Individual and community level determinants of short birth interval in Ethiopia: A multilevel analysis

Desalegn Markos Shifti; Catherine Chojenta; Elizabeth G Holliday; Deborah Loxton

doi:10.1371/journal.pone.0227798

. 2020 Jan 14;15(1):e0227798. doi: 10.1371/journal.pone.0227798

Individual and community level determinants of short birth interval in Ethiopia: A multilevel analysis

Desalegn Markos Shifti ^1,^2,^*, Catherine Chojenta ², Elizabeth G Holliday ³, Deborah Loxton ²

Editor: Kannan Navaneetham⁴

PMCID: PMC6959604 PMID: 31935262

Abstract

Background

The World Health Organization recommends a minimum of 33 months between two consecutive live births to reduce the risk of adverse maternal and child health outcomes. However, determinants of short birth interval have not been well understood in Ethiopia.

Objective

The aim of this study was to assess individual- and community-level determinants of short birth interval among women in Ethiopia.

Methods

A detailed analysis of the 2016 Ethiopian Demographic and Health Survey data was performed. A total of 8,448 women were included in the analysis. A two-level multilevel logistic regression analysis was used to identify associated individual- and community-level factors and estimate between-community variance.

Results

At the individual-level, women aged between 20 and 24 years at first marriage (AOR = 1.37; 95% CI: 1.18–1.60), women aged between 25 and 29 years at first marriage (AOR = 1.65; 95% CI: 1.20–2.25), having a husband who attended higher education (AOR = 1.32; 95% CI: 1.01–1.73), being unemployed (AOR = 1.16; 95% CI: 1.03–1.31), having an unemployed husband (AOR = 1.23; 95% CI: 1.04–1.45), being in the poorest wealth quintile (AOR = 1.82; 95% CI: 1.39–2.39), being in the poorer wealth quintile (AOR = 1.58; 95% CI: 1.21–2.06), being in the middle wealth quintile (AOR = 1.61; 95% CI: 1.24–2.10), being in the richer wealth quintile (AOR = 1.54; 95% CI: 1.19–2.00), increased total number of children born before the index child (AOR = 1.07; 95% CI: 1.03–1.10) and death of the preceding child (AOR = 1.97; 95% CI: 1.59–2.45) were associated with increased odds of short birth interval. At the community-level, living in a pastoralist region (AOR = 2.01; 95% CI: 1.68–2.39), being a city dweller (AOR = 1.75; 95% CI: 1.38–2.22), high community-level female illiteracy (AOR = 1.23; 95% CI: 1.05–1.45) and increased distance to health facilities (AOR = 1.32; 95% CI: 1.11–1.56) were associated with higher odds of experiencing short birth interval. Random effects showed significant variation in short birth interval between communities.

Conclusion

Determinants of short birth interval are varied and complex. Multifaceted intervention approaches supported by policy initiatives are required to prevent short birth interval.

Introduction

A preceding birth interval is defined as the amount of time between the birth of the child under study (index child) and the immediately preceding birth [1, 2]. It has received increased attention in demography and public health research because of its implications for fertility and maternal and child health [3].

To reduce the risk of adverse maternal and child health outcomes, the most recent World Health Organization (WHO) recommendation for a healthy pregnancy interval is at least two years (24 months) or a birth-to-birth interval of 33 months, assuming nine months gestation [4]. Short birth interval is associated with increased risk of adverse birth outcomes, such as preterm birth [5, 6], low birth weight [5, 6], small size for gestational age [5], congenital anomalies [7, 8], autism [9], and infant mortality [5, 10]. It is also associated with a higher risk of maternal morbidities, including preeclampsia, high blood pressure, and premature rupture of membranes [11].

Worldwide, birth interval practices differ widely [12]. Particularly, women in developing countries often have shorter birth intervals than they would personally prefer [13]. For instance, in sub-Saharan Africa, mothers would prefer a median birth interval of 38.9 months, which is more than 6.2 months longer than their actual median birth interval [14]. Similarly, evidence from Demographic Health Survey data (DHS) of 52 different countries showed that more than two-thirds of births occurred at durations of less than 30 months since the previous birth [15].

A variety of factors influences a woman’s birth interval, including her personal characteristics and the health status of her previous child [12]. Traditional practices, particularly breastfeeding, postpartum abstinence and cultural norms also affect birth spacing [12]. Other factors affecting birth interval are rooted in social and cultural norms, including reproductive histories and behaviors of individual women, utilization of reproductive health services and other background factors [16].

In Ethiopia, the prevalence of short birth interval ranges from 21.7% (national estimate) [17] to 57.6% (district level estimate) [18]. In addition, there were inconsistencies among studies in identifying determinants of short birth interval, such as wealth index [18–20], maternal age [19, 21], and the number of children [19, 21]. Moreover, previous studies [18–21] mainly focused on individual-level determinants of short birth interval. However, some of the individual-level factors such as marital status, maternal age at marriage, maternal age at birth, death of preceding child, sex of preceding child, exposure to media, and distance to a health facility have not been well studied in Ethiopia. Furthermore, community-level factors such as regions where women reside, community-level female literacy, and community-level poverty have not yet been investigated. However, human population has a complex community structure and its characteristics can also affect an individual’s health behavior, including birth interval practices. Hence, it is important to assess the effects of individual- and community-level factors together to understand the complex nature of short birth interval, which can then inform national-level policy direction as well as individual- and community-based interventions, such as mass campaigns and dissemination of health education messages.

Taking into account changing demographics, such as high fertility [17] in Ethiopia, the increasing proportion of short birth intervals [18, 20] with a marked variation among regions ranging from 10.6% in Amhara region to 45.9% in Somali region [17], our objective was to identify individual- and community-level determinants of short birth interval in Ethiopia. The findings of this study will provide pertinent information for policy makers, program planners and other stakeholders, which in turn may help to design and implement appropriate interventions at different levels to prevent short birth interval and improve women’s and children’s health in general.

Methods

Study area and setting

An in-depth secondary data analysis was conducted using the 2016 Ethiopian Demographic and Health Survey (EDHS) data. The EDHS is carried out every five years to provide health and health-related indicators at the national and regional levels in Ethiopia. The 2016 EDHS is the fourth national survey conducted in all parts of Ethiopia (in nine regional states [Tigray, Afar, Amhara, Oromia, Somali, Benishangul-Gumuz, Southern Nations, Nationalities and Peoples’ Region (SNNPR), Gambella, and Harari] and two administrative cities [Addis Ababa and Dire Dawa]). Administratively, regions in Ethiopia are divided into Zones, and Zones into administrative units called Woredas. Each Woreda is further subdivided into the lowest administrative unit, called Kebeles. During the 2007 census, each Kebele was subdivided into census enumeration areas (EAs), which were convenient for the implementation of the census.

Study design and sampling

The 2016 EDHS was cross-sectional by design. It employed two-stage stratified cluster sampling based on a sampling frame of the 2007 Population and Housing Censuses in which EAs were the sampling units for the first stage and households for the second stage (Fig 1). The detailed sampling procedure is presented in the full EDHS report [17]. The current study included individual-level data for 8,448 women who had at least two live births during the five years preceding 2016 as well as community characteristics of 640 clusters. Women who had never been married were excluded from the analysis, since women who have multiple births out of wedlock are unlikely to plan their births in the same way as married women. Data collection took place from 18 January 2016 to 27 June 2016.

Outcome variable

The outcome variable for this study was short birth interval. Short birth interval was defined as an interval less than 33 months between two successive live births (33 months = 24 months of birth to conception period + 9 months duration of pregnancy). Preceding birth intervals exceeding 33 months were defined as non-short birth interval, which is consistent with the WHO recommendations [4].

Exposure variables

The socio-ecological model provides the theoretical foundation for selecting the exposure variables. The model asserts that human health behavior, the experience of short birth interval in the current study, is influenced at multiple levels within the social and physical environment, such as the individual-, household-, community- and societal/policy-level factors [22–24]. The framework also assumes that interactions between individuals and their environment are reciprocal, implying that individuals are influenced by their environment and the environment is influenced by individuals [25]. For instance, an individual woman’s literacy status is affected by community-level female literacy and vice versa. These variables, in turn, may independently affect women’s experience of short birth interval. An increased understanding of the impact of individual- and community-level factors on the risk of short birth interval can inform policy makers and health programmers by providing insight into how to prevent the risk of short birth interval among individuals as well as groups of women in a community. Assessing the impact of societal/policy-level factors on the risk of short birth interval, however, is beyond the scope of the current study.

Details of the definition and coding of the individual- and community-level factors included in the current study are presented online (see S1 Table). Variables were selected based on reviewed literature [2, 18–21, 26–31]. Individual factors included maternal age at first marriage, maternal age at birth of the preceding child, marital status, polygyny status, maternal educational level, husbands’/partners’ education level, maternal occupation, husbands’/partners’ occupation, wealth quintile, sex of preceding child, total number of children born before the index child, survival status of preceding child, exposure to media (television, radio and newspaper; each separately) and distance to health facility. Separated, divorced, and widowed women were considered in the analysis. Since data regarding birth interval of women were collected from their most recent childbirth in the five years preceding the survey, women may have become widowed or divorced or separated before the time of the survey but after they already became eligible to provide birth interval information. Community-level factors included place of residence, contextual regions, community-level female illiteracy, community-level poverty, and community-level distance to a health facility. The 11 regions of Ethiopia are essentially delineated for administrative purposes, and in this study they were categorized into three contextual regions: pastoralist, agrarian, and city (which were defined on the basis of the cultural and socio-economic backgrounds of their populations) [32]. Except for place of residence and geographic region, the EDHS did not capture variables that describe community characteristics. Therefore, additional community-level characteristics were constructed by aggregating the individual-level characteristics within their clusters. Aggregate values were classified as high or low based on the distribution in each community. The cut-point was defined as the mean if the aggregate variable was normally distributed and the median if not. For example, community-level poverty refers to the proportion of poor households (two lowest wealth quintiles) in the community (cut-off at median proportion). Community-level factors describe groups of populations living in similar settings.

The assumption of independence of observation has been taken as a basis to determine whether variables were analysed at the individual or community-level. If the observations at the individual-level are independent, variables were treated as individual-level factors. Whereas if the observations were clustered into higher levels of units or if several women had shared features (e.g., place of residence and contextual region, that could have the same effect on short birth interval among women in that locality), then the variables were analysed at the community-level.

Data analysis

Descriptive analysis

Due to the non-proportional allocation of the sample to different regions as well as urban and rural areas, and the possible differences in response rates, as per the recommendation of the EDHS, proportions and frequencies were estimated after applying sample weights to the data to adjust for disproportionate sampling and non-responses. A detailed explanation of the weighting procedure can be found in the EDHS report [17]. To examine the crude association between the individual- and community-level factors separately with short birth interval, p-values were calculated using Pearson’s chi-squared test. A p-value of less than 0.05 was set for statistical significance of an association.

Multivariable multilevel analysis

The 2016 EDHS employed multistage cluster sampling and data are hierarchical (i.e., mothers are nested within households, and households are nested within clusters). As a result of the sampling approach, mothers within the same cluster may be more similar to each other than mothers in the rest of the country. To account for this clustering, two-stage multivariable multilevel logistic regression analysis was used to estimate the effects of individual- and community-level determinants on short birth interval and to estimate the between-cluster variability in the odds of short birth interval. For the multilevel logistic regression analysis, the Stata syntax xtmelogit was employed [33]. Four models were constructed in this analysis. Model I (the empty model) was fitted without explanatory variables to estimate random variation in the intercept and the intra-cluster correlation coefficient (ICC; i.e., to evaluate the extent of the cluster variation in short birth interval). Model II was constructed to examine the effects of individual-level characteristics, while Model III was fitted to assess the effects of community-level characteristics. Finally, model IV adjusted for the individual- and community-level variables simultaneously. The fixed effect sizes of individual- and community-level factors on short birth interval were expressed using the adjusted odds ratios (AOR) with 95% confidence interval (CI). A p-value of <0.05 was used to declare statistical significance.

Total number children born before the index child was considered as a continuous variable. This was done after checking for the linearity assumption with the log-odds of short birth interval, which is a binary response variable. Multicollinearity was also checked among the exposure variables using the variance inflation factor (VIF). If the values of VIF were lower than 10, then the collinearity problem was considered to be unlikely. The VIF for birth order was 15.80 and for the total number of children born before the index child was 15.77, which indicates the presence of collinearity. Therefore, we removed the variable birth order from the model and the VIF became less than 2 for each variable included in the model (see S2 Table).

The random effects were measured by the intra-class correlation coefficient (ICC), median odds ratio (MOR) and proportional change in variance (PCV). The ICC was calculated to evaluate whether the variation in short birth interval is primarily within or between communities [34]. This is because individuals living in the same neighbourhood (community) may be more similar to each other than individuals living in other neighbourhoods as they share a number of economic, social, and other neighbourhood characteristics that may condition similar health status [35]. The ICC for two-level binary data was calculated using the following formula [34, 36]:

I C C = \frac{V_{a}}{V_{a} + \frac{π^{2}}{3}}

Where V_a (τ²) is community-level variance and $\frac{π^{2}}{3}$ is individual-level variance (VI) equal to 3.29. It takes a value between 0 and 1. A high ICC value indicates that neighbourhoods are important in understanding individual differences in short birth interval [37]. The MOR is defined as the median value of the odds ratio between the area at highest risk and the area at the lowest risk when randomly picking out two areas and it depends directly on the area-level variance [34]. It can be calculated using the following formula:

M O R = (e x p \sqrt{2 X V_{a} X 0.6745}) \approx e x p (0.95 \sqrt{V_{A}})

Where V_a is the area-level variance, and 0.6745 is the 75^th centile of the cumulative distribution function of the normal distribution with mean 0 and variance 1. In this study, MOR shows the extent to which the individual probability of experiencing short birth interval is determined by the residential area [34]. The PCV is used to measure the total variation attributed to individual-level factors and area-level factors in the multilevel model [34, 38]. It was calculated using the following mathematical equation:

P C V = \frac{V_{n - 1} - V_{n - 2}}{V_{n - 1}}

Where V_n−1 is the neighbourhood variance in the empty model and V_n−2 is the neighbourhood variance in the subsequent model.

Model fit statistics

Akaike’s information criterion (AIC) and Schwarz’s Bayesian information criteria (BIC) were used to assess goodness of fit and inform the selection of nested models (individual- and community-level model). The AIC and BIC values were compared in successive models and the model with the lowest value was considered to be the best-fit model [33, 39]. Unlike statistical methods that employ hypothesis testing approaches such as deviance differences and log-likelihood ratio statistics, AIC and BIC penalize the deviance for a larger number of parameters (overparameterization) [39–41]. Thus, they provide more protection against overfitting the model to the data. In addition, area under receiver operating characteristics (AUC) curve was used to assess the predictive quality of each model [42].

Statistical analysis was performed using Stata version 14 statistical software (StataCorp. Stata Statistical Software: Release 14. College Station, TX: StataCorp LP. 2015).

Ethics statement

The 2016 EDHS was approved by the National Research Ethics Review Committee of Ethiopia and ICF Macro International. Permission from The DHS Program was obtained to use the 2016 EDHS data. This analysis was approved by the University of Newcastle Human Research Ethics Committee (H-2018-0332).

Results

Sociodemographic and other health related characteristics

Overall, 8,448 women at level 1 nested within 640 clusters at level 2 were included in the analysis. The proportion of short birth intervals based on individual- and contextual-level background characteristics of the study participants are shown in Table 1. The mean (standard deviation) age of women who participated in the study was 30.82 (6.1). Among women who experienced short birth interval, 79.4% were 19 years old or under at their first marriage, 76.4% were uneducated, 75.1% were not working, and 28.8% were from the poorest households.

Table 1. Weighted proportion of short birth interval by individual-level factors, EDHS 2016.

Variables	Weighted proportion		P value
Variables	Non-short birth interval	Short birth interval
Maternal age at first marriage (years)			<0.001 <0.001
≤19	3499 (83.6)	3324 (79.4)
20–24	600 (12.7)	697 (16.3)
25–29	116 (2.8)	142 (3.6)
30+	46 (0.9)	24 (0.7)
Maternal age at birth of the preceding child (years)
≤19	712 (16.7)	598 (13.8)
20–24	1383 (31.3)	1391 (32.6)
25–29	1117 (26.6)	1143 (27.9)
30–34	723 (16.9)	706 (16.4)
35+	326 (8.5)	349 (9.3)
Marital status
Separated/Divorced/Widowed	227 (4.2)	144 (2.9)	<0.001
Married	4034 (95.8)	4043 (97.1)	<0.001
Polygyny status (n = 8077)			<0.001
Yes	552 (11.4)	774 (14.0)
No	3467 (88.2)	3257 (85.4)
Don’t know	15 (0.4)	12 (0.5)
Maternal educational level			<0.001
No education	2871 (71.9)	3201 (76.4)
Primary	1019 (22.9)	779 (20.7)
Secondary	250 (3.4)	131 (1.8)
Higher	121 (1.7)	76 (1.1)
Husband’s/partner’s education level (n = 8077)			<0.001
No education	2023 (52.0)	2308 (51.6)
Primary	1341 (37.3)	1210 (40.3)
Secondary	365 (6.2)	279 (4.5)
Higher	281 (4.1)	217 (2.6)
Don’t know	24 (0.4)	29 (1.0)
Maternal occupation			<0.001
Not working	2949 (71.1)	3164 (75.1)
Working	1312 (28.9)	1023 (24.9)
Husband’s/partner’s occupation (n = 8077)			<0.001
Not working	350 (6.5)	545 (9.5)
Working	3639 (93.0)	3466 (89.8)
Don’t know	45 (0.5)	32 (0.6)
Wealth quintiles			<0.001
Poorest	1348 (22.0)	1999 (28.8)
Poorer	759 (22.1)	719 (24.8)
Middle	658 (21.0)	545 (21.5)
Richer	589 (19.1)	466 (16.6)
Richest	907 (15.8)	458 (8.3)
Sex of preceding child			0.33
Male	2162 (50.9)	2169 (52.0)
Female	2099 (49.1)	2018 (48.0)
Survival status of preceding child			<0.001
Yes	4099 (96.0)	3855 (92.7)
No	162 (4.0)	332 (7.3)
Distance to a health facility			<0.001
Big problem	2279 (58.0)	2530 (67.3)
Not big problem	1982 (42.0)	1657 (32.7)
Watched television			<0.001
Yes	946 (18.9)	538 (12.9)
No	3315 (81.1)	3649 (87.1)
Listened to radio			<0.001
Yes	1086 (26.6)	734 (23.0)
No	3179 (73.4)	3453 (77.0)
Read newspapers			<0.001
Yes	296 (5.6)	143 (3.8)
No	3965 (94.4)	4044 (96.2)

Open in a new tab

As depicted in Table 2, the vast majority of women with short birth interval resided in rural areas (94.0%) and agrarian regions (87.8%).

Table 2. Weighted proportion of short birth interval by community-level factors, EDHS 2016.

Variables	Weighted proportion		P value
Variables	Non-short birth interval	Short birth interval	P value
Place of residence			<0.001
Urban	818 (11.3)	468 (6.0)
Rural	3443 (88.7)	3719 (94.0)
Contextual regions			<0.001
Agrarian	2310 (93.0)	1629 (87.8)
Pastoralist	1338 (4.4)	2065 (10.7)
City dweller	613 (2.6)	493 (1.5)
Distance to a health facility as a big problem			<0.001
High	2233 (59.0)	2630 (72.5)
Low	2028 (41.0)	1557 (27.5)
Community-level poverty			<0.001
High	1902 (34.7)	2564 (42.2)
Low	2359 (65.3)	1623 (57.8)
Community-level female literacy			<0.001
High	1861 (48.4)	2357 (50.2)
Low	2400 (51.6)	1830 (49.8)

Open in a new tab

The overall prevalence of short birth interval in Ethiopia was 45.8% (95% CI: 42.91–48.62).

Determinants of short birth interval among women

Multilevel analysis (fixed effect analysis)

Table 3 presents the results of multivariable multilevel logistic regression analysis (fixed effects). In the full model adjusted for individual- and community-level factors, maternal age at first marriage, maternal educational level, maternal occupation, husbands’/partners’ occupation, wealth quintile, total number of children born before the index child, death of the preceding child, contextual regions, community-level illiteracy, community-level poverty, and community-level increased distance to health facility were each associated with short birth interval among women.

Table 3. Individual- and community-level determinants of short birth interval in Ethiopia using multilevel logistic regression analysis, EDHS 2016.

