Disparities in obstetric danger sign knowledge among pregnant women in postwar Eastern Tigray,Ethiopia: Robust logistic regression

Abstract

Background:

Maternal death is a foremost public health concern in underdeveloped countries, particularly in postwar Eastern Tigray, Ethiopia, where healthcare infrastructure is limited. Identifying obstetric risk factors is crucial to reducing unnecessary maternal mortality. Disparities in knowledge of obstetric danger signs are a critical barrier to reducing maternal mortality, particularly in fragile, post-conflict settings.

Objective:

The main goal is to measure the knowledge disparity of pregnant women about obstetric danger signs and associated factors at postwar Eastern Tigray, Ethiopia, health facility centers using penalized and Bayesian logistic regression.

Methods:

We conducted a health-facility-based cross-sectional study. Data were collected from 102 pregnant women, primarily from rural areas, using structured interviews. Predictors included socioeconomic factors (education, residence, and income), healthcare access (distance to a health facility), and obstetric history (parity). Knowledge of obstetric danger signs was the outcome, classified as “good” (⩾3 signs identified) or “poor.” Associations were analyzed using penalized logistic regression and Bayesian logistic regression.

Results:

Descriptive statistics revealed profound disparities in knowledge: prevalence was 92.3% in urban versus 12.7% in rural residents, 93.5% for the shortest versus 5.1% for the longest travel time, and 5.6% for the lowest versus 91.7% for higher education levels. Penalized regression identified urban residence as the strongest predictor (adjusted odds ratio = 27.4, p = 0.002). Bayesian analysis affirmed significant associations for education Level 2 (median odds ratio = 78.2, 95% credible interval: 9.7, 1042.4). The longest travel time was associated with a 97% reduction in the odds of knowledge (odds ratio = 0.03, 95% credible interval: 0.002, 0.34). Model comparison confirmed the superiority of penalized regression over the failed standard model.

Conclusion:

Knowledge of obstetric danger signs is predominantly determined by structural access factors specifically urban residence and travel time to health facilities followed by educational attainment. These findings highlight the urgent need for targeted, equity-focused interventions that address geographical and infrastructural barriers to health information in postwar Eastern Tigray, Ethiopia.

Keywords

Antenatal care Bayesian logistic regression Firth’s penalized logistic regression knowledge of obstetric danger signs pregnant women

Introduction

The maternal death remains a distressing worldwide health burden, concentrated tremendously in developing countries where obstetric complications are a foremost leading cause of mortality among pregnant women of reproductive age.^1,2 The World Health Organization estimates that approximately 15% of all pregnant women will grow a potentially life-threatening complication, including severe hemorrhage, pregnancy-induced hypertension, or prolonged labor.³ In the 25-year period from 1990 to 2015, an estimated 10.7 million reproductive aged mothers died from such pregnancy complications worldwide, with 99% of these deaths happening in developing countries.⁴ In Ethiopia, the situation is predominantly severe; the most recent Ethiopian Demographic and Health Survey point out a high motherly death ratio, with approximately 4 maternal losses per 1000 births.⁵ This crisis continues regardless of the global promise under Sustainable Development Goal 3 to reduce the worldwide maternal death ratio to less than 70 per 100,000 live births by 2030.^4,6

A life-threatening and modifiable cause subsidizing to these preventable maternal losses is the extensive prevalence of lack of awareness and knowledge of obstetric danger signs among reproductive pregnant women. Danger signs comprising severe vaginal bleeding, loss of consciousness, high fever, convulsions, swollen hands/face, and blurred vision are acute indicators that signal the prerequisite for instant medical intervention.^7,8 The little consciousness of these signs during pregnancy, labor, and the postnatal period leads directly to delays in seeking care, a vital determining factor of maternal death.^2,9 Distressingly, maternal loss is often not alleged as a substantial personal health risk within a lot of communities, however, rather acknowledged as a natural or inevitable result of childbirth.¹⁰ As a result, education about potential pregnancy complications is not simply informative but is a key pillar for enabling informed decision-making and timely healthcare access.^2,6

Many studies conducted through different regions of Ethiopia disclose momentous disparities in the knowledge of obstetric danger signs, underlining both progress and persistent gaps. Research in facility-based settings, which characteristically grasp pregnant women already, engaged with the health system, reported relatively higher knowledge levels of obstetric danger signs. Notably, research in Arba Minch Town found that 68.4% of reproductive women discerned about pregnancy-related danger signs, with vaginal bleeding being almost collectively familiar (98.37%).⁸ Likewise, a research in Harar town conveyed that 70% of antenatal care (ANC) attendees were conscious about obstetric danger signs, even though 30% could not name a single sign.¹¹ However, these facts contrast severely with results from community-based studies, which provide a more illustrative depiction of the overall population. A research in the Erer district of the Somali region institutes that merely 15.5% of reproductive women were familiar with obstetric danger signs through all obstetric periods (gravidity, childbearing, and postnatal).¹² This unambiguous discrepancy underlines that knowledge is not consistently dispersed and that reproductive women who do not access health facilities regularly, especially those who are most vulnerable, remain profoundly underserved.

The existing literature has steadily recognized a multifactorial framework of predictors related with a reproductive women’s knowledge of obstetric danger signs. Socioeconomic and demographic factors are foundational, with maternal education evolving as one of the solidest and furthermost consistent predictors. The reproductive women with secondary or higher education level have been shown to be significantly more knowledgeable, with odds ratios (ORs) ranging from 1.46 to 4.42 relative to their illiterate counterparts.^13,14 Furthermore, the occupation and access to mass media are also significant, as evinced in Gedo Town, where having mass media in the house made reproductive women three times more likely to have sensible knowledge about obstetric danger signs.¹³ Further than individual characteristics, healthcare access and engagement are essential. Moreover, reproductive women of urban residence,⁴ earlier initiation of ANC,¹⁵ a higher number of ANC visits,¹⁶ and receiving direct health education for the duration of healthcare⁴ are all powerfully and positively correlated with higher knowledge levels of obstetric danger signs. This form of evidence realistically explains the pathways over which knowledge is acquired, stressing the intertwined roles of socioeconomic privilege and health system interaction.

Regardless of this strong evidence, critical research gaps remain, above all, concerning delicate and post-conflict settings. Furthermost existing literatures and studies in Ethiopia have been conducted in comparatively stable, accessible regions. There is a profound shortage of intensive research in communities recovering from war, such as those in Eastern Tigray. These situations present unique, compounded vulnerabilities: health infrastructure is often smashed or demolished, routine health facilities like ANC are harshly disrupted, inhabitants may be displaced internally or externally, and social networks are broken. The well-known predictors of knowledge such as ANC attendance or access to a health services may be rendered out of use or may function another way in such a setting. It is blurred whether the known sociodemographic predictors maintain their strength or if new, conflict-related factors (e.g. displacement status, trauma, and loss of household assets) emerge as most important determining factors. Besides, the extreme disparities in knowledge observed between facility and community studies in unchanging sceneries propose that in a postwar setting, where facility access is restricted for many, the overall knowledge gap could be terrible and its drivers more intricate.

Hence, this article is designed to directly discourse these contextual research gaps. Its principal purpose is to pinpoint and quantify, through rigorously modeling, the predictors of knowledge disparity concerning obstetric danger signs, specifically among pregnant women in a high risk settings of postwar community in Eastern Tigray, Northern Ethiopia using robust regression approaches, specifically Firth’s penalized logistic regression and Bayesian logistic regression approaches. By concentrating on these fragile settings, the article tested the applicability of the recognized multifactorial framework incorporating sociodemographic, economic, and obstetric service utilization variables in a context of acute recovery. The findings determined which determining factors remain significant controls for intervention when systems are destabilized and illuminated any unique obstacles enacted by the post-conflict setting. Eventually, this study seeks to make context-specific evidence that can notify the design of targeted, resilient educational and health service delivery strategies. Such strategies are urgently required to bridge the contextual knowledge gap, alleviate the “three delays,” and contribute to reducing maternal death in line with nationwide and worldwide targets, even in the furthermost inspiring settings.

Methodology

Study design

The study used health facility-based analytical cross-sectional research design to assess the knowledge of obstetric danger signs among pregnant mothers attending ANC. Well-administered questionnaire was used as a data collection tool, and consecutive sampling of pregnant women was employed. The study considered the volunteer pregnant women who had been visiting at 46 health facility centers of eastern Tigray, Ethiopia for more than four times. The Cronbach’s alpha test was used to check the reliability and validity of the questionnaire or survey data. We applied the Firth’s penalized binary logistic regression, regularizations for robustness, and Bayesian binary logistic regression to overcome the failure of standard model and its instability of results, separation problem, and multicollinearity problems in standard binary logistic regression analysis.

Study area and period

This study was conducted in the postwar settings of Eastern Tigray, Northern Ethiopia. It was undertaken as part of a larger project entitled “A Rigorous Statistical Analysis of Health Disparities in Postwar Communities in Eastern Tigray, Ethiopia,” which was implemented over a 2-year period from December 2023 to June 2025. The current article was derived from the whole project undertakings and documentation and is reported here as a separate research article.

Study populations

The study considered total of 310 volunteer pregnant women who had been attending ANC for at least 4 visiting times at the health facility centers during the data collection period as a target population. The sampling frame consisted of all 46 health facility centers offering ANC services in Eastern Tigray, Ethiopia

Inclusion-exclusion criteria

Inclusion

The eligible participants were pregnant women who had attended at least four antenatal care (4 + ANC) visits. The study was conducted in health facilities across the Eastern Zone of Tigray, Ethiopia. However, due to security constraints arising from military occupation in parts of the region, the study was limited to 16 of the 18 districts of the eastern Tigray region of the Ethiopia, plus one primary hospital in the Irob district, all of which were under governmental control and accessible except Zalambesa district and partially Irob district were occupied by the Eritrean troops during the data collection period.

Exclusion

The study excluded pregnant women who attended fewer than four ANC visits and pregnant women who were severely ill or who had no willingness to provide informed consent. Geographically, the most areas of two districts namely, Zalambesa district and the Irob district except one primary hospital in Dawhan town were also excluded due to the occupation of most areas within these two districts by Eritrean troops and were rendered non-functional as a result of war damage during the period of data collection.

Sample size determination and sampling technique

n = \frac{Z_{\frac{α}{2}}^{2} P (1 - P) N}{e^{2} (N - 1) + Z_{\frac{α}{2}}^{2} P (1 - P)},

where n is the sample size, N is the population size, e = 0.08 is the margin of error, p = 0.5 is the population proportion required for optimal sample size, and $Z = 1.96$ is the standard variate at given 95% confidence level (CI). Therefore, the sample size is calculated as:

\begin{matrix} n = \frac{Z_{\frac{α}{2}}^{2} P (1 - P) N}{e^{2} (N - 1) + Z_{\frac{α}{2}}^{2} P (1 - P)} \\ = \frac{{1.96}^{2} * 0.5 (1 - 0.5) * 310}{{0.08}^{2} (310 - 1) + {1.96}^{2} * 0.5 (1 - 0.5)} \\ = \frac{297.724}{2.9380} = 101.34 \approx 102 . \end{matrix}

Thus, the required sample size was calculated using the formula for a single proportion with finite population correction. With a total estimated population of 310 pregnant women, an assumed proportion of 50%, a 95%CI, and a margin of error of 8%, the minimum sample required was 102 participants. An 8% margin of error was adopted as a pragmatic balance between precision and feasibility in the immediate postwar setting, where logistical and security constraints precluded a stricter margin. To reflect the operational reality of concentrated service access with only urban general hospitals functional immediately following the conflict, a purposive sampling strategy was employed. Quotas were weighted by patient volume across the tiered, accessible facilities: five pregnant women from each non-hospital health center, six from each primary hospital, ten from Wukro General Hospital, and 13 from Adigrat General Hospital. While this non-probability approach limits statistical generalizability to a fully recovered health system, it ensures that the sample is contextually representative of where ANC was actually delivered during the acute recovery phase, providing operationally valid evidence for urgent health system rehabilitation.

Study variables

Dependent variables: Knowledge of pregnant women about obstetric danger signs.

Independent variables: Sociodemographic and economic factors: residence (rural, urban), distance measured using the time taken to reach at the health facility centers in hours, religion, ethnic group, marital status, educational level, occupation, monthly income, and husband support.

Obstetric history: Parity of pregnancy, age at first delivery, and place of most recent delivery.

Operational definitions

Obstetric Danger signs of pregnancies are warning signs that women encounter during pregnancy.

Danger signs: Presence of a condition that increases the chances of a pregnant woman and/or her unborn child dying or having poor health.

Mothers: Refers to women who are pregnant and/ or who have a child.

Knowledge of obstetric danger signs—means the basic information that the mothers have regarding obstetric danger signs.

Knowledge of obstetric danger signs was determined by measuring the recognition of explicit clinical signs through four critical periods. Accordingly, the phases and their key signs were defined as:—Phase 1 (Pregnancy): vaginal bleeding, swollen hands or face, severe headache, blurred vision, and lower abdominal pain. Phase 2 (Labor/Childbirth): severe vaginal bleeding, prolonged labor (>12 h), convulsions, trouble breathing, and retained placenta. Phase 3 (Postpartum): severe vaginal bleeding, foul-smelling vaginal discharge, and fever. Phase 4 (Neonatal): pitched cry, trouble feeding/suckling, convulsions, and loss of consciousness, hyperthermia, problematic breathing, jaundice, and failure to pass urine/stool within 24 h. This sorting was based on established maternal and newborn health references.^4,16,17

Consequently, knowledge of obstetric danger signs was categorized using a scoring system adapted from a study in Southern Tanzania.¹⁸ Women were measured across four phases of their pregnancy. They were classified as having no knowledge (mentioning no signs in any phase), low knowledge (up to three signs, with ⩾1 from each phase), moderate knowledge (4–7 signs, with ⩾1 from each phase), or adequate knowledge (⩾8 signs). For final analysis, these categories were dichotomized into good knowledge and poor knowledge.

The study assessed reproductive women’s knowledge of obstetric danger signs during pregnancy at 46 health facility center. The pregnant women were asked to mention three key signs. The research classified women as having good or poor knowledge, with good knowledge being achieved when they could mention these signs.^4,16,17

Good knowledge refers to those participants who mentioned at least three obstetric danger signs.^4,16,17

Poor knowledge refers to those participants who cannot mention at least three obstetric danger signs.

Data collection instruments and procedure

The study involved collecting data in the ANC Room, obtaining consent from respondents through verbal and written means, and preparing a structured questionnaire in English and Tigrigna for better understanding. The questionnaire was then translated back to English for consistency, followed by face-to-face interviews. The questionnaire was developed by the research team based on relevant literature and expert input and administered to 102 participants (32.9%) of the target population (n = 310). The instrument was not pilot-tested prior to data collection. However, Internal consistency reliability was assessed post data collection using Cronbach’s alpha, which yielded a value of 0.859, indicating good internal consistency.

Data processing and analyzing

The descriptive statistics for data was analyzed using SPSS version 20 and categorized based on information type, displayed through tables, frequencies, percentages, and graphs, and reviewed by supervisors and invigilators. R-statistical software was used for the penalized binary logistic regression analysis, regularizations, and Bayesian binary logistic regression analysis.

Statistical model frameworks

The chi-square method

The chi-square technique is used to find out whether there is association between row and column.

Chi-square test of independence following steps:

H₀: the row variable and column variable are independent.

H₁: the row variable and column variable are dependent.

At level of significance alpha (α):

Assumption

✓ The sample must be randomly selected from the population.

✓ The population must be normally distributed for the variable understudy.

✓ The observation must be independent of each other.

✓ The expected frequency for each category is at least five. If there is a category with all expected frequency of less than 5, either increase the sample size or combine two or more categories make each expected frequency at least five.

✓ Test statistics for the test of Independent

Let O _{i j} respect the observed number counts in the ( $i j$ )th cell, and let $E i j$ represent the expected number of counts in the ( $i j$ )th cell. Then:

X_{C a l}^{2} = \frac{\sum_{i = 1}^{k} \sum_{j = 1}^{p} {(O_{i j} - E_{i j})}^{2}}{E_{i j}}

$O_{ij} =$ the observed frequency

$E_{I j} =$ the expected frequency.

$X_{C a l}^{2}$ as x² (r–1) (c–1) to show the above distribution relationship. Test statistic x²ca follows a chi-square distribution with (r–1) (c–1) degree of freedom.

Binary logistic regression

The study utilized a binary logistic regression model to evaluate the factors influencing knowledge of obstetric danger signs, a model suitable for binary outcome variables:

\begin{array}{l} π (x) = \frac{e^{{β_{0} + β_{1} X_{1} + β_{2} X_{2} + β_{3} X_{3} \dots + β_{k} X_{k}}}}{1 + e^{{β_{0} + β_{1} X_{1} + β_{2} X_{2} + β_{3} X_{3} \dots + β_{k} X_{k}}}} \\ = \frac{e^{X^{T} β}}{1 + e^{X^{T}, β i}}, \end{array}

(1)

The OR is obtained for this is:

\begin{array}{l} l o g i t (π (x)) = \log (\frac{π (x)}{1 - π (x)}) \\ = β_{0} + β_{1} X_{1} + β_{2} X_{2} + β_{3} X_{3} \dots + β_{k} X_{k} \\ = X^{T} β, \end{array}

(2)

where $X^{T} = (X_{1}, X_{2}, X_{3}, \dots, X_{k})$ are k number of predictors, $β = (β_{0}, β_{1}, β_{2}, \dots, β_{k})$ are unknown regression coefficients parameters, $π (x)$ is the probability of having good knowledge of obstetric danger signs, and $1 - π (x)$ is the probability of having poor knowledge of obstetric danger signs.

Results and discussions

Results

This section includes a brief overview and explanation of the outcomes of an institutional cross-sectional survey study designed to examine reproductive women’s knowledge of obstetric risk symptoms during pregnancy and the associated factors. These include sociodemographic variables and obstetric history. Primary data were acquired cross-sectionally using an interview-based questionnaire and thoroughly processed. Data were then manipulated and analyzed. The quantitative data were summarized using descriptive and inferential statistics, which were then presented in an ordered manner utilizing frequency tables, percentages, graphs, and charts.

The authors used the chi-square test and binary logistic regression model to investigate the association between women’s understanding of obstetric risk indicators during pregnancy and the associated factors. The study sought to determine the impact of these variables on women’s awareness of obstetric risk indications during pregnancy. Furthermore, the study intended to assess the degree and direction of the association between reproductive women’s knowledge of obstetric danger signs and the associated parameters.

The reliability and validly or internal consistency of the questionnaire are presented in Table 6 of Supplemental Material. The Cronbach’s alpha was calculated to assess internal consistency and validity, yielding a value of ( $α = 0.859$ ). As this exceeds the conventional threshold or benchmark of 70%, the scale verified good reliability and validity for use in this cross-sectional survey study.

General descriptions of sociodemographic factors

The sociodemographic information was presented in both tabular and graphical versions. The tables and charts show how these factors are distributed using frequencies and percentages. The study’s conclusions are based on the objectives and analysis of replies supplied by pregnant women visiting health centers in eastern Tigray, Ethiopia. This study recruited 102 pregnant women from the reproductive population. As a result, the descriptive summary statistics, graphical representations, chi-square test for independence, and binary logistic regression model findings are presented in the relevant figures and tables in this section.

Table 1 illustrates the frequencies and percentages of sociodemographic characteristics of the women. Thus, the table illustrates the counts and percentages of sociodemographic characteristics of pregnant women coming for antenatal service at health centers in eastern Tigray.

Table 1.

Sociodemographic history.

Sociodemographic history		Frequency	Percent
Age	15–19	7	6.9
	20–24	33	32.4
	25–29	25	24.5
	30–34	17	16.7
	35–39	12	11.8
	40 and above	8	7.8
	Total	102	100.0
Marital status	Single	1	1.0
	Married	96	94.1
	Widowed	2	2.0
	Divorced	3	2.9
	Total	102	100.0
Religion	Orthodox	102	100.0
Ethnicity	Tigray	102	100.0
Occupation	Housewife	71	69.6
	Gov’t employee	9	8.8
	Private employee	18	17.6
	Business	4	3.9
	Total	102	100.0
Educational status	Illiterates	54	52.9
	Read and write	29	28.4
	Elementary school	12	11.8
	Secondary school & above	7	6.9
	Total	102	100.0
Monthly income	<500 Birr	72	70.6
	501–1000 Birr	26	25.5
	1001–1500 Birr	2	2.0
	1501–2000 Birr	2	2.0
	Total	102	100.0
Husband support	Yes	97	95.1
	No	4	3.9
	Sometimes	1	1.0
	Total	102	100.0
In what ways did your husband support you in attending ANC?	Economically	63	64.3
	Morally	11	11.2
	Both economically & morally	24	24.5
	Total	98	100
Family size	1–3	37	36.3
	4–6	51	50.0
	>6	14	13.7
	Total	102	100.0

ANC: Antenatal care.

To begin, the percentages of women’s age coming for antenatal checkups the highest number of women about 33 (32.4%) of the total were between ages 20 and 24 years old followed by about 25 (24.5%) between ages 25–29 and 17 (16.7%) within ages 30–34. Women aged 15–19 represented the smallest proportion (6.9%, n = 7), followed by those aged over 40 (7.8%, n = 8). Generally, about 80.4% of the women are aged less than or equal to 34 years old. This means that most of the women following antenatal checkups were young. Figure 1 in Supplemental Material depicts similar facts about age distributions of reproductive women.

Regarding marital status, among the total women following antenatal service at health centers, only one was single. Likewise, 5 (4.9%) of the women were widowed and divorced. However, the maximum number of women following antenatal service at the health centers was 96 (94.1%). In terms of religion and ethnicity, all the women were Orthodox and Tigray, respectively. In terms of occupation, a large percentage of women 71 (69.6%) were housewives followed by private employees 18 (17.6%). Figure 2 in Supplemental Material depicts similar facts about the distribution of the marital status of reproductive women.

In terms of education, most of the women following antenatal service were illiterate, covering 54 (52.9%) of the total respondents, followed by 29 (28.4%) who can read and write. Overall, 93.1% of the women had completed elementary school or less. Moreover, the monthly income of most of the women who were interviewed is less than 500 birrs, which is 72 (70.6%) of the total. Overall, about 96.1% of the women’s monthly income is less than or equal to 1000 birrs. In terms of support, most of the women 97 (95.1%) were being supported by their husbands. Among the women who were being supported by their husbands, about 63 (64.29%) were supported economically. In terms of family size, about half 51 (50%) of the women had 4–6 number of family size. For further details, see Table 1.

The pie chart in Figure 1 shows that many of the women, about 63 (61.76%), were rural residents. The remaining 39 (38.24%) resided in urban areas. This means the dominant number of women cannot access health facilities easily.

Figure 1.

Pie chart for residence of reproductive women at health facility centers.

The pie chart in Figure 2 illustrates the distance from the health facility center on foot in hours. Many pregnant women, about 39 (38.24%) of the total, responded that they travel long distances to get health facilities. This was compounded by travel times exceeding 2 h to the nearest health facility during pregnancy. The remaining 32 (31.37%) and 31 (30.39%) of the pregnant women responded that they travel 1–2 h and less than 1 h, respectively. Therefore, many women travel long distances to get health facilities during their pregnancy time.

Figure 2.

Pie chart for distance from health facility center in hours.

General descriptions of obstetric history

The sampled reproductive women who were attending antenatal healthcare were asked to answer about their obstetric history in brief. Based on the responses of the sampled women, the general descriptions of obstetric history were summarized in Table 2.

Table 2.

Distribution of obstetric characteristics among participants.

Obstetric history		Frequency	Percent
How many times have you been pregnant?	None	18	17.6
	One time	28	27.5
	Two times	26	25.5
	Three and above	30	29.4
	Total	102	100.0
How many children do you have?	No	5	4.9
	One	30	29.4
	Two	17	16.7
	Three	13	12.7
	>Three	37	36.3
	Total	102	100.0
How many years are between your deliveries on average?	<1 year	1	1.0
	1 year	32	31.4
	2 year	69	67.6
	Total	102	100.0
What was your age at first delivery?	<15 year	21	20.6
	16–25 year	79	77.5
	>25 year	1	1.0
	I don’t know	1	1.0
	Total	102	100.0
At what gestational age was your first visit? In months?	1–3 months	32	31.4
	4–6 months	68	66.7
	I don’t remember	2	2.0
	Total	102	100.0
Pregnancies that resulted in a baby that was born Alive?	0	4	3.9
	1	30	29.4
	2	20	19.6
	above 2	48	47.1
	Total	102	100.0
Pregnancies resulted in a baby that was born dead?	0	96	94.1
	1	6	5.9
	Total	102	100.0
Who makes the final decision about where you must give birth?	Husband	10	9.8
	Myself	92	90.2
	Total	102	100.0
Place of most recent delivery?	Home	33	32.4
	health center	67	65.7
	Hospital	2	2.0
	Total	102	100.0
If it is home, why?	No transportation	25	71.4
	Precipitated labor	7	20.0
	Lack of money	1	2.9
	Did not want to go to health facility	2	5.7
	Total	35	100

Table 2 briefly illustrates the obstetric history of the reproductive women who were attending antenatal healthcare. In terms of the number of times women were pregnant, many of the women were pregnant at least three times covering about 30 (29.4%) of the total respondents, followed by 28 (27.5%) and 26 (25.5%) of the women were being pregnant for one time and two times respectively. Overall, about 70.6% of the total women were pregnant at most two times or a maximum of two times. Moreover, many reproductive women, about 37 (36.3%), had more than three children, followed by 30 (29.4%) of the sampled reproductive women who had only one child. Overall, about 63.7% of the women had children of three and below.

In terms of the average number of years between each delivery, many of the women reported that about 69 (67.6%) of them had on average 2 years of delivery intervals or gaps followed by 32 (29.4%) of them had only 1-year delivery intervals. In total, about 32.4% of them had less than or equal to 1-year delivering interval in between. Of most of the pregnant women at health centers, 79 (77.5%) of the total reported that their first delivery was at the age between 16–25 years old, followed by 21 (20.6%) of them giving their first birth at the age less than 15 years old. In total, about 98% of the pregnant women at health centers reported that their first delivery was at age less than or equal to 25 years old.

This study clearly showed that the women around eastern Tigray give birth at an early age. The most reproductive women who were attending antenatal healthcare at nearby health facility center had reported that about 68 (66.7%) of the total had visited the health facility center for ANC follow-up at a gestational age of 4–6 months followed by 32 (31.4%) of them visited the health center at a gestational age of 1–3 months. About 2% of the mothers did not remember their first ANC visit. In total, about 98% of the pregnant women reported that their gestational age was less than or equal to 6 months when they visited the health facility center for the first time for their ANC follow-up.

Moreover, about 48 (47.1%) of the respondents had more than 2 live births. Among total births, 6 (5.9%) of the babies were born dead. The decision on where to give birth was made independently by 90.2% of women (n = 92), and by their husbands for the remaining 9.8% (n = 10). Furthermore, 67 (65.6%) of the mothers gave birth at health centers, 33 (32.4%) gave birth at home, and the remaining 2% at the hospital. The reason for home delivery was mostly lack of transportation 25 (71.4%), and the second most reported was precipitated labor 7 (20%).

General descriptions of knowledge of obstetric danger signs

Knowledge of reproductive women toward obstetric danger signs during pregnancy was measured through the information levels of the women. First, the reproductive women attending antenatal healthcare at health centers were asked to answer questions regarding their information about obstetric danger signs, and then, they were asked to mention the common obstetric danger signs during pregnancy. Further details are displayed in Table 3 of Supplemental Material.

Table 3 illustrates the frequency distributions of danger signs during pregnancy listed by the reproductive women attending ANC at the health center. About 71 (77.17%) of the reproductive women reported that bleeding at any time during pregnancy was a key obstetric danger sign. Similarly, about 64 (69.57%) of them reported leaking of fluid from the birth canal as a key obstetric danger sign during pregnancy. The third highest number of respondents, 23 (25%), reported swelling of the body as a key obstetric danger sign during pregnancy. For further details, see Table 3 and also refer to the details in Table 3 of Supplemental Material.

Table 3.

Frequency distributions of danger signs during pregnancy.

Name the danger signs you know during pregnancy	Frequency	Percent
1. Bleeding at any time during pregnancy	71	77.17
2. Leaking of fluid from the birth canal	64	69.57
3. Swelling of the body.	23	25
4. Persistent vomiting especially from the fourth month of pregnancy onwards	9	9.78
5. Persistent headache or blurred vision	13	14.13
6. Absent or decreased fetal movements	7	7.61
7. Unusual abdominal pain	1	1.1

The pie chart in Figure 3 illustrates the knowledge level of pregnant women about obstetric danger signs. More than half, 58 (56.86%), of the pregnant women at the health center had poor knowledge of obstetric danger signs. The remaining 44 (43.14%) of the pregnant women had good knowledge of obstetric danger signs.

Figure 3.

Pie chart for knowledge level of pregnant women of obstetric danger signs.

The descriptive analysis summarized in Table 1 of Supplemental Material reveals stark disparities in correct knowledge of obstetric danger signs across subgroups. Knowledge about obstetric danger signs was prominently higher among respondents with higher education (Levels 2–4: 82.8%–91.7% vs. Level 1: 5.6%), urban residents (92.3% vs. rural: 12.7%), and those with the shortest travel time to a health facility (93.5%). A strong positive gradient was also observed with parity, where knowledge about obstetric danger signs increased from 5.6% in primiparous women to 80.0% in women with parity ⩾4.

Test of independence or associations using chi-square test

Table 4 illustrates the association between knowledge of danger signs and other attributes (factors), namely, age, education, parity, residence, distance from health center, and occupation. Based on the results of chi-square tests, the formal tests are written in the following paragraphs.

Table 4.

Test of association (chi-square tests) between attributes and knowledge about danger signs during pregnancy.

Chi-square tests for knowledge level of pregnant women versus attributes
Attributes (factors)	Knowledge level of pregnant women
Attributes (factors)	Value	df	Asymp. Sig. (2-sided)
Age
Pearson chi-square	2.337	5	0.801
Likelihood ratio	2.367	5	0.796
Education
Pearson chi-square	42.573	3	0.000
Likelihood ratio	48.012	3	0.000
Parity
Pearson chi-square	42.573	3	0.000
Likelihood ratio	48.012	3	0.000
Residence
Pearson chi-square	62.237	1	0.000
Likelihood ratio	70.364	1	0.000
Distance from health center
Pearson chi-square	55.169	2	0.000
Likelihood ratio	65.636	2	0.000
Occupation
Pearson chi-square	1.916	3	0.590
Likelihood ratio	1.960	3	0.581
N of valid cases	102

Null and alternative hypotheses

H₀: There is no association between knowledge of obstetric danger signs and the age of women.

H₁: There is an association between knowledge of obstetric danger signs and the age of women.

According to the results of the chi-square test displayed in Table 4, the null hypothesis was not rejected; since the p-value for the chi-square test is greater than the level of significance ( $α = 0.05$ ), there is no significant association between knowledge of danger signs and age of the women (p-value = 0.801) at 5% level of significance. This result indicated that age does not affect being knowledgeable about the obstetric danger signs during pregnancy.

Null and alternative hypotheses

H₀: There is no association between knowledge of obstetric danger signs and parity.

H₁: There is an association between knowledge of obstetric danger signs and parity.

Chi-square results in Table 4 also show that the number of times being pregnant (parity) has a significant association with knowledge of obstetric danger signs during pregnancy (p-value < 0.001) by rejecting the null hypothesis in favor of the alternative hypothesis.

Null and alternative hypotheses

H₀: There is no association between knowledge of obstetric danger signs and residence.

H₁: There is an association between knowledge of obstetric danger signs and residence.

Chi-square results in Table 4 further show that residence has a significant association with knowledge of obstetric danger signs during pregnancy (p-value < 0.001) by rejecting the null hypothesis in favor of the alternative hypothesis. This indicates that living in rural and urban areas has a significant effect on being knowledgeable about obstetric danger signs during pregnancy.

Null and alternative hypotheses

H₀: There is no association between knowledge of obstetric danger signs and distance from the health centers.

H₁: There is an association between knowledge of obstetric danger signs and distance from health centers.

The chi-square test result revealed that the association between knowledge of obstetric danger signs and distance from the health center was significant. In this sense, the distance from the health centers has a significant effect on the knowledge of obstetric danger signs (p-value < 0.001).

Null and alternative hypotheses

H₀: There is no association between knowledge of obstetric danger signs and occupation of women.

H₁: There is an association between knowledge of obstetric danger signs and the occupation of women.

However, the occupation of the women (p-value = 0.590) has no significant association with the knowledge of danger signs during pregnancy. Therefore, knowledge of obstetric danger signs during pregnancy does not depend on the occupation of women. To show this, the results of binary logistic regressions are summarized in Table 5.

Table 5.

Binary logistic regression for knowledge of women toward obstetric danger signs during pregnancy.

Variables in the equation
	B	SE	Wald	df	Sig.	Exp(B)	95%CI for EXP(B)
	B	SE	Wald	df	Sig.	Exp(B)	Lower	Upper
Edu (Read & write)	4.402	0.771	32.586	1	0.000	81.600	18.002	369.878
Edu (Elementary school)	5.231	1.202	18.952	1	0.000	187.000	17.743	1970.824
Edu (Secondary & above)	4.625	1.233	14.076	1	0.000	102.000	9.105	1142.606
Constant	−2.833	0.594	22.743	1	0.000	0.059
Parity (one time)	0.713	1.197	0.355	1	0.551	2.040	0.195	21.295
Parity (two-time)	3.303	1.105	8.934	1	0.003	27.200	3.118	237.280
Parity (three & above)	4.220	1.126	14.051	1	0.000	68.000	7.487	617.575
Constant	−2.833	1.029	7.581	1	0.006	0.059
Residence (Urban)	4.413	0.710	38.614	1	0.000	82.500	20.511	331.835
Constant	−1.928	0.378	25.958	1	0.000	0.145
Distance (<1 h)	5.592	1.030	29.458	1	0.000	268.250	35.608	2020.828
Distance (1–2 h)	2.538	0.810	9.813	1	0.002	12.658	2.586	61.957
Constant	−2.918	0.726	16.154	1	0.000	0.054

Simple binary logistic regression for knowledge of obstetric danger signs

The simple binary logistic regression was applied to predict the dichotomous outcome, namely, the knowledge level of reproductive women on the associated factors. The results are summarized in Table 6.

Table 6.

Firth’s penalized logistic regression estimates for predictors of correct knowledge.

Predictor	Coefficient (β)	SE	p-value	OR	95%CI
(Intercept)	−4.14	2.19	0.058	0.016	(0.000, 1.15)
Education: Level 2	2.61	1.17	0.027	13.57	(1.33, 653.26)
Education: Level 3	2.02	1.57	0.166	7.53	(0.46, 6843.25)
Education: Level 4	2.97	1.88	0.096	19.45	(0.64, 1.15e + 05)
Parity: 2	2.53	2.71	0.581	12.59	(0.01, 6.25e + 05)
Parity: 3	4.31	2.36	0.058	74.41	(0.91, 5.32e + 05)
Parity: 4	3.93	2.35	0.164	51.05	(0.33, 3.05e + 05)
Residence: Urban	3.31	1.17	0.002	27.36	(2.96, 1.92e + 04)
Travel Time: 2	−2.02	1.34	0.167	0.13	(0.002, 2.18)
Travel Time: 3	−3.70	1.73	0.085	0.03	(0.000, 1.50)

OR: Odds ratio; CI: Confidence interval.

Reference categories: Education = Level 1, Parity = 1, Residence = Rural, Travel Time = 1.

Model: Likelihood ratio χ²(9) = 107.62, p < 0.001, n = 102.

Method: Profile penalized likelihood ratio test for p-values and confidence intervals.

The results of simple binary logistic regressions in Table 5 show that educational status, parity, residence, and distance from health facility in hours are significantly associated with or related to knowledge of women toward obstetric danger signs during pregnancy (p-value < 0.001). However, the knowledge of women toward obstetric danger signs during pregnancy was not significantly associated with the age of women and occupation of women at a 5% level of significance.

The results of Binary logistic regression in Table 5 show that the maternal educational status was found to have a significant association with knowledge of women toward obstetric danger signs during pregnancy time. For instance, mothers who can read and write were 81.6 times more likely to have good knowledge than illiterate mothers (OR = 81.6; 95%CI: 18.002, 369.878). Similarly, the mothers who attended elementary school (OR = 187; 95%CI: 17.743, 1970.82) and secondary school and above (OR = 102; 95%CI: 9.105, 1142.606) were 187 times and 102 times more likely to have good knowledge than illiterate mothers, respectively.

Furthermore, mothers who gave birth once were 2.04 times more likely to have good knowledge than mothers who did not give birth before (OR = 2.04, 95%CI: 0.195, 21.295). However, the effect was not significant as the 95%CI for OR includes one in the interval and the p-value for Wald’s test (Wald’s Test = 0.355, p-value = 0.551) was greater than a 5% level of significance. Similarly, mothers who gave birth twice were 27.2 times more likely to have good knowledge than mothers who did not give birth before (OR = 27.2, 95%CI: 3.118, 237.28). Likewise, the mothers who gave birth three times or more were 68 times more likely to have good knowledge compared to mothers who did not give birth before (OR = 68, 95%CI: 7.487, 617.575). The result sought as mothers give birth as many times as possible, they experienced good knowledge of obstetric danger signs during pregnancy time. To this generalization, when mothers gave birth twice and above, they significantly experienced good knowledge about obstetric danger signs during pregnancy time.

Moreover, the mothers who live in urban areas were 82.5 times more likely to have good knowledge of obstetric danger signs during pregnancy than the mothers who live in rural areas (OR = 82.5; 95%CI: 20.511, 331.835). Similarly, mothers who travel long distances are more likely to have poorer knowledge than those who travel short distances. For instance, the mothers who travel less than 1 h were 268.25 times more likely to have good knowledge compared to mothers who travel more than 2 h. Similarly, the mothers who travel 1 up to 2 h were more likely to have good knowledge than the women who travel more than 2 h on foot. However, the standard simple binary logistic regression results were unreliable, exhibiting signs of overfitting, multicollinearity, and separation. Therefore, more robust analytical methods, including penalized logistic regression, regularizations, and Bayesian logistic regression, were required for result and conclusion stability.

The contingency tables of Table 2 in Supplemental Material confirmed the patterns observed in the descriptive statistics, detailing the raw cell counts. The summarized data revealed extreme disparities, with most categories covering cells with small counts (<5). Remarkably, the relationship appears quasi-complete for Residence (only three pregnant women from urban respondents lacked knowledge about obstetric danger signs) and Travel Time (only two pregnant women respondents with the shortest travel time lacked knowledge about obstetric danger signs, while only two with the longest travel time had knowledge about obstetric danger signs). These small cell counts disrupt the assumptions of large-sample tests like the Pearson’s chi-square and may lead to untrustworthy p-values. Hence, analytical strategies applying exact statistical methods (e.g. Fisher’s Exact Test) or penalized regression are required for valid inference. The patterns strongly suggest that these variables are potent predictors of knowledge status about obstetric danger signs.

Table 4 in Supