiparous women often choose to give birth at home for the gain of privacy and believe they will not face problems and are familiar with childbirth [64]. Furthermore, institutional delivery seeking behavior is affected by the delivery service satisfaction of the preceding pregnancies.

Among the community-level factors, women from the community with high media exposure had higher odds of institutional delivery. This was supported by prior studies [2, 43, 53, 56]. The possible reason is that health information may enhance health-seeking habits through different electronic and print media, as information about what service is available, where and when to get the services, as well as the advantages and risks of accessing specific services, can be transmitted through such media [65]. The study revealed that the place of residence was found to be a significant predictor of institutional delivery. Women living in rural areas had a higher likelihood of having institutional delivery than rural residents. It was consistent with studies in Ethiopia [66, 67], Bangladesh [53, 56], and Nepal [10]. The possible explanation could be due to urban women had better access to maternal health care services and alternative service provisions like the use of private sectors, and get access to transportation at a reasonable cost and time as compared to rural women [68]. Furthermore, urban residents are closer to information about the health benefits of institutional delivery. Besides, women in city administrations (Addis Ababa and Dire-Dawa) had higher odds of institutional delivery as compared to those from pastoral regions. The consistent result has been reported in Ethiopia [67, 69]. The possible justification is health facilities are easily accessible and highly concentrated in Addis Ababa and Dire-Dawa. But women in pastoral regions have poor access to education and are not permanent residents and because of these in these areas, there is limited availability and accessibility of maternal health services such as institutional delivery.

Strength and limitations of the study

This study had strengths. First, the study was based on weighted data to make it representativeness at national and regional levels: therefore, it can be generalized to all women who gave birth during the study period. Besides, the study was based on an advanced (appropriate) model, by taking into account the clustering effect, to get reliable standard error and estimate. Moreover, the use of GIS and SaTScan statistical tests helps to detect similar and statistically significant hotspot areas of institutional delivery and design effective public health programs. But this study was not without limitations. The SaTScan detect only circular clusters, irregularly shaped clusters were not identified. Besides, the GPS data (Latitude and Longitude) taken at enumeration area were displaced to 5 Km in urban areas and 10 Km in Rural areas for the privacy issue, this could bias our spatial result. Furthermore, the EDHS survey did not incorporate clinically confirmed data; rather, it relied on mothers or caregivers reports and might have the possibility of social desirability and recall bias (27). Furthermore, due to the cross-sectional nature of the data, the temporal relationship can’t be established.

Conclusions

Institutional delivery utilization in Ethiopia was very low. The spatial distribution of institutional delivery was significantly varied in Ethiopia. The significant hotspot areas with a low prevalence of institutional delivery (high home delivery) were detected in the Harari, south Oromia, and most parts of Somali regions. Parity, preceding birth interval, maternal education, number of ANC visits, wealth status, residence, and region were found to be significantly associated with institutional delivery. Therefore, public health interventions targeting significant hotspot areas (areas with a low prevalence of institutional delivery) is essential to enhance institutional delivery and reduce maternal and newborn mortality. Besides, governmental and non-governmental organizations should scale up maternal health programs to rural and poorest women. For future researchers, it is good to incorporate maternal and community knowledge, attitude, and behavior towards maternal health care service utilization by using a mixed approach (qualitative and quantitative studies) to have a deeper understanding of the factors that impede them to give birth at the health facility.

12 Jun 2020

PONE-D-19-33251

Individual and community-level determinants and spatial distribution of institutional delivery in Ethiopia, 2016: Spatial and multilevel analysis

PLOS ONE

Dear Mr Tesema,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

We would appreciate receiving your revised manuscript by Jun 26 2020 11:59PM. When you are ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter.

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). This letter should be uploaded as separate file and labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. This file should be uploaded as separate file and labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. This file should be uploaded as separate file and labeled 'Manuscript'.

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

We look forward to receiving your revised manuscript.

Kind regards,

Professor Khaled Khatab, Ph.D.

Academic Editor

PLOS ONE

Additional Editor Comments (if provided):

1) Some similar studies have discussed this issue in the same country before. See, for example, Mekonnen, Z.A., Lerebo, W.T., Gebrehiwot, T.G. et al. Multilevel analysis of individual and community-level factors associated with institutional delivery in Ethiopia. BMC Res Notes 8, 376 (2015). https://doi.org/10.1186/s13104-015-1343-1; Mezmur M, Navaneetham K, Letamo G, Bariagaber H (2017) Individual, household and contextual factors associated with skilled delivery care in Ethiopia: Evidence from Ethiopian demographic and health surveys. PLoS ONE 12(9): e0184688. https://doi.org/10.1371/journal.pone.0184688.

1.a) So what this study added to the current knowledge?

1.b) A further comparisons with the above studies need to be included in the discussion section.

1.c) Also, the literature review needs to be more robust and to include all the similar that tackled this issue in Ethiopia.

2) It was not mentioned why the manuscript has focused on this only or this period only? Was there any inclusion/exclusion criteria considered?

4) Neither strength nor the limitations of this study were mentioned?

5) The future work plan was not mentioned and how we plan to overcome the limitations of this study in the future.

Journal requirements:

When submitting your revision, we need you to address these additional requirements:

1.    Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at http://www.plosone.org/attachments/PLOSOne_formatting_sample_main_body.pdf and http://www.plosone.org/attachments/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. We note that Figures 1 and 4-6 in your submission contain map images which may be copyrighted. All PLOS content is published under the Creative Commons Attribution License (CC BY 4.0), which means that the manuscript, images, and Supporting Information files will be freely available online, and any third party is permitted to access, download, copy, distribute, and use these materials in any way, even commercially, with proper attribution. For these reasons, we cannot publish previously copyrighted maps or satellite images created using proprietary data, such as Google software (Google Maps, Street View, and Earth). For more information, see our copyright guidelines: http://journals.plos.org/plosone/s/licenses-and-copyright.

We require you to either (1) present written permission from the copyright holder to publish these figures specifically under the CC BY 4.0 license, or (2) remove the figures from your submission:

1.    You may seek permission from the original copyright holder of Figures 1 and 4-6 to publish the content specifically under the CC BY 4.0 license.

We recommend that you contact the original copyright holder with the Content Permission Form (http://journals.plos.org/plosone/s/file?id=7c09/content-permission-form.pdf) and the following text:

“I request permission for the open-access journal PLOS ONE to publish XXX under the Creative Commons Attribution License (CCAL) CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). Please be aware that this license allows unrestricted use and distribution, even commercially, by third parties. Please reply and provide explicit written permission to publish XXX under a CC BY license and complete the attached form.”

Please upload the completed Content Permission Form or other proof of granted permissions as an "Other" file with your submission.

In the figure caption of the copyrighted figure, please include the following text: “Reprinted from [ref] under a CC BY license, with permission from [name of publisher], original copyright [original copyright year].”

2.    If you are unable to obtain permission from the original copyright holder to publish these figures under the CC BY 4.0 license or if the copyright holder’s requirements are incompatible with the CC BY 4.0 license, please either i) remove the figure or ii) supply a replacement figure that complies with the CC BY 4.0 license. Please check copyright information on all replacement figures and update the figure caption with source information. If applicable, please specify in the figure caption text when a figure is similar but not identical to the original image and is therefore for illustrative purposes only.

The following resources for replacing copyrighted map figures may be helpful:

USGS National Map Viewer (public domain): http://viewer.nationalmap.gov/viewer/

The Gateway to Astronaut Photography of Earth (public domain): http://eol.jsc.nasa.gov/sseop/clickmap/

Maps at the CIA (public domain): https://www.cia.gov/library/publications/the-world-factbook/index.html and https://www.cia.gov/library/publications/cia-maps-publications/index.html

NASA Earth Observatory (public domain): http://earthobservatory.nasa.gov/

Landsat: http://landsat.visibleearth.nasa.gov/

USGS EROS (Earth Resources Observatory and Science (EROS) Center) (public domain): http://eros.usgs.gov/#

Natural Earth (public domain): http://www.naturalearthdata.com/

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: No

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: No

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Summary:

The research focus though not new, it has helped to fill some of the identified gaps in the existing direction of the study.

The authors have demonstrated good understanding of the subject and had written the manuscript professionally and intelligently. Efforts put into the writing the manuscripts and the analysis have shown clearly that the authors are skilful and experienced researchers. The methodology adopted in the data analysis fits appropriately to the nature of the data and in fulfilment of the aim/objectives of the study. However, clearer research question(s) need to be established. The authors would need to consult experienced language editor to enhance the flow of the manuscripts.

Some minor errors/concerns need to be addressed. I have chronicle some the concerns I feel the authors need to address to enhance the quality of the paper and to meet the required standard of the intended journal (see the attached document).

Reviewer #2: Overview

This paper presents an interesting approach to understanding variation in institutional delivery in Ethiopia using both multilevel modeling and GIS. While the modeling strategy is interesting, the paper would be improved if the authors spent more time discussing the two approaches (the multilevel modeling and the GIS approach) in tandem and how they build on one another. More synthesis on the use of the added value and joint examination of the results would be interesting and would strengthen the paper.

General comments:

• The authors will need to carefully proof read the manuscript. In some places, it is difficult to follow because of grammatical errors. It may help to have an outside person edit the document for grammar. I’ve pulled a few examples here: page 1 line 63: “For example, the estimated 130,000 maternal deaths happened in 2017 in those countries [3].” Page 1 lines 71-72, “plentiful numbers of women in developing countries give birth at homes.” Page 2 line 90: “So far, Ethiopia has made a lot to curb maternal...”

• Page 1 line 67: I think it would be more accurate to say that strengthening facility delivery or skilled attendance would “reduce” not “alleviate” the burden of preventable maternal death.

• Page 2, line 96: is the low facility attendance due to limited health facilities, bad infrastructure, etc?

Methods

• There is a lot of background information provided in the “study design, setting and period” section. The authors may want to consider integrating some of the background about Ethiopia into the background section, as it is somewhat difficult to follow in the methods section.

• The authors should indicate that they are using DHS data and describe it from the beginning of the methods section when they first discuss the survey.

• Line 151: missing closed parentheses

• The authors may want to consider including the general equation that they use. It is somewhat confusing for the authors to describe the outcome variable as Yi given that multilevel models usually require a nested structure, so that individual i is usually nested in community j and so forth. Also, how is region a community level variable? Is that not a third level of the model? Or, are the authors including dummies for region as fixed effects?

• The authors should more clearly specify how they defined the level 2 community variables. As it stands, it is unclear how the authors measured community level media exposure. It would also be helpful for them to cite other papers that have used EA in DHS as a proxy for community.

• The calculation of the PCV in binary multilevel models is not straightforward due to the level-1 variation of the binary model. Can the authors describe in full what approach they used?

• The authors may want to consider condensing some of the material and reorganizing some of the subheadings in the methods section. For example, the data collection procedure section could be combined with the description of the survey. As of now, some of the information is duplicative and the section is somewhat difficult to follow.

• What statistical program did the authors use and using what estimation technique?

Results

• Does the fact that DHS throws the spatial coordinates of GPS influence to ensure that the exact location of the clusters is not revealed influence the results in any way? In particular, page 11 (the section beginning at line 292)?

• The interpretation of the odds ratio from the null model is confusing (line 305). I assume that 7.01 refers to the intercept? The authors may want to ensure that their interpretation is correct, and if so, explain it in a way that is more clear.

• The authors state that “About 73 percent of the variability in institutional delivery was explained by the full model.” Is this the model overall? What about the different levels?

• In the methods section, the authors describe “region” as a community-level variable. Does region refer to urban/rural as indicated in the results section?

• The authors may want to discuss community level variables separately from individual level variables as it is somewhat confusing to understand the results.

Discussion

• An important piece missing from the discussion is a synthesis of the results and a discussion of why context matters. The authors do a good job of summarizing the results and then discussing them in light of other literature, but it would be interesting for the authors to focus on understanding the combined results from both analytical approaches to deepen the understanding of the role of community context in Ethiopia.

• The authors do not discuss the results of the very interesting GIS analysis in the discussion.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files to be viewed.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at figures@plos.org. Please note that Supporting Information files do not need this step.

Submitted filename: Reviewed Comments on the PLOSONE Manuscript-Latest.docx

Click here for additional data file.

31 Jul 2020

PLOS ONE

Point by point response for editors/reviewers comments

Manuscript title: Individual and community-level determinants and spatial distribution of institutional delivery in Ethiopia, 2016: Spatial and multilevel analysis

Manuscript ID: PONE-D-19-33251

Dear editor/reviewer.

Dear all,

We would like to thank you for this constructive, building, and improvable comments on this manuscript that would improve the substance and content of the manuscript. We considered each comment and reviewers on the manuscript thoroughly. Our point-by-point responses for each comment and questions are described in detail on the following pages.

Response to editors comments

1. Some similar studies have discussed this issue in the same country before. See, for example, Mekonnen, Z.A., Lerebo, W.T., Gebrehiwot, T.G. et al. Multilevel analysis of individual and community-level factors associated with institutional delivery in Ethiopia. BMC Res Notes 8, 376 (2015). https://doi.org/10.1186/s13104-015-1343-1; Mezmur M, Navaneetham K, Letamo G, Bariagaber H (2017) Individual, household and contextual factors associated with skilled delivery care in Ethiopia: Evidence from Ethiopian demographic and health surveys. PLoS ONE 12(9): e0184688. https://doi.org/10.1371/journal.pone.0184688.

a) So what this study added to the current knowledge?

Authors’ response: Thank you editor for the concerns. These two previous studies were conducted based on EDHS 2005 and 2011 data to investigate the individual and community level determinants of institutional delivery using multilevel analysis. But both studies are failed to capture the spatial distribution of institutional delivery using ArcGIS and SaTScan analysis to identify the significant hotspot areas where institutional delivery was low even if studies done on the prevalence of institutional delivery in different parts of Ethiopia revealed that the prevalence has been varied across the country. Besides, in this study, the data were not weighted even if the DHS statistician recommended using weighted data based on strata, PSU, and weighting variables to restore the representativeness as well as to get reliable standard error and estimate. Therefore, we investigated the individual and community level determinants and spatial distribution of institutional delivery in Ethiopia based on the most recent EDHS data. Thus, exploring the spatial distribution of institutional delivery is important to identify significant primary and secondary hotspot areas where institutional delivery is low, this could help to design targeted public health interventions to the identified hotspot areas to enhance health facility delivery to reduce preventable maternal and newborn mortality. Furthermore, hence the data we used were weighted the estimates are reliable. (see the Backroung section, line 95-106, page 6)

b) A further comparisons with the above studies need to be included in the discussion section.

Authors’ response: Thank you editor for the comments. We had incorporated it. (see the revised document)

c) Also, the literature review needs to be more robust and to include all the similar that tackled this issue in Ethiopia

Authors’ response: Thank you, editor. We had incorporated previous study findings conducted on institutional delivery by extensively searching literature. (See the revised manuscript)

2. It was not mentioned why the manuscript has focused on this only or this period only? Was there any inclusion/exclusion criteria considered?

Authors’ response: Thank you editor for the concerns. We have used the EDHS 2016 data for this study since this Survey is the most recent in Ethiopia. For this study, we excluded the respondents where the outcome variable (place of delivery) was missed and for the spatial analysis, we excluded those women in the Enumeration Areas (EAs) with zero latitudes and longitude.

3. Neither strength nor the limitations of this study were mentioned?

Authors’ response: Thank you editor, we have included strength and limitations of the study in the revised manuscript. (See the Strengths and limitation section, line 427-438, page 21)

4. The future work plan was not mentioned and how we plan to overcome the limitations of this study in the future.

Authors’ response: Thank you editor, we have incorporated in the revised manuscript. (See the revised manuscript, Conclusion section, line 448-452, page 22)

5. We note that Figures 1 and 4-6 in your submission contain map images which may be copyrighted. All PLOS content is published under the Creative Commons Attribution License (CC BY 4.0), which means that the manuscript, images, and Supporting Information files will be freely available online, and any third party is permitted to access, download, copy, distribute, and use these materials in any way, even commercially, with proper attribution. For these reasons, we cannot publish previously copyrighted maps or satellite images created using proprietary data, such as Google software (Google Maps, Street View, and Earth). For more information, see our copyright guidelines: http://journals.plos.org/plosone/s/licenses-and-copyright.

Authors’ response: Thank you editor for the concern. The map is not copyrighted rather we have done using ArcGIS and SaTScan software based on the shapefile of Ethiopia received from Ethiopian Central Statistical Agency (CSA) by explaining the purpose of the study and GPS data (longitude and latitude) from measure DHS program by explaining the objective of the study through online requesting and allow us to access the shapefile and GPS data. Now we cite the source of the shapefile since it is needed to explore the spatial distribution of institutional delivery. Therefore, the maps presented in our study are not copyrighted rather it was our spatial analysis result.

Response to reviewers comments

Reviewer#1

1. The research focus though not new, it has helped to fill some of the identified gaps in the existing direction of the study.

The authors have demonstrated good understanding of the subject and had written the manuscript professionally and intelligently. Efforts put into the writing the manuscripts and the analysis have shown clearly that the authors are skilful and experienced researchers. The methodology adopted in the data analysis fits appropriately to the nature of the data and in fulfilment of the aim/objectives of the study. However, clearer research question(s) need to be established. The authors would need to consult experienced language editor to enhance the flow of the manuscripts.

Authors’ response: Thank you, reviewer, for the comments. The research questions in our study were, 1) to identify the individual and community level factors associated with institutional delivery, 2) to explore the spatial distribution of institutional delivery, and to identify significant hotspot areas where institutional delivery utilization is low to design targeted public health interventions. We had extensively edited and modified the entire document. (see the revised manuscript)

2. Some minor errors/concerns need to be addressed. I have chronicle some the concerns I feel the authors need to address to enhance the quality of the paper and to meet the required standard of the intended journal

Authors’ response: We thank the reviewer for your great effort for the betterment of our work. We accepted your comments and modified accordingly. (see the revised manuscript)

Reviewer# 2

1. this paper presents an interesting approach to understanding variation in institutional delivery in Ethiopia using both multilevel modeling and GIS. While the modeling strategy is interesting, the paper would be improved if the authors spent more time discussing the two approaches (the multilevel modeling and the GIS approach) in tandem and how they build on one another. More synthesis on the use of the added value and joint examination of the results would be interesting and would strengthen the paper.

Authors’ response: Thank you, reviewer. We synthesize both the use of spatial and multilevel modeling and for further we cite references. (See the revised manuscript)

2. The authors will need to carefully proof read the manuscript. In some places, it is difficult to follow because of grammatical errors. It may help to have an outside person edit the document for grammar. I’ve pulled a few examples here: page 1 line 63: “For example, the estimated 130,000 maternal deaths happened in 2017 in those countries [3].” Page 1 lines 71-72, “plentiful numbers of women in developing countries give birth at homes.” Page 2 line 90: “So far, Ethiopia has made a lot to curb maternal...” • Page 1 line 67: I think it would be more accurate to say that strengthening facility delivery or skilled attendance would “reduce” not “alleviate” the burden of preventable maternal death. • Page 2, line 96: is the low facility attendance due to limited health facilities, bad infrastructure, etc?

Authors’ response: Thank you, reviewer, for the comments. we had extensively edited the whole document with the help of language experts at university of Gondar. (See the revised manuscript)

3. There is a lot of background information provided in the “study design, setting and period” section. The authors may want to consider integrating some of the background about Ethiopia into the background section, as it is somewhat difficult to follow in the methods section. The authors should indicate that they are using DHS data and describe it from the beginning of the methods section when they first discuss the survey. Line 151: missing closed parentheses

Authors’ response: Thank you, reviewer. we have modified it. (see the revised manuscript)

4. The authors may want to consider including the general equation that they use. It is somewhat confusing for the authors to describe the outcome variable as Yi given that multilevel models usually require a nested structure, so that individual i is usually nested in community j and so forth. Also, how is region a community level variable? Is that not a third level of the model? Or, are the authors including dummies for region as fixed effects?

Authors’ response: Thank you, reviewer. we incorporated the general equation of the multilevel model we fitted. We use Enumeration area/clusters as a random effect as level two since women within the selected enumeration area are more correlated than the women from different clusters. In EDHS except for region, distance to the health facility, and residence the other variables were collected at the individual level. We tried to consider the region as the third level of the model but the result was the same as two level multilevel model and besides, we didn't get any variable collected at the region level. We categorized the region into four groups as Pastoral, Semi pastoral, City administration, and Agrarian in the multilevel analysis and considered as a community level variable at level 2.

5. The authors should more clearly specify how they defined the level 2 community variables. As it stands, it is unclear how the authors measured community level media exposure. It would also be helpful for them to cite other papers that have used EA in DHS as a proxy for community.

Authors’ response: Thank you, reviewer. We included how we generated community-level variables. In EDHS data except for region, residence, and distance to health facility these variables were collected at individual levels. The EDHS data has hierarchical nature means women are nested within-cluster therefore we want to assess the individual and cluster level variables that affect institutional delivery. Then to assess whether community media exposure has a significant effect on institutional delivery we aggregated media exposure collected at the individual level to community level/cluster level. Then we categorized as high community media exposure and low community media exposure based on the national median value since it was not normally distributed. (See the revised manuscript)

6. The calculation of the PCV in binary multilevel models is not straightforward due to the level-1 variation of the binary model. Can the authors describe in full what approach they used?

Authors’ response: thank you, reviewer, for the comments. The proportional change in variance (PCV) measures the total variation attributed by individual-level factors and area-level factors in the multilevel model. PCV is used to show the total variability explained by the final model (model with individual and community level variable simultaneously) relative to the null model, it is like the coefficient of determination (R2) in the linear regression model. It is interpreted as the total variability of institutional delivery explained by the final model. As you know in the multilevel binary logistic regression model the level-1 variation is constant (π2/3) unlike the linear regression since there are no residuals. Therefore, we calculated by substracted the variance of institutional delivery in the final model from the variance in the null mode divided by the variance in the null model.

7. The authors may want to consider condensing some of the material and reorganizing some of the subheadings in the methods section. For example, the data collection procedure section could be combined with the description of the survey. As of now, some of the information is duplicative and the section is somewhat difficult to follow.

Authors’ response: Thank you, reviewer, we organized to combined related subheadings together. (See the revised manuscript)

8. What statistical program did the authors use and using what estimation technique?

Authors’ response: Thank you reviewer. For analysis we used melogit STATA command (melogit institutional_delivery i.ANC i.region i.parity i.residence i.communty_media_exposure i.maternal_education i.materna_age i.husband_education i.birth_interval i.wealth_status [pw=wgt] || v001:, or), using maximum likelihood estimation technique. To calculate the MOR and ICC we install the xtmrho stata command and run after running the full model.

9. Results

• Does the fact that DHS throws the spatial coordinates of GPS influence to ensure that the exact location of the clusters is not revealed influence the results in any way? In particular, page 11 (the section beginning at line 292)?

Authors’ response: thank you, reviewer, for the comments. In EDHS as GPS data (latitude and longitude) taken in the Enumeration area level were displaced up to 5km in Urban area, and up to 10 km in rural areas because of privacy issues, therefore, this could bias our findings and we acknowledge in the limitation sections of the study. (See the revised manuscript, limitation section, line 434-436, page 21)

10. The interpretation of the odds ratio from the null model is confusing (line 305). I assume that 7.01 refers to the intercept? The authors may want to ensure that their interpretation is correct, and if so, explain it in a way that is more clear.

Authors' response: Thank you, reviewer. This MOR result in the null model. MOR quantifies the variation between clusters (the second level variations) by comparing two women from two randomly chosen, different clusters. The result was MOR= 7.01, 95% CI: 6.02, 9.17 it indicates that if we randomly select two women from two different clusters. A woman from the cluster with a high likelihood of institutional had 7.01 times higher odds of having institutional delivery compared with women form clusters with lower institutional delivery. Therefore, this finding was not the intercept rather it is another method of assessing heterogeneity across clusters like ICC.

11. The authors state that “About 73 percent of the variability in institutional delivery was explained by the full model.” Is this the model overall? What about the different levels?

Authors’ response: Thank you, reviewer, for the concern. PCV is used to measure the total variation in institutional delivery explained by the final model (a model with both individual and community-level factors simultaneously) in relative to the null model ( Model without independent variables).

Therefore, about 73% of the total variation in institutional delivery was explained by the overall model.

12. In the methods section, the authors describe “region” as a community-level variable. Does region refer to urban/rural as indicated in the results section?

Authors' response: Thank you, reviewer, for the comments. In Ethiopia, there are 9 regions and 2 city administrations. In our study, we categorized the region into 4 groups. 1, pastoralist region (Benishangul, Somali, Gambella, and Afar), Semipastorlaist (Oromia, SNNPR), Agrarian (Amhara and Tigray) and City administration (Addis Ababa, Dire Dawa, and Harari) based on literature. Rural and Urban were for the variable residence.

13. The authors may want to discuss community level variables separately from individual level variables as it is somewhat confusing to understand the results.

Authors' response: Thank you, reviewer. we have discussed it separately, first, we discussed the individual variables significantly associated with institutional delivery. (See the Discussion section, line 347-425, page 17-21)

14. Discussion

• An important piece missing from the discussion is a synthesis of the results and a discussion of why context matters. The authors do a good job of summarizing the results and then discussing them in light of other literature, but it would be interesting for the authors to focus on understanding the combined results from both analytical approaches to deepen the understanding of the role of community context in Ethiopia.

Authors' response: Thank you, reviewer. we have written out in the revised manuscript.

15. • The authors do not discuss the results of the very interesting GIS analysis in the discussion.

Authors' response: Thank you, reviewer. we have discussed it. (See the Discussion section, line 361-370, page 18)

Submitted filename: Point by point response.docx

Click here for additional data file.

28 Sep 2020

PONE-D-19-33251R1

Individual and community-level determinants, and spatial distribution of institutional delivery in Ethiopia, 2016: Spatial and multilevel analysis

PLOS ONE

Dear Mr Tesema,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by 7th of November 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

We look forward to receiving your revised manuscript.

Kind regards,

Prof Khaled Khatab, Ph.D.

Academic Editor

PLOS ONE

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #2: (No Response)

Reviewer #3: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #2: Yes

Reviewer #3: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #2: I Don't Know

Reviewer #3: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #2: Yes

Reviewer #3: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #2: No

Reviewer #3: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #2: The authors have done a good job in addressing many of my previous comments, but there still remain several issues for the authors to address.

General

There are still some grammatical issues that the authors should address prior to publication. In particular, there are still grammatical mistakes, misplaced words, and incorrect/inconsistent use of tense.

Methods

1. The organization of the methods section is still somewhat confusing. The multilevel equation specified by the authors appears under the outcome variable description when it seems like the discussion of the model would be more appropriate under the description of multilevel analysis that appears later in the methods section.

2. The authors describe their calculation of the PCV as being reliant on calculation the total variance. The authors should include how they calculate the total variance, given that the VPC is more difficult to calculate in binary-response models due to the different scale of the level 1 variance because of use of the link function.

3. Furthermore, the authors may want to discuss the implications of their choice of estimation technique (maximum likelihood) on the estimates of variance, given that maximum likelihood estimation tends to underestimate the variance at higher levels in multilevel binary models, and often MCMC estimation is used instead when the variance parameters are of substantive interest. The authors may want to consider adding this in the limitations section.

Results

4. The authors calculate the total variance explained by the addition of the covariates in the model, but they do not decompose the variance to examine the variance explained at the community level. This would be helpful, and again goes back to my previous comment asking the authors about how they calculated the total variation in the model.

Discussion

1. The discussion still requires some grammatical editing and certain parts are difficult to follow. In particular, paragraph 2.

Reviewer #3: Authors have addressed concerned earlier. Having reviewed the current version of the manuscript I can confidently confirmed that the article has now significantly met the set conditions for it to be published in this journal. Thus, I hereby recommend that the manuscript be accepted for publication. The only area I wish the authors should try and adjust is in the equations in lines 135-142. Using equation editor could help to fine-tune the notations/subscripts of the respective equation's parameters.

For record however, I commend the authors for the efforts channeled into writing the manuscript and I believe their scholastic contributions into the research community would receive wider acceptability.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #2: No

Reviewer #3: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

8 Oct 2020

Point by point response for editors/reviewers comments

Manuscript title: Individual and community-level determinants, and spatial distribution of institutional delivery in Ethiopia, 2016: Spatial and multilevel analysis

Manuscript ID: PONE-D-19-33251R1

Dear editor/reviewer.

Dear all,

We would like to thank you for this constructive, building, and improvable comments on this manuscript that would improve the substance and content of the manuscript. We considered each comment and clarification questions of editors and reviewers on the manuscript thoroughly. Our point-by-point responses for each comment and questions are described in detail on the following pages. Further, the details of changes were shown by track changes in the supplementary document attached.

Response to reviewers comment

Reviewer # 2

1. General

There are still some grammatical issues that the authors should address prior to publication. In particular, there are still grammatical mistakes, misplaced words, and incorrect/inconsistent use of tense.

Authors’ response: Thank you reviewer for the comments. We extensively edited the whole document for the grammatical error, typical errors, and sentence structure with the help of language experts at UOG. (See the revised manuscript)

2. Methods

2.1. The organization of the methods section is still somewhat confusing. The multilevel equation specified by the authors appears under the outcome variable description when it seems like the discussion of the model would be more appropriate under the description of multilevel analysis that appears later in the methods section.

Authors’ response: Thank you reviewer for raising the important comments. We accept your comments and we placed the equation in the multilevel analysis section. (See the Method section, line 191- 202, page 10)

2.2. . The authors describe their calculation of the PCV as being reliant on calculation the total variance. The authors should include how they calculate the total variance, given that the VPC is more difficult to calculate in binary-response models due to the different scale of the level 1 variance because of use of the link function.

Authors’ response: Thank you reviewer for the comments. As you stated, it is difficult to calculate VPC/ICC is more difficult since unlike the linear regression model the individual-level variance in logistic regression is assumed to follow the standard logistic distribution with mean 0 and variance of π2/3 (3.29). So, the individual level variance is constant that is π2/3, we add the cluster level variance with π2/3 to get the total variance in the null model and in the final model. Then we calculate PCV by using the formula (variance in the null model – variance in the final model)/variance in the null model.

2.3. Furthermore, the authors may want to discuss the implications of their choice of estimation technique (maximum likelihood) on the estimates of variance, given that maximum likelihood estimation tends to underestimate the variance at higher levels in multilevel binary models, and often MCMC estimation is used instead when the variance parameters are of substantive interest. The authors may want to consider adding this in the limitations section

Authors’ response: Thank you reviewer for the concerns. We used the maximum likelihood estimation technique to estimate the estimates of variance since we used the classical multilevel binary logistic regression and the sample is large. We plan to use MCMC to estimate the variance using MCMC but to do this it needs the posterior distribution (bayesian approach). And now we have checked by simulating the data and estimate the variance using MCMC but the value is the same as the MLE findings since the sample is adequate and the prevalence is 26.2.

2.4. Results, The authors calculate the total variance explained by the addition of the covariates in the model, but they do not decompose the variance to examine the variance explained at the community level. This would be helpful, and again goes back to my previous comment asking the authors about how they calculated the total variation in the model.

Authors’ response: Thank you reviewer for concerns. PCV is commonly done in multilevel analysis to examine by how much the full model explains the variations of institutional delivery as compared to the null model. And we calculate the PCV for three models (model1 model2, and model 3) in relative to the null model. Logically we need that the PCV value to be highest in the final model as the number of covariates incorporated is increased, based on these the final model was 73% indicates that the final model explains the variation in institutional delivery by 73%, it is like the coefficient of determination in the linear regression model. So, this 73% is explained by the mutual effects of the individual as well as community-level variables, and it is difficult to partition how much is explained by the individual and how much is by the community level variables. Just PCV is supportive, beyond that we have used deviance, MOR, ICC, and LR-test to check whether the model is improved or not. Just to know how much is attributable to the community level and how much is by community level we can get in model 1 and model 2 as we presented in the table.

3. Discussion

1. The discussion still requires some grammatical editing and certain parts are difficult to follow. In particular, paragraph 2.

Authors’ response: Thank you the reviewer for the comments. We extensively modify it. (See the revised manuscript)

Reviewer #3

1. The authors have addressed concerned earlier. Having reviewed the current version of the manuscript I can confidently confirmed that the article has now significantly met the set conditions for it to be published in this journal.

Thus, I hereby recommend that the manuscript be accepted for publication. The only area I wish the authors should try and adjust is in the equations in lines 135-142. Using equation editor could help to fine-tune the notations/subscripts of the respective equation's parameters

Authors’ response: Thank you reviewer for the comments. We accept the comment and we put in equation form. (See the revised manuscript)

Submitted filename: Point by point response.docx

Click here for additional data file.

30 Oct 2020

Individual and community-level determinants, and spatial distribution of institutional delivery in Ethiopia, 2016: Spatial and multilevel analysis

PONE-D-19-33251R2

Dear Mr Tesema,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter, and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double-check that your user information is up-to-date. If you have any billing-related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Professor Khaled Khatab, Ph.D.

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

4 Nov 2020

PONE-D-19-33251R2

Individual and community-level determinants, and spatial distribution of institutional delivery in Ethiopia, 2016: Spatial and multilevel analysis

Dear Dr. Tesema:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Professor Khaled Khatab

Academic Editor

PLOS ONE