Variables	Model 2^a AOR (95% CI)	Model 3^b AOR (95% CI)	Model 4^c AOR (95% CI)
Maternal age at first marriage (reference:≤19 years)
20–24	1.37 (1.18, 1.60)^***		1.37 (1.18, 1.60)^***
25–29	1.63 (1.19, 2.23)^**		1.65 (1.20, 2.25)^**
30+	0.69 (0.38, 1.24)		0.68 (0.38, 1.21)
Maternal age at birth of the preceding child (reference: ≤19)
20–24	1.07 (0.91, 1.25)		1.09 (0.93, 1.27)
25–29	0.89 (0.74, 1.08)		0.94 (0.78, 1.13)
30–34	0.78 (0.62, 0.98)^**		0.83 (0.66, 1.05)
35+	0.82 (0.61, 1.10)		0.89 (0.66, 1.20)
Polygyny status (reference: not polygyny)
Polygyny	1.09 (0.95, 1.26)		1.01 (0.88, 1.17)
Don’t know	1.50 (0.62, 3.62)		1.62 (0.67, 3.90)
Maternal educational level (reference: No education)
Primary	0.99 (0.86, 1.13)		1.03 (0.89, 1.19)
Secondary	0.81 (0.60, 1.09)		0.83 (0.61, 1.12)
Higher	1.22 (0.80, 1.87)		1.24 (0.81, 1.90)
Husband’s/partner’s education level (reference: No education)
Primary	1.07 (0.95, 1.21)		1.12 (0.99, 1.2)
Secondary	1.19 (0.96, 1.47)		1.16 (0.94, 1.44)
Higher	1.36 (1.04, 1.78)^**		1.32 (1.01, 1.73)^**
Don’t know	1.38 (0.74, 2.55)		1.41 (0.76, 2.62)
Maternal occupation (reference: working)
Not working	1.17 (1.04, 1.32)^**		1.16 (1.03, 1.31)^**
Husband’s/partner’s occupation (reference: working)
Not working	1.30 (1.09, 1.53)^**		1.23 (1.04, 1.45)^**
Don’t know	0.55 (0.33, 0.92)^**		0.54 (0.32, 0.90)^**
Wealth quintile (reference: richest)
Poorest	2.60 (2.08, 3.26)^***		1.82 (1.39, 2.39)^***
Poorer	1.86 (1.47, 2.34)^***		1.58 (1.21, 2.06)^**
Middle	1.77 (1.40, 2.24)^***		1.61 (1.24, 2.10)^***
Richer	1.67 (1.33, 2.11)^***		1.54 (1.19, 2.00)^**
Sex of preceding child (reference: male)
Female	0.96 (0.87, 1.05)		0.96 (0.87, 1.06)
Total number of children born before the index child	1.07 (1.04, 1.11)^***		1.07 (1.03, 1.10)^***
Survival of preceding child (reference: yes)
No	2.01 (1.62, 2.50)^***		1.97 (1.59, 2.45)^***
Watched television (reference: yes)
No	1.09 (0.91, 1.30)		1.06 (0.88, 1.27)
Listened to radio (reference: yes)
No	1.16 (1.01, 2.34)^**		1.14 (0.99, 1.31)
Read newspapers (reference: yes)
No	1.15 (0.88, 1.51)		1.10 (0.84, 1.44)
Distance to a health facility to get medical help (reference: not big problem)
Big problem	1.02 (0.91, 1.15)		0.94 (0.83, 1.06)
Community-level factors
Place of residence (reference: urban)
Rural		1.72 (1.36, 2.17)^***	1.23 (0.93, 1.63)
Contextual regions (reference: Agrarian)
Pastoralist		2.24 (1.88, 2.67)^***	2.01 (1.68, 2.39)^***
City dweller		1.68 (1.32, 2.13)^***	1.75 (1.38, 2.22)^***
Community-level female illiteracy (reference: low)
High		1.29 (1.10, 1.48)^**	1.23 (1.05, 1.45)^**
Community-level poverty (reference: low)
High		1.23 (1.02, 1.48)^**	1.13 (0.93, 1.37)
Community-level distance to the health facility as a big problem (reference: low)
High		1.38 (1.17, 1.62)^***	1.32 (1.11, 1.56)^**

Open in a new tab

***P value <0.001

** P value <0.05

AOR = Adjusted Odds Ratio; CI = Confidence Interval

Model 2^a is adjusted for individual-level factors

Model 3^b is adjusted for community-level factors

Model 4^c is the final model adjusted for individual- and community-level factors

N.B. Model 1 (Empty model) was fitted without determinant variables and is not included in this table (but is in Table 4).

Individual-level factors

The odds of short birth interval among women aged between 20 and 24 years at first marriage were 37% higher (AOR = 1.37; 95% CI: 1.18–1.60) compared with women whose age at their first marriage was 19 years or under. Likewise, the odds of short birth interval among women aged between 25 and 29 years at first marriage were 65% higher (AOR = 1.65; 95% CI: 1.20–2.25) compared with women whose age at first marriage was 19 years or under. The odds of short birth interval among not working women was 16% higher (AOR = 1.16; 95% CI: 1.03–1.31) than for working women. The odds of short birth interval were 23% higher (AOR = 1.23; 95% CI: 1.04–1.45) among women whose husband/partner was not working compared with women who had a working husband/partner. On the other hand, the odds of short birth interval were 46% lower (AOR = 0.54; 95% CI: 0.32–0.90) among women who did not know their partner’s working status than women whose husband/partner was working. The odds of short birth interval among women from the poorest households were 1.82 times higher (AOR = 1.82; 95% CI: 1.39–2.39), from poorer households 1.58 times higher (AOR = 1.58; 95% CI: 1.21–2.06), from middle income households 1.61 times higher (AOR = 1.61; 95% CI: 1.24–2.10) and from the richer households 1.54 times higher (AOR = 1.54; 95% CI: 1.19–2.00) than women from the richest households. As the number of total children born before the index child increased by one, the odds of short birth interval increased by 7% (AOR = 1.07; 95% CI: 1.03–1.10). The odds of short birth interval among women whose preceding child did not survive were 1.97 times higher (AOR = 1.97; 95% CI: 1.59–2.45) compared with women whose preceding child survived.

Community-level factors

Women residing in a pastoralist region had a two-fold higher odds (AOR = 2.01; 95% CI: 1.68–2.39) and women residing in cities had a 1.75-fold higher odds (AOR = 1.75; 95% CI: 1.38–2.22) of experiencing short birth interval as compared with women residing in an agrarian region. Similarly, women residing in communities with a high proportion of illiterate women had 23% higher odds (AOR = 1.23; 95% CI: 1.05–1.45) of experiencing short birth interval when compared with women residing in communities with a low proportion of illiterate women. Women who lived in a community with increased distance to health facility had a 32% higher odds (AOR = 1.32; 95% CI: 1.11–1.56) of experiencing short birth interval compared with women who lived in a community with close to a health facility.

Multilevel analysis (random-effects analysis)

Table 4 presents quantities based on random effects. The prevalence rate of short birth interval varied across communities (τ² = 0.75, p = <0.001). The empty model revealed that 18.6% of the total variance in the odds of short birth interval was accounted for by between-cluster variation of characteristics (ICC = 0.186). The between-cluster variability declined over successive models, from 18.6% in the empty model to 11.8% in the individual-level only model, 11.2% in the community-level only model, and 9.5% in the final (combined) model. The proportional change in variance indicated that the addition of predictors to the empty model explained an increased proportion of variation in short birth interval. Similar to ICC values, the combined model showed a higher PCV, i.e., 53.3% of variance in short birth interval could be explained by the combined factors at the individual- and community-levels. Furthermore, the MOR confirmed that short birth interval was attributed to community-level factors. The MOR for short birth interval was 2.28 in the empty model, which indicated the presence of variation between communities (clustering) in short birth interval since MOR was two times higher than the reference (MOR = 1). The unexplained community variation in short birth interval decreased to a MOR of 1.75 when all factors were added to the empty model. This implies that even though individual- and community-level factors were considered, the effect of clustering is still statistically significant in the full model.

Table 4. Results from random intercept model (measure of variation) for short birth interval at cluster level using multilevel logistic regression analysis.

Random effects (Measure of variation for short birth interval)	Model 1^a	Model 2^b	Model 3^c	Model 4^d
Community-level variance (SE)	0.75 (0.08)	0.44 (0.06)	0.41 (0.05)	0.35 (0.05)
P value	<0.001	<0.001	<0.001	<0.001
ICC (%)	18.6	11.8	11.2	9.5
Explained variance (PCV) (%)	Reference	41.3	45.3	53.3
MOR (95% CI)	2.28 (2.10, 2.47)	1.88 (1.74, 2.03)	1.85 (1.71, 1.98)	1.75 (1.62, 1.89)
Model fit statistics
AIC	11185.68	10503.02	10975.21	10414.58
BIC	11199.76	10726.92	11031.54	10680.46
AUC (95% CI)	0.75 (0.74, 0.76)	0.74 (0.73, 0.76)	0.74 (0.73, 0.75)	0.74 (0.73, 0.75)

Open in a new tab

SE = Standard Error; DIC = Deviance Information Criterion; ICC = Intra-Class Correlation; PCV = Percentage Change in Variance; MOR = Median Odds Ratio; CI = Confidence Interval; AIC = Akaike’s Information Criterion; BIC = Schwarz’s Bayesian Information Criteria; AUC = Areas under the receiver operating characteristic

Model 1^a is the null model, a baseline model without any determinant variable

Model 2^b is adjusted for individual-level factors

Model 3^c is adjusted for community-level factors

Model 4^d is the final model adjusted for individual- and community-level factors

Model fit statistics

As shown in Table 4 (model fit statistics), the values of AIC and BIC showed subsequent reduction which indicates each model represents a significant improvement over the previous model and it points to the goodness of fit of the final model built in the analysis. Therefore, the final (combined) model that incorporated individual- and community-level factors was selected for predicting the occurrence of short birth interval among women. The predictive utility of the final model was fair, with an AUC of 0.74.

Discussion

The purpose of this paper was to understand the individual- and community-level determinants of short birth interval in a developing country, Ethiopia. Our findings reveal that close to half of women (45.7%) in Ethiopia had a short birth interval. The difference in the prevalence of short birth intervals between the current study and the one reported in the 2016 EDHS (21.7%) [17] is due to the different definitions used for short birth interval. The EDHS considered a birth interval of less than 24 months a short birth interval, whereas our study defined it as less than 33 months, which is in accordance with the WHO recommendations [4]. In the current study, short birth interval births were associated with different individual- and community-level characteristics. At the individual-level, maternal age at first marriage, maternal occupation, husband’s/partner’s education, husband’s/partner’s occupation, household wealth index, the total number of children born before the index child, and survival status of the preceding child were associated with short birth interval. At the community-level, women living in cities and pastoralist regions, in community with a high level of female illiteracy and women who lived far away from a health facility were more likely to experience short birth intervals.

In this study, it was found that the odds of short birth interval among women aged between 20 and 24 years and 25 and 29 years when they first married were higher compared to women who had married at 19 years or under. A previous study conducted in Asian countries found that marriage age influences birth spacing [43]. In contrast, a previous study conducted in Ethiopia reported a non-significant association [19] This inconsistency could be due to the difference in the study setup. Unlike the current study, which was based on nationally representative data, the previous study was restricted to a rural area and single district. Literature reports that age at marriage can influence both the quantum and tempo of fertility [44].

Household wealth index was one of the independent predictors of short birth interval. Women from the poorest, poorer, middle and richer households were more likely to experience short birth interval compared with women from the richest household. This finding is consistent with the existing literature [19]. While it is likely that women with fewer resources have less access to health information and contraception services and thus may not adhere to the scientifically recommended birth intervals, more detailed research is needed to understand this finding.

The study found that death of the preceding child increased the likelihood of short birth interval. It is possible that death of the preceding child ends breastfeeding, resulting in women returning to menses and resuming ovulation sooner; some couples may unintentionally have their next child within a shorter birth interval. On the other hand, some couples may intentionally try to have another child as soon as possible. This finding is consistent with the findings of previous studies performed in Ethiopia [20, 27] and Bangladesh [45].

Our study also found that characteristics of clusters in which women live have an effect on women’s experience of short birth interval, independent of their individual characteristics. For instance, residing in a different contextual region such as a pastoralist area and a city affects the odds of women experiencing a short birth interval. Living in a pastoralist region and a city was associated with higher odds of short birth interval. Short birth interval in a pastoralist community could be attributable to their nomadic lifestyle, which hinders access to health services infrastructure, including contraception and media regarding health information. On the other hand, the likelihood of short birth interval among women residing in cities may be explained by the relatively busy lifestyle and expensive cost of living in cities, which may force women to reduce their childbearing period and engage in income-generating activities to earn a living and support their family.

Unlike the effect of individual-level educational status on short birth interval, women living in communities with a high proportion of female illiteracy had 23% higher odds of short birth interval relative to women living in communities with a low proportion of female illiteracy. This result may imply a negative effect of community-level maternal illiteracy on shaping birth interval practice. The effect of community-level female illiteracy may operate through a wide range of mechanisms, such as poor access to information, health illiteracy, family planning unawareness, and contraceptive non-utilization.

The longer the distance to a health facility from the community where the women lived, the higher the odds that women experienced short birth interval. The existing literature [46–48] shows clear connections between distance to a health facility and family planning utilization which is vital to ensuring an optimum birth interval. However, it is important to note that the distance to a health facility in this study was assessed based on women’s reports, which may over- or underestimate the actual distance to the health facilities.

The results of our study should be interpreted in light of the following limitations. First, since the information on birth interval was recorded retrospectively, it might be prone to recall bias. Second, the analyses were conducted using data collected in a cross-sectional survey, which prevents causal inferences. Hence, there is a need for prospective research.

Despite the abovementioned limitations, the study has numerous strengths. First, the dataset used in this analysis was the most recent and nationally representative survey that covered all regions and administrative cities of the country. Second, unlike previously conducted small-scale studies that mainly focused on individual-level factors, the current study considered community-level factors. Additionally, this study applied multilevel modeling to accommodate the hierarchical nature of the EDHS data.

Conclusion

The study showed that nearly half of women experience short birth interval in Ethiopia. Not only individual characteristics of women, but also community-level factors determine women’s experience of short birth interval. Individual-level characteristics, including maternal age at first marriage between 20 and 24, and between 25 and 29, maternal occupation, husband’s/partner’s education, husband’s/partner’s occupation, household wealth index, the total number of children born before the index child, and death of the preceding child were found to be independent predictors of short birth interval. Similarly, community-level variables, such as contextual region, community-level female illiteracy, and community-level increased distance to health facility were found to be significantly associated with short birth interval. The random-effects analysis result also revealed that community-level random effects remained significant after controlling for the effects of both individual- and community-level variables, indicating that the experience of short birth interval among women depends on community contexts.

Thus, the prevention of short birth interval requires a multifaceted intervention approach. Family planning programs should pay particular attention to women with the identified risk factors for short birth interval. An approach that may initiate discussion on optimum birth interval is community mobilization through women’s groups (e.g., Women’s Development Army). The effectiveness of the Women’s Development Army in improving maternal and child health services’ outcomes has been documented elsewhere [49, 50]. Furthermore, mass campaigns and health information dissemination, particularly in city and pastoralist communities, regarding optimum birth interval is required.

Implications for policy practice and further research

The government should set a clear target to reduce the magnitude (45.7%) of short birth interval in a defined period. Birth interval messages as per the WHO recommendation should be incorporated into antenatal care, postnatal care, and immunization services to improve women’s awareness. Additionally, integrating the WHO recommended birth interval message under the health extension program, which is a community-based health care delivery program [51], could help address women and their husbands/partners in each household. Considering the finding that unemployment affects birth interval, the Ethiopian government should develop initiatives to improve the working status of women and their husbands/partners. Moreover, planners and policymakers should make further efforts to alleviate poverty at the household level. Since community-level illiteracy showed negative externality on women’s experience of short birth interval, the Ministry of Health should work in collaboration with the Ministry of Education to improve the literacy status of women in the community. In addition, cultural beliefs related to birth interval practice should be investigated. Policy makers should focus on improving access to health facilities to improve family planning service utilization and reduce short birth interval. Longitudinal data would enable better tracking of birth interval practice in the country through reducing recall bias and providing data to estimate the causal effect of exposure variables on short birth interval.

Supporting information

S1 Table. Description of variables used in the analysis.

(DOCX)

Click here for additional data file.^{(33.8KB, docx)}

S2 Table. The results of multicollinearity analysis.

(DOCX)

Click here for additional data file.^{(29.9KB, docx)}

Acknowledgments

We are grateful to The DHS Program for allowing us to use the EDHS data for this study.

Data Availability

All data are available from the DHS Program database (https://dhsprogram.com/data/available-datasets.cfm). Any researcher can request for access to the data without any restriction.

Funding Statement

The authors received no specific funding for this work.

References

1.Rutstein SO. Effects of preceding birth intervals on neonatal, infant and under‐five years mortality and nutritional status in developing countries: evidence from the demographic and health surveys. International Journal of Gynecology & Obstetrics. 2005;89(S1). [DOI] [PubMed] [Google Scholar]
2.Fallahzadeh H, Farajpour Z, Emam Z. Duration and determinants of birth interval in Yazd, Iran: a population study. Iranian journal of reproductive medicine. 2013;11(5):379 [PMC free article] [PubMed] [Google Scholar]
3.Yoder PS, Joe L.Lugalla P., Richard F. Sambaiga. Determinants of the Duration of Birth Intervals in Tanzania: Regional Contrasts and Temporal Trends DHS Qualitative Research Studies No 19 Calverton, Maryland, USA: ICF International; 2013. [Google Scholar]
4.World Health Organization. Report of a WHO Technical Consultation on Birth Spacing. Geneva, Switzerland 13–15 June 2005.
5.Grisaru-Granovsky S, Gordon E-S, Haklai Z, Samueloff A, Schimmel MM. Effect of interpregnancy interval on adverse perinatal outcomes—a national study. Contraception. 2009;80(6):512–8. 10.1016/j.contraception.2009.06.006 [DOI] [PubMed] [Google Scholar]
6.Adam I, Ismail MH, Nasr AM, Prins MH, Smits LJ. Low birth weight, preterm birth and short interpregnancy interval in Sudan. The Journal of Maternal-Fetal & Neonatal Medicine. 2009;22(11):1068–71. [DOI] [PubMed] [Google Scholar]
7.Chen I, Jhangri GS, Chandra S. Relationship between interpregnancy interval and congenital anomalies. American journal of obstetrics and gynecology. 2014;210(6):564 e1-8. 10.1016/j.ajog.2014.02.002 [DOI] [PubMed] [Google Scholar]
8.Kwon S, Lazo-Escalante M, Villaran M, Li C. Relationship between interpregnancy interval and birth defects in Washington State. Journal of Perinatology. 2012;32(1):45 10.1038/jp.2011.49 [DOI] [PubMed] [Google Scholar]
9.Cheslack-Postava K, Liu K, Bearman PS. Closely spaced pregnancies are associated with increased odds of autism in California sibling births. Pediatrics. 2011;127:246–53. 10.1542/peds.2010-2371 [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Hussaini KS, Ritenour D, Coonrod DV. Interpregnancy intervals and the risk for infant mortality: a case control study of Arizona infants 2003–2007. Maternal and child health journal. 2013;17(4):646–53. 10.1007/s10995-012-1041-8 [DOI] [PubMed] [Google Scholar]
11.DaVanzo J, Razzaque A, Rahman M, Hale L, Ahmed K, Khan MA, et al. The effects of birth spacing on infant and child mortality, pregnancy outcomes, and maternal morbidity and mortality in Matlab, Bangladesh. Technical Consultation and Review of the Scientific Evidence for Birth Spacing. 2004. [Google Scholar]
12.Setty-Venugopal V, Upadhyay UD. Birth spacing: three to five saves lives. Population Reports Series L: Issues in World Health. 2002(13):1–23. [PubMed] [Google Scholar]
13.Abdel-Tawab N, Lūzā SF, Zaki A. Helping Egyptian women achieve optimal birth spacing intervals through fostering linkages between family planning and maternal/child health services: Citeseer; 2008. [Google Scholar]
14.Rutstein SO. Trends in Birth Spacing DHS Comparative Reports No 28 Calverton, Maryland, USA: ICF Macro; 2011. [Google Scholar]
15.Rutstein SO. Further evidence of the effects of preceding birth intervals on neonatal infant and under-five-years mortality and nutritional status in developing countries: Evidence from the Demographic and Health Surveys. Macro International Inc; 2008. [DOI] [PubMed] [Google Scholar]
16.RamaRao S, Townsend J, Askew I. Correlates of inter-birth intervals: implications of optimal birth spacing strategies in Mozambique. 2006. [Google Scholar]
17.Central Statistical Agency (CSA) [Ethiopia] and ICF. Ethiopia Demographic and Health Survey 2016. Addis Ababa, Ethiopia, and Rockville, Maryland, USA: CSA and ICF. 2016.
18.Yohannes S, Wondafrash M, Abera M, Girma E. Duration and determinants of birth interval among women of child bearing age in Southern Ethiopia. BMC pregnancy and childbirth. 2011;11(1):38. [DOI] [PMC free article] [PubMed] [Google Scholar]
19.Hailu D, Gulte T. Determinants of Short Interbirth Interval among Reproductive Age Mothers in Arba Minch District, Ethiopia. International journal of reproductive medicine. 2016;2016. [DOI] [PMC free article] [PubMed] [Google Scholar]
20.Tsegaye D, Shuremu M, Bidira K. Practice of child spacing and its associated factors among women of child bearing age (15 to 49 years) in Illubabor zone, South West Ethiopia. International Journal of Nursing and Midwifery. 2017;9(7):102–8. [Google Scholar]
21.Begna Z, Assegid S, Kassahun W, Gerbaba M. Determinants of inter birth interval among married women living in rural pastoral communities of southern Ethiopia: a case control study. BMC pregnancy and childbirth. 2013;13(1):116. [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Sallis JF, Owen N, Fisher E. Ecological models of health behavior. Health behavior: Theory, research, and practice. 2015;5:43–64. [Google Scholar]
23.McLaren L, Hawe P. Ecological perspectives in health research. Journal of Epidemiology & Community Health. 2005;59(1):6–14. [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Violence Prevention Alliance. The ecological Framework. 2018 [Available from: http://www.who.int/violenceprevention/approach/ecol.
25.Salihu HM, Wilson RE, King LM, Marty PJ, Whiteman VE. Socio-ecological model as a framework for overcoming barriers and challenges in randomized control trials in minority and underserved communities. International Journal of MCH and AIDS. 2015;3(1):85 [PMC free article] [PubMed] [Google Scholar]
26.De Jonge HC, Azad K, Seward N, Kuddus A, Shaha S, Beard J, et al. Determinants and consequences of short birth interval in rural Bangladesh: a cross-sectional study. BMC pregnancy and childbirth. 2014;14(1):427. [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Tessema GA, Zeleke BM, Ayele TA. Birth interval and its predictors among married women in Dabat District, Northwest Ethiopia: A retrospective follow up study. African journal of reproductive health. 2013;17(2):39–45. [PubMed] [Google Scholar]
28.Bassey G, Nyengidiki TK, Dambo ND. Determinants of interpregnancy interval among parturient in Port Harcourt, Nigeria. Sahel Medical Journal. 2016;19(4):180. [Google Scholar]
29.Chirwa TF, Mantempa JN, Kinziunga FL, Kandala JD, Kandala N-B. An exploratory spatial analysis of geographical inequalities of birth intervals among young women in the Democratic Republic of Congo (DRC): a cross-sectional study. BMC pregnancy and childbirth. 2014;14(1):271. [DOI] [PMC free article] [PubMed] [Google Scholar]
30.Exavery A, Mrema S, Shamte A, Bietsch K, Mosha D, Mbaruku G, et al. Levels and correlates of non-adherence to WHO recommended inter-birth intervals in Rufiji, Tanzania. BMC pregnancy and childbirth. 2012;12(1):152. [DOI] [PMC free article] [PubMed] [Google Scholar]
31.Karkee R, Lee AH. Birth spacing of pregnant women in Nepal: a community-based study. Frontiers in Public Health. 2016;4:205 10.3389/fpubh.2016.00205 [DOI] [PMC free article] [PubMed] [Google Scholar]
32.FMoH. Health System Special Support. 2015 [Available from: http://www.moh.gov.et/ejcc/en/health-system-special-support.
33.Goldstein H. Multilevel statistical models 4th ed: UK, University of Bristol; 2011. [Google Scholar]
34.Merlo J, Chaix B, Ohlsson H, Beckman A, Johnell K, Hjerpe P, et al. A brief conceptual tutorial of multilevel analysis in social epidemiology: using measures of clustering in multilevel logistic regression to investigate contextual phenomena. J Epidemiol Community Health. 2006;60(4):290–7. 10.1136/jech.2004.029454 [DOI] [PMC free article] [PubMed] [Google Scholar]
35.Merlo J. Multilevel analytical approaches in social epidemiology: Measures of health variation compared with traditional measures of association. Journal of Epidemiology and Community Health. 2003; 57(8):550–1. 10.1136/jech.57.8.550 [DOI] [PMC free article] [PubMed] [Google Scholar]
36.Raykov T, Marcoulides GA. Intraclass correlation coefficients in hierarchical design studies with discrete response variables: A note on a direct interval estimation procedure. Educational and psychological measurement. 2015;75(6):1063–70. 10.1177/0013164414564052 [DOI] [PMC free article] [PubMed] [Google Scholar]
37.Merlo J, Chaix B, Yang M, Lynch J, Råstam L. A brief conceptual tutorial of multilevel analysis in social epidemiology: linking the statistical concept of clustering to the idea of contextual phenomenon. Journal of Epidemiology & Community Health. 2005;59(6):443–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
38.Merlo J, Yang M, Chaix B, Lynch J, Råstam L. A brief conceptual tutorial on multilevel analysis in social epidemiology: investigating contextual phenomena in different groups of people. Journal of Epidemiology & Community Health. 2005;59(9):729–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
39.Vrieze SI. Model selection and psychological theory: a discussion of the differences between the Akaike information criterion (AIC) and the Bayesian information criterion (BIC). Psychological methods. 2012;17(2):228 10.1037/a0027127 [DOI] [PMC free article] [PubMed] [Google Scholar]
40.Gagné P, Dayton CM. Best regression model using information criteria. Journal of Modern Applied Statistical Methods. 2002;1(2):57. [Google Scholar]
41.Akaike H. A new look at the statistical model identification Selected Papers of Hirotugu Akaike: Springer; 1974. p. 215–22. [Google Scholar]
42.Carter JV, Pan J, Rai SN, Galandiuk S. ROC-ing along: Evaluation and interpretation of receiver operating characteristic curves. Surgery. 2016;159(6):1638–45. 10.1016/j.surg.2015.12.029 [DOI] [PubMed] [Google Scholar]
43.MacQuarrie KLD. Marriage and Fertility Dynamics: The Influence of Marriage Age on the Timing of First Birth and Birth Spacing DHS Analytical Studies No 56 Rockville, Maryland, USA: ICF International; 2016. [Google Scholar]
44.Padmadas SS, Hutter I, Willekens F. Compression of women's reproductive spans in Andhra Pradesh, India. International family planning perspectives. 2004:12–9. 10.1363/3001204 [DOI] [PubMed] [Google Scholar]
45.van Soest A, Saha UR. Relationships between infant mortality, birth spacing and fertility in Matlab, Bangladesh. PloS one. 2018;13(4):e0195940 10.1371/journal.pone.0195940 [DOI] [PMC free article] [PubMed] [Google Scholar]
46.Wang W, Mallick L. Understanding the relationship between family planning method choices and modern contraceptive use: an analysis of geographically linked population and health facilities data in Haiti. BMJ Global Health. 2019;4(Suppl 5):e000765 10.1136/bmjgh-2018-000765 [DOI] [PMC free article] [PubMed] [Google Scholar]
47.Shiferaw S, Spigt M, Seme A, Amogne A, Skrøvseth S, Desta S, et al. Does proximity of women to facilities with better choice of contraceptives affect their contraceptive utilization in rural Ethiopia? PloS one. 2017;12(11):e0187311 10.1371/journal.pone.0187311 [DOI] [PMC free article] [PubMed] [Google Scholar]
48.Silumbwe A, Nkole T, Munakampe MN, Milford C, Cordero JP, Kriel Y, et al. Community and health systems barriers and enablers to family planning and contraceptive services provision and use in Kabwe District, Zambia. BMC health services research. 2018;18(1):390 10.1186/s12913-018-3136-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
49.Yitbarek K, Abraham G, Morankar S. Contribution of women’s development army to maternal and child health in Ethiopia: a systematic review of evidence. BMJ open. 2019;9(5):e025937 10.1136/bmjopen-2018-025937 [DOI] [PMC free article] [PubMed] [Google Scholar]
50.Damtew ZA, Karim AM, Chekagn CT, Zemichael NF, Yihun B, Willey BA, et al. Correlates of the Women’s Development Army strategy implementation strength with household reproductive, maternal, newborn and child healthcare practices: a cross-sectional study in four regions of Ethiopia. BMC pregnancy and childbirth. 2018;18(1):373. [DOI] [PMC free article] [PubMed] [Google Scholar]
51.Caglia J, Kearns A, Langer A. Health Extension Workers in Ethiopia: Delivering Community-based Antenatal and Postnatal Care. Boston, MA: Maternal Health Task Force; 2014. [Google Scholar]

PLoS One. doi: 10.1371/journal.pone.0227798.r001

Decision Letter 0

Kannan Navaneetham

21 Sep 2019

PONE-D-19-21556

Individual and community level determinants of short birth interval in Ethiopia: a multilevel analysis

PLOS ONE

Dear Mr. Shifti,

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.

The manuscript has been reviewed by two reviewers and their comments are appended below. The reviewers have raised number of concerns including the lack of theory for conceptualization of the framework of the study, sample inclusion criteria for the analysis and measurement of certain variables among others. I believe that the comments would be helpful to revise the manuscript.

We would appreciate receiving your revised manuscript by Nov 05 2019 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,

Kannan Navaneetham

Academic Editor

PLOS ONE

Journal Requirements:

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

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.journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and http://www.journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

[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: Yes

**********

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

Reviewer #1: Yes

Reviewer #2: Yes

**********

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: The paper by author(s) presents an exhaustive analysis of statistical tools for the analysis of individual and community level determinants of short birth interval in Ethiopia.

This work may be accepted, it is recommended to go through a minor review.

i. Methods: Study area and setting: An in-depth secondary data analysis was conducted using the 2016 Ethiopian Demographic and Health Survey (EDHS) data. and Study design and sampling: the study used data from the 2016 Ethiopian Demographic and Health Survey (EDHS). This may be presents similar meaning.

ii. Why authors used Akaike’s Information Criterion (AIC), Schwarz’s Bayesian information criteria (BIC) and Log-likelihood tests to assess goodness of fit and inform the selection of nested models (individual and community level model)? Why not R2 or adjusted R2 or Mallows Cp or mean sum square error (MSE) or others tools? If authors used any references I think it will be more reliable.

iii. I think authors used fully/mutually adjusted Odds Ratio. If authors describe shortly about this it will be more reliable.

Reviewer #2: Comments on “Individual and community level determinants of short birth interval in Ethiopia: a multilevel analysis” (ID: PONE-D-19-21556)

This paper covers a relevant question: what are the factors associated with short intervals? There is a large body of work examining the effects of short intervals on various outcomes, but we have much less research on the determinants of those intervals in the first place. The paper takes a novel and informative approach to addressing this question by implementing a multilevel mixed effects logit model. I feel that the paper is not ready for publication, as there are a number of issues that need to be addressed.

General comments:

1) The most important overarching weakness of the paper is the lack of any framework to motivate the study and to interpret the results. The authors need to spell out a theoretical argument for how birth intervals are determined. What is the decision-making process? What kinds of factors affect those decisions? How do the major elements included in the analysis (e.g. SES, biological, community, and others) ultimately determine the durations of birth intervals? This will help the authors to interpret their results in some kind of a structured way that can guide the reader. As it is now, the reader is bombarded with many coefficients of varying magnitudes and signs with only brief post hoc arguments justifying them. It would be much more helpful to have an understanding of why we should care about the variables chosen for the model, why the effects are predictable or surprising, and what kind of policy recommendations are called for.

2) The analysis should be restricted to only women with a partner. These make up the great majority of multiparous women (95% in your sample) and these are the individuals which can be realistically targeted with policy interventions. This is because women who have multiple births out of wedlock in a culturally conservative society are unlikely to be planning these births in the same way (if at all). Furthermore, the marital status category (unpartnered) must necessarily be collinear with categories of other variables indicating that the woman is not married. For example, the category for husband’s education indicating ‘not partnered’.

3) The continuous variables (maternal age at marriage) should probably be operationalized in a non-linear way. This can be done either through polynomial transformations (e.g. age at marriage ^2 or higher) or through categorization. Categories are probably a better choice for the presentation of the results. I recommend that the authors examine the data to see if this is appropriate.

4) Does maternal age at birth refer to her age at the birth of the index child or the previous child? If it is her age at the birth of the index child, this variable does not make sense to include in the model. This is because her age at birth is determined by the length of the interval (i.e. your outcome variable). If it is her age at the birth of the previous child, you need to spell this out in the text. Likewise, children ever born should refer to the number of children born prior to the birth of the index child (i.e. not including the index child). It was not clear from the paper or supporting documentation.

5) If I have understood the results correctly, it seems like community-level factors actually have very little influence on birth interval length once individual-level characteristics are controlled for. The ICC for the full model with individual and community factors was only 2.9%. In the conclusion and policy implications, however, you have interpreted community-level factors as a major concern. Is this justified given the low ICC?

6) The paper will benefit greatly from language editing for clarity and precision.

Specific comments:

Line 46: The definition of birth intervals is not quite right. First, the interval is not defined as the number of months between births, but the amount of time. Months are simply the unit that some authors express that time in. It can equivalently be expressed in years (in continuous form) or days. Second, the authors are defining a preceding birth interval here. That is fine, because that is what the paper is about. I would simply suggest that they rephrase the definition as the definition of a preceding interval instead of birth interval in general. This is because there can also be a succeeding birth interval, which would be the interval following the birth of the index child.

Line 70: How can the prevalence range from 21% to 57.6%? Does this mean across specific sub-populations or over time?

Line 223: Age at birth should be reported in more meaningful categories than the ones provided here. The age group 18-34 is enormous and captures the large majority of births. Why not present these as standard 5-year age groups?

Line 228: ‘Magnitude’ should really be ‘prevalence’.

Line 302 – 314: If you are using the same data as the EDHS, I do not think there is a need to explain why your estimates of the prevalence of short intervals differ from the report. It is clear that you use a different definition of ‘short intervals’ than the DHS program, and that this is the reason why your prevalence is higher. The comparisons to Tanzania, DR Congo, and Bangladesh are not really relevant here. I would recommend dropping this paragraph entirely, and earlier in the paper when you present the prevalence you estimate, just add a line stating that the definition you use is different from the DHS definition, and this is why your estimates are higher.

Table 1: Age at first marriage is not included in the table.

Table 3: Where is model 1? The columns start from model 2. I would also set the reference category of education as ‘No education’. Only 2.5% of women had higher education.

**********

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: Yes: Benojir Ahammed

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.

PLoS One. 2020 Jan 14;15(1):e0227798. doi: 10.1371/journal.pone.0227798.r002

Author response to Decision Letter 0

14 Nov 2019

Editorial comments: When submitting your revision, we need you to address these additional requirements. 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.journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and http://www.journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

Response: Our manuscript has been revised to meet PLOS ONE's style requirements.

Reviewer #1 comments: The paper by author(s) presents an exhaustive analysis of statistical tools for the analysis of individual and community level determinants of short birth interval in Ethiopia.

Response: Thank you.

Reviewer #1 comments: Methods: Study area and setting: An in-depth secondary data analysis was conducted using the 2016 Ethiopian Demographic and Health Survey (EDHS) data. and Study design and sampling: the study used data from the 2016 Ethiopian Demographic and Health Survey (EDHS). This may be presents similar meaning.

Response: We agree with the reviewer and we removed the repetition from the ‘study design and sampling’ section.

Reviewer #1 comments: Why authors used Akaike’s Information Criterion (AIC), Schwarz’s Bayesian information criteria (BIC) and Log-likelihood tests to assess goodness of fit and inform the selection of nested models (individual and community level model)? Why not R2 or adjusted R2 or Mallows Cp or mean sum square error (MSE) or others tools? If authors used any references, I think it will be more reliable.

Response: Thank you. Generally, R-squared (R2), Mean Squared Error (MSE) and other tools (such as Residual Standard Error (RSE), Mean Absolute Error (MAE)) are sensitive to the inclusion of additional variables in the model, even if those variables don’t explain significant variation in the outcome. For instance, including additional variables in the model will always increase the R2 and reduce the MSE. In contrast, Akaike’s Information Criterion (AIC), and Schwarz’s Bayesian information criteria (BIC) penalise the deviance for models with a larger number of parameters, and thus provide more protection against overfitting the model to the data, relative to approaches based on hypothesis testing (e.g., deviance difference, log-likelihood ratio statistic). Therefore, using a more robust metric to guide the model choice is recommended. Adjusted R2, AIC, BIC, and Mallows Cp are among the most commonly used metrics for measuring regression model quality and for comparing models. It is because of the above-mentioned reasons, we used AIC and BIC. In this revised version, we removed Log-likelihood tests. Therefore, in this revised version of the manuscript, we restricted the model fit statistics based on Akaike’s Information Criteria (AIC) and Schwarz’s Bayesian Information Criteria (BIC). We have added a brief statement explaining the robust nature of AIC and BIC under the ‘model fit statistics’ section of the manuscript (Line 230-236).

Reviewer #1 comments: I think authors used fully/mutually adjusted Odds Ratio. If authors describe shortly about this it will be more reliable.

Response: We agree and have mentioned that ‘adjusted odds ratio’ was used to express the fixed effect sizes of the individual- and community-level factors on short birth interval (Line 197)

Reviewer #2 comments: This paper covers a relevant question: what are the factors associated with short intervals? There is a large body of work examining the effects of short intervals on various outcomes, but we have much less research on the determinants of those intervals in the first place. The paper takes a novel and informative approach to addressing this question by implementing a multilevel mixed effects logit model. I feel that the paper is not ready for publication, as there are a number of issues that need to be addressed.

Response: We have carefully edited the manuscript to address the issues raised by you and other reviewers

Reviewer #2 comments: The most important overarching weakness of the paper is the lack of any framework to motivate the study and to interpret the results. The authors need to spell out a theoretical argument for how birth intervals are determined. What is the decision-making process? What kinds of factors affect those decisions? How do the major elements included in the analysis (e.g. SES, biological, community, and others) ultimately determine the durations of birth intervals? This will help the authors to interpret their results in some kind of a structured way that can guide the reader. As it is now, the reader is bombarded with many coefficients of varying magnitudes and signs with only brief post hoc arguments justifying them. It would be much more helpful to have an understanding of why we should care about the variables chosen for the model, why the effects are predictable or surprising, and what kind of policy recommendations are called for.

Response: The socio-ecological model of health behavior provides the theoretical foundation of the current study. We have now incorporated detail about this theoretical foundation into the manuscript (line 126-139). In addition, as mentioned in the method section (line 165-170), the assumption of independence of observation has been taken as a basis to determine which variables are to be analysed at the individual- and community-levels.

Reviewer #2 comments: The analysis should be restricted to only women with a partner. These make up the great majority of multiparous women (95% in your sample) and these are the individuals which can be realistically targeted with policy interventions. This is because women who have multiple births out of wedlock in a culturally conservative society are unlikely to be planning these births in the same way (if at all). Furthermore, the marital status category (unpartnered) must necessarily be collinear with categories of other variables indicating that the woman is not married. For example, the category for husband’s education indicating ‘not partnered’.

Response: The variable ‘Marital status’ in Demographic and Health Survey data describes the marital status of the women at the time of the survey not their marital status in the last five years. In our previous version of the manuscript, the ‘not married’ category included women who had never been married, as well as women who were separated, divorced, and widowed. In light of your feedback, in this revised version, ‘marital status’ is now categorized into ‘married’ and ‘separated/divorced/widowed.

With the exception of ‘never married’ women, we assumed that all other unpartnered women (i.e. separated, divorced, or widowed) who were eligible for the birth interval question at the time of the survey conceived their multiple births from with their late/ex-husband. Since data regarding birth interval were collected from their childbirths in the five years preceding the survey, women may become widowed or divorced or separated shortly (or any time) before the time of the survey but after they already become eligible for providing birth interval information. Therefore, separated, divorced, and widowed women were retained for the analysis. In addition, while separated, divorced, and widowed women were with their late/ex-husbands, there were likely factors present that limited them from achieving an optimum birth interval and these factors should be investigated. Moreover, since some of separated/divorced/widowed women were among those who contributed to the magnitude of short birth interval in the country, including them in the analysis could help to inform policy and health programmers. However, as you hypothesized that ‘women who have multiple births out of wedlock in a culturally conservative society may be unlikely to be planning their births in the same way’, we excluded ‘never married’ (i.e. 12 (0.14%)) women from the analysis in this revised version of the manuscript. Thank you for your careful review. We have described the exclusion of ‘never-married women’ from the analysis (line 115 to 117). In addition, the reasons for retaining separated/divorced/widowed women are described under the exposure variable section (line 147-148). Regarding your concern regarding collinearity among ‘Not partnered’ categories and not married women:

In our previous version of the manuscript, since separated, divorced, and widowed women (i.e. 371 women) were not eligible to answer for their polygyny status, their husband/partner’s education, and their husband's/partner’s occupation, there was systematic missingness for these variables. To address the missing values, we replaced them with a ‘Not partnered’ category values. However, your concern regarding collinearity led us to consider the variable as it is (i.e. with the systematic missingness). Thank you for that. Now, we performed a complete case analysis (CCA). Evidence has shown that CCA provides unbiased results when the chance of being a complete case does not depend on the outcome (i.e. the chance of being complete case for the variable husbands’ education, husbands’ occupation, and polygyny status does not depend on whether the women experienced short birth interval or not) (Hughes RA, Heron J, Sterne JA, Tilling K. Accounting for missing data in statistical analyses: multiple imputation is not always the answer. International journal of epidemiology. 2019;1:11.). Therefore, we believe the concern of collinearity has been addressed. In addition, the results of multicollinearity analysis are annexed as supplementary material (S2 Table. The results of multicollinearity analysis. (DOCX)).

Reviewer #2 comments: The continuous variables (maternal age at marriage) should probably be operationalized in a non-linear way. This can be done either through polynomial transformations (e.g. age at marriage ^2 or higher) or through categorization. Categories are probably a better choice for the presentation of the results. I recommend that the authors examine the data to see if this is appropriate.

Response: We agree with the reviewer and have presented ‘maternal age at first marriage’ in 5-year age categories. However, the highest category is restricted to maternal age at marriage of 30 years and above since the number of observations after 30 years of maternal age becomes small.

Reviewer #2 comments: Does maternal age at birth refer to her age at the birth of the index child or the previous child? If it is her age at the birth of the index child, this variable does not make sense to include in the model. This is because her age at birth is determined by the length of the interval (i.e. your outcome variable). If it is her age at the birth of the previous child, you need to spell this out in the text. Likewise, children ever born should refer to the number of children born prior to the birth of the index child (i.e. not including the index child). It was not clear from the paper or supporting documentation.

Response: Thank you. Previously, we considered maternal age at the birth of the index child. In this revised manuscript, we used maternal age at the birth of the preceding child. Since it does not fulfill the linearity assumption, we presented it in 5-year age categories.

Reviewer #2 comments: If I have understood the results correctly, it seems like community-level factors actually have very little influence on birth interval length once individual-level characteristics are controlled for. The ICC for the full model with individual and community factors was only 2.9%. In the conclusion and policy implications, however, you have interpreted community-level factors as a major concern. Is this justified given the low ICC?

Response: We appreciate your observation. However, an ICC value of 2.9% while small, it is not zero. It indicates that the experience of short birth interval among women still depends on community contexts despite controlling individual- and community-level factors in the final model.

Moreover, in this revised of the manuscript, the magnitude of ICC in the final model increased (from 2.9% to 9.5%). This change could be attributed to the changes made in coding some of exposure variables such as maternal age at first marriage, maternal age at birth of preceding child and total number of children born before the index child. Keeping this change in mind, even though individual- and community-level factors were controlled in the final model, 9.5% of the total variance in the odds of short birth interval was accounted for by between-cluster variation of characteristics. This implies that experience of short birth interval among women still depends on community contexts. As it is known, an ICC of 0 or closer to 0 would suggest that the clusters/community are similar to random samples taken from the country and suggests that clusters/community are not relevant to understand the variation in short birth interval. Whereas a high ICC value informs us that clusters/community are very important in understanding the individual differences in short birth interval (Merlo J, Chaix B, Yang M, Lynch J, Råstam L. A brief conceptual tutorial of multilevel analysis in social epidemiology: linking the statistical concept of clustering to the idea of contextual phenomenon. Journal of Epidemiology & Community Health. 2005;59(6):443-9.). In line with this, our interest is to show that nearly 10% of the total individual differences in short birth interval were due to cluster/community variation. In the manuscript (line 404-407), the statement was presented as follow:

“The random-effect analysis result had also revealed that community-level random effects remained significant after controlling for both individual- and community-level variables, indicating that the experience of short birth interval among women depends on community contexts.”

Reviewer #2 comments: The paper will benefit greatly from language editing for clarity and precision.

Response: The paper has been carefully revised by native English speakers.

Reviewer #2 comments: Line 46: The definition of birth intervals is not quite right. First, the interval is not defined as the number of months between births, but the amount of time. Months are simply the unit that some authors express that time in. It can equivalently be expressed in years (in continuous form) or days. Second, the authors are defining a preceding birth interval here. That is fine, because that is what the paper is about. I would simply suggest that they rephrase the definition as the definition of a preceding interval instead of birth interval in general. This is because there can also be a succeeding birth interval, which would be the interval following the birth of the index child.

Response: The definition has been corrected accordingly.

Reviewer #2 comments: Line 70: How can the prevalence range from 21% to 57.6%? Does this mean across specific sub-populations or over time?

Response: The variation of the magnitude was observed across sub-populations. The 21% prevalence of short birth interval is based on the 2016 Ethiopia Demographic and Health Survey while 57.6% prevalence was based on the 2010 study conducted in one district; Lemo district. For clarity, it is rewritten as follows (line 74-75):

‘In Ethiopia, the prevalence of short birth interval ranges from 21.0% (national estimate) (17) to 57.6% (district level estimate) (18).’

Reviewer #2 comments: Line 223: Age at birth should be reported in more meaningful categories than the ones provided here. The age group 18-34 is enormous and captures the large majority of births. Why not present these as standard 5-year age groups?

Response: Age at birth is presented in the 5-year age categories as per your recommendation.

Reviewer #2 comments: Line 228: ‘Magnitude’ should really be ‘prevalence’.

Response: Corrected accordingly.

Reviewer #2 comments: Line 302 – 314: If you are using the same data as the EDHS, I do not think there is a need to explain why your estimates of the prevalence of short intervals differ from the report. It is clear that you use a different definition of ‘short intervals’ than the DHS program, and that this is the reason why your prevalence is higher. The comparisons to Tanzania, DR Congo, and Bangladesh are not really relevant here. I would recommend dropping this paragraph entirely, and earlier in the paper when you present the prevalence you estimate, just add a line stating that the definition you use is different from the DHS definition, and this is why your estimates are higher.

Response: This paragraph is entirely dropped from the paper. The reason for the difference is also described as follow (line 325-329):

‘The difference in the prevalence of short birth intervals between the current study and the one reported in the 2016 EDHS (21.7%) (17) is due to the different definitions used for short birth interval. The EDHS considered a birth interval of less than 24 months a short birth interval, whereas our study defined it as less than 33 months, which is in accordance with the WHO recommendations (4).’

Reviewer #2 comments: Table 1: Age at first marriage is not included in the table.

Response: It is incorporated in this revised manuscript.

Reviewer #2 comments: Table 3: Where is model 1? The columns start from model 2.

Response: As it is mentioned in the method section, under the heading ‘multivariable multilevel analysis’ (Line 190-193), Model I (Empty model) was fitted without explanatory variables to estimate random variation in the intercept and the intra-cluster correlation coefficient (ICC). Table 3 presents quantities based on fixed effects only. That is why column describing model 1 was not included in table 3. The finding of model 1 is presented in table 4. However, for the sake of clarity, the reason for not including ‘Model I’ in table 3 has been added as a footnote (line 608).

Reviewer #2 comments: I would also set the reference category of education as ‘No education’. Only 2.5% of women had higher education.

Response: The reference category of education has been changed as per your suggestion.

Attachment

Submitted filename: Response to Reviewers.docx

Click here for additional data file.^{(46.6KB, docx)}

PLoS One. doi: 10.1371/journal.pone.0227798.r003

Decision Letter 1

Kannan Navaneetham

31 Dec 2019

Individual and community level determinants of short birth interval in Ethiopia: a multilevel analysis

PONE-D-19-21556R1

Dear Dr. Shifti,

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

Within one week, you will receive an e-mail containing information on the amendments required prior to publication. When all required modifications have been addressed, you will receive a formal acceptance letter and your manuscript will proceed to our production department and be scheduled for publication.

Shortly after the formal acceptance letter is sent, an invoice for payment will follow. To ensure an efficient production and billing process, please log into Editorial Manager at https://www.editorialmanager.com/pone/, click the "Update My Information" link at the top of the page, and update your user information. 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 enable them to help maximize its impact. If they will be preparing press materials for this manuscript, you must inform our press team as soon as possible and 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.

With kind regards,

Kannan Navaneetham

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

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 #1: All comments have been addressed

**********

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

Reviewer #1: Yes

**********

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

Reviewer #1: Yes

**********

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

Reviewer #1: Yes

**********

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

Reviewer #1: Yes

**********

6. Review Comments to the Author

Reviewer #1: (No Response)

**********

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 #1: No

PLoS One. doi: 10.1371/journal.pone.0227798.r004

Acceptance letter

Kannan Navaneetham

6 Jan 2020

PONE-D-19-21556R1

Individual and community level determinants of short birth interval in Ethiopia: a multilevel analysis

Dear Dr. Shifti:

I am 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 notify them about your upcoming paper at this point, to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, 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.

For any other questions or concerns, please email plosone@plos.org.

Thank you for submitting your work to PLOS ONE.

With kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Professor Kannan Navaneetham

Academic Editor

PLOS ONE

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

S1 Table. Description of variables used in the analysis.

(DOCX)

Click here for additional data file.^{(33.8KB, docx)}

S2 Table. The results of multicollinearity analysis.

(DOCX)

Click here for additional data file.^{(29.9KB, docx)}

Attachment

Submitted filename: Response to Reviewers.docx

Click here for additional data file.^{(46.6KB, docx)}

Data Availability Statement

All data are available from the DHS Program database (https://dhsprogram.com/data/available-datasets.cfm). Any researcher can request for access to the data without any restriction.

[pone.0227798.ref001] 1.Rutstein SO. Effects of preceding birth intervals on neonatal, infant and under‐five years mortality and nutritional status in developing countries: evidence from the demographic and health surveys. International Journal of Gynecology & Obstetrics. 2005;89(S1). [DOI] [PubMed] [Google Scholar]

[pone.0227798.ref002] 2.Fallahzadeh H, Farajpour Z, Emam Z. Duration and determinants of birth interval in Yazd, Iran: a population study. Iranian journal of reproductive medicine. 2013;11(5):379 [PMC free article] [PubMed] [Google Scholar]

[pone.0227798.ref003] 3.Yoder PS, Joe L.Lugalla P., Richard F. Sambaiga. Determinants of the Duration of Birth Intervals in Tanzania: Regional Contrasts and Temporal Trends DHS Qualitative Research Studies No 19 Calverton, Maryland, USA: ICF International; 2013. [Google Scholar]

[pone.0227798.ref004] 4.World Health Organization. Report of a WHO Technical Consultation on Birth Spacing. Geneva, Switzerland 13–15 June 2005.

[pone.0227798.ref005] 5.Grisaru-Granovsky S, Gordon E-S, Haklai Z, Samueloff A, Schimmel MM. Effect of interpregnancy interval on adverse perinatal outcomes—a national study. Contraception. 2009;80(6):512–8. 10.1016/j.contraception.2009.06.006 [DOI] [PubMed] [Google Scholar]

[pone.0227798.ref006] 6.Adam I, Ismail MH, Nasr AM, Prins MH, Smits LJ. Low birth weight, preterm birth and short interpregnancy interval in Sudan. The Journal of Maternal-Fetal & Neonatal Medicine. 2009;22(11):1068–71. [DOI] [PubMed] [Google Scholar]

[pone.0227798.ref007] 7.Chen I, Jhangri GS, Chandra S. Relationship between interpregnancy interval and congenital anomalies. American journal of obstetrics and gynecology. 2014;210(6):564 e1-8. 10.1016/j.ajog.2014.02.002 [DOI] [PubMed] [Google Scholar]

[pone.0227798.ref008] 8.Kwon S, Lazo-Escalante M, Villaran M, Li C. Relationship between interpregnancy interval and birth defects in Washington State. Journal of Perinatology. 2012;32(1):45 10.1038/jp.2011.49 [DOI] [PubMed] [Google Scholar]

[pone.0227798.ref009] 9.Cheslack-Postava K, Liu K, Bearman PS. Closely spaced pregnancies are associated with increased odds of autism in California sibling births. Pediatrics. 2011;127:246–53. 10.1542/peds.2010-2371 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0227798.ref010] 10.Hussaini KS, Ritenour D, Coonrod DV. Interpregnancy intervals and the risk for infant mortality: a case control study of Arizona infants 2003–2007. Maternal and child health journal. 2013;17(4):646–53. 10.1007/s10995-012-1041-8 [DOI] [PubMed] [Google Scholar]

[pone.0227798.ref011] 11.DaVanzo J, Razzaque A, Rahman M, Hale L, Ahmed K, Khan MA, et al. The effects of birth spacing on infant and child mortality, pregnancy outcomes, and maternal morbidity and mortality in Matlab, Bangladesh. Technical Consultation and Review of the Scientific Evidence for Birth Spacing. 2004. [Google Scholar]

[pone.0227798.ref012] 12.Setty-Venugopal V, Upadhyay UD. Birth spacing: three to five saves lives. Population Reports Series L: Issues in World Health. 2002(13):1–23. [PubMed] [Google Scholar]

[pone.0227798.ref013] 13.Abdel-Tawab N, Lūzā SF, Zaki A. Helping Egyptian women achieve optimal birth spacing intervals through fostering linkages between family planning and maternal/child health services: Citeseer; 2008. [Google Scholar]

[pone.0227798.ref014] 14.Rutstein SO. Trends in Birth Spacing DHS Comparative Reports No 28 Calverton, Maryland, USA: ICF Macro; 2011. [Google Scholar]

[pone.0227798.ref015] 15.Rutstein SO. Further evidence of the effects of preceding birth intervals on neonatal infant and under-five-years mortality and nutritional status in developing countries: Evidence from the Demographic and Health Surveys. Macro International Inc; 2008. [DOI] [PubMed] [Google Scholar]

[pone.0227798.ref016] 16.RamaRao S, Townsend J, Askew I. Correlates of inter-birth intervals: implications of optimal birth spacing strategies in Mozambique. 2006. [Google Scholar]

[pone.0227798.ref017] 17.Central Statistical Agency (CSA) [Ethiopia] and ICF. Ethiopia Demographic and Health Survey 2016. Addis Ababa, Ethiopia, and Rockville, Maryland, USA: CSA and ICF. 2016.

[pone.0227798.ref018] 18.Yohannes S, Wondafrash M, Abera M, Girma E. Duration and determinants of birth interval among women of child bearing age in Southern Ethiopia. BMC pregnancy and childbirth. 2011;11(1):38. [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0227798.ref019] 19.Hailu D, Gulte T. Determinants of Short Interbirth Interval among Reproductive Age Mothers in Arba Minch District, Ethiopia. International journal of reproductive medicine. 2016;2016. [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0227798.ref020] 20.Tsegaye D, Shuremu M, Bidira K. Practice of child spacing and its associated factors among women of child bearing age (15 to 49 years) in Illubabor zone, South West Ethiopia. International Journal of Nursing and Midwifery. 2017;9(7):102–8. [Google Scholar]

[pone.0227798.ref021] 21.Begna Z, Assegid S, Kassahun W, Gerbaba M. Determinants of inter birth interval among married women living in rural pastoral communities of southern Ethiopia: a case control study. BMC pregnancy and childbirth. 2013;13(1):116. [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0227798.ref022] 22.Sallis JF, Owen N, Fisher E. Ecological models of health behavior. Health behavior: Theory, research, and practice. 2015;5:43–64. [Google Scholar]

[pone.0227798.ref023] 23.McLaren L, Hawe P. Ecological perspectives in health research. Journal of Epidemiology & Community Health. 2005;59(1):6–14. [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0227798.ref024] 24.Violence Prevention Alliance. The ecological Framework. 2018 [Available from: http://www.who.int/violenceprevention/approach/ecol.

[pone.0227798.ref025] 25.Salihu HM, Wilson RE, King LM, Marty PJ, Whiteman VE. Socio-ecological model as a framework for overcoming barriers and challenges in randomized control trials in minority and underserved communities. International Journal of MCH and AIDS. 2015;3(1):85 [PMC free article] [PubMed] [Google Scholar]

[pone.0227798.ref026] 26.De Jonge HC, Azad K, Seward N, Kuddus A, Shaha S, Beard J, et al. Determinants and consequences of short birth interval in rural Bangladesh: a cross-sectional study. BMC pregnancy and childbirth. 2014;14(1):427. [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0227798.ref027] 27.Tessema GA, Zeleke BM, Ayele TA. Birth interval and its predictors among married women in Dabat District, Northwest Ethiopia: A retrospective follow up study. African journal of reproductive health. 2013;17(2):39–45. [PubMed] [Google Scholar]

[pone.0227798.ref028] 28.Bassey G, Nyengidiki TK, Dambo ND. Determinants of interpregnancy interval among parturient in Port Harcourt, Nigeria. Sahel Medical Journal. 2016;19(4):180. [Google Scholar]

[pone.0227798.ref029] 29.Chirwa TF, Mantempa JN, Kinziunga FL, Kandala JD, Kandala N-B. An exploratory spatial analysis of geographical inequalities of birth intervals among young women in the Democratic Republic of Congo (DRC): a cross-sectional study. BMC pregnancy and childbirth. 2014;14(1):271. [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0227798.ref030] 30.Exavery A, Mrema S, Shamte A, Bietsch K, Mosha D, Mbaruku G, et al. Levels and correlates of non-adherence to WHO recommended inter-birth intervals in Rufiji, Tanzania. BMC pregnancy and childbirth. 2012;12(1):152. [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0227798.ref031] 31.Karkee R, Lee AH. Birth spacing of pregnant women in Nepal: a community-based study. Frontiers in Public Health. 2016;4:205 10.3389/fpubh.2016.00205 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0227798.ref032] 32.FMoH. Health System Special Support. 2015 [Available from: http://www.moh.gov.et/ejcc/en/health-system-special-support.

[pone.0227798.ref033] 33.Goldstein H. Multilevel statistical models 4th ed: UK, University of Bristol; 2011. [Google Scholar]

[pone.0227798.ref034] 34.Merlo J, Chaix B, Ohlsson H, Beckman A, Johnell K, Hjerpe P, et al. A brief conceptual tutorial of multilevel analysis in social epidemiology: using measures of clustering in multilevel logistic regression to investigate contextual phenomena. J Epidemiol Community Health. 2006;60(4):290–7. 10.1136/jech.2004.029454 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0227798.ref035] 35.Merlo J. Multilevel analytical approaches in social epidemiology: Measures of health variation compared with traditional measures of association. Journal of Epidemiology and Community Health. 2003; 57(8):550–1. 10.1136/jech.57.8.550 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0227798.ref036] 36.Raykov T, Marcoulides GA. Intraclass correlation coefficients in hierarchical design studies with discrete response variables: A note on a direct interval estimation procedure. Educational and psychological measurement. 2015;75(6):1063–70. 10.1177/0013164414564052 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0227798.ref037] 37.Merlo J, Chaix B, Yang M, Lynch J, Råstam L. A brief conceptual tutorial of multilevel analysis in social epidemiology: linking the statistical concept of clustering to the idea of contextual phenomenon. Journal of Epidemiology & Community Health. 2005;59(6):443–9. [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0227798.ref038] 38.Merlo J, Yang M, Chaix B, Lynch J, Råstam L. A brief conceptual tutorial on multilevel analysis in social epidemiology: investigating contextual phenomena in different groups of people. Journal of Epidemiology & Community Health. 2005;59(9):729–36. [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0227798.ref039] 39.Vrieze SI. Model selection and psychological theory: a discussion of the differences between the Akaike information criterion (AIC) and the Bayesian information criterion (BIC). Psychological methods. 2012;17(2):228 10.1037/a0027127 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0227798.ref040] 40.Gagné P, Dayton CM. Best regression model using information criteria. Journal of Modern Applied Statistical Methods. 2002;1(2):57. [Google Scholar]

[pone.0227798.ref041] 41.Akaike H. A new look at the statistical model identification Selected Papers of Hirotugu Akaike: Springer; 1974. p. 215–22. [Google Scholar]

[pone.0227798.ref042] 42.Carter JV, Pan J, Rai SN, Galandiuk S. ROC-ing along: Evaluation and interpretation of receiver operating characteristic curves. Surgery. 2016;159(6):1638–45. 10.1016/j.surg.2015.12.029 [DOI] [PubMed] [Google Scholar]

[pone.0227798.ref043] 43.MacQuarrie KLD. Marriage and Fertility Dynamics: The Influence of Marriage Age on the Timing of First Birth and Birth Spacing DHS Analytical Studies No 56 Rockville, Maryland, USA: ICF International; 2016. [Google Scholar]

[pone.0227798.ref044] 44.Padmadas SS, Hutter I, Willekens F. Compression of women's reproductive spans in Andhra Pradesh, India. International family planning perspectives. 2004:12–9. 10.1363/3001204 [DOI] [PubMed] [Google Scholar]

[pone.0227798.ref045] 45.van Soest A, Saha UR. Relationships between infant mortality, birth spacing and fertility in Matlab, Bangladesh. PloS one. 2018;13(4):e0195940 10.1371/journal.pone.0195940 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0227798.ref046] 46.Wang W, Mallick L. Understanding the relationship between family planning method choices and modern contraceptive use: an analysis of geographically linked population and health facilities data in Haiti. BMJ Global Health. 2019;4(Suppl 5):e000765 10.1136/bmjgh-2018-000765 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0227798.ref047] 47.Shiferaw S, Spigt M, Seme A, Amogne A, Skrøvseth S, Desta S, et al. Does proximity of women to facilities with better choice of contraceptives affect their contraceptive utilization in rural Ethiopia? PloS one. 2017;12(11):e0187311 10.1371/journal.pone.0187311 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0227798.ref048] 48.Silumbwe A, Nkole T, Munakampe MN, Milford C, Cordero JP, Kriel Y, et al. Community and health systems barriers and enablers to family planning and contraceptive services provision and use in Kabwe District, Zambia. BMC health services research. 2018;18(1):390 10.1186/s12913-018-3136-4 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0227798.ref049] 49.Yitbarek K, Abraham G, Morankar S. Contribution of women’s development army to maternal and child health in Ethiopia: a systematic review of evidence. BMJ open. 2019;9(5):e025937 10.1136/bmjopen-2018-025937 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0227798.ref050] 50.Damtew ZA, Karim AM, Chekagn CT, Zemichael NF, Yihun B, Willey BA, et al. Correlates of the Women’s Development Army strategy implementation strength with household reproductive, maternal, newborn and child healthcare practices: a cross-sectional study in four regions of Ethiopia. BMC pregnancy and childbirth. 2018;18(1):373. [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0227798.ref051] 51.Caglia J, Kearns A, Langer A. Health Extension Workers in Ethiopia: Delivering Community-based Antenatal and Postnatal Care. Boston, MA: Maternal Health Task Force; 2014. [Google Scholar]

PERMALINK

Individual and community level determinants of short birth interval in Ethiopia: A multilevel analysis

Desalegn Markos Shifti

Catherine Chojenta

Elizabeth G Holliday

Deborah Loxton

Roles

Abstract

Background

Objective

Methods

Results

Conclusion

Introduction

Methods

Study area and setting

Study design and sampling

Fig 1. The flowchart for the sampling and data extraction procedure, EDHS 2016 EDHS, Ethiopian Demographic and Health Survey.

Outcome variable

Exposure variables

Data analysis

Descriptive analysis

Multivariable multilevel analysis

Model fit statistics

Ethics statement

Results

Sociodemographic and other health related characteristics

Table 1. Weighted proportion of short birth interval by individual-level factors, EDHS 2016.

Table 2. Weighted proportion of short birth interval by community-level factors, EDHS 2016.

Determinants of short birth interval among women

Multilevel analysis (fixed effect analysis)

Table 3. Individual- and community-level determinants of short birth interval in Ethiopia using multilevel logistic regression analysis, EDHS 2016.

Individual-level factors

Community-level factors

Multilevel analysis (random-effects analysis)

Table 4. Results from random intercept model (measure of variation) for short birth interval at cluster level using multilevel logistic regression analysis.

Model fit statistics

Discussion

Conclusion

Implications for policy practice and further research

Supporting information

Acknowledgments

Data Availability

Funding Statement

References

Decision Letter 0

Kannan Navaneetham

Roles

Author response to Decision Letter 0

Decision Letter 1

Kannan Navaneetham

Roles

Acceptance letter

Kannan Navaneetham

Roles

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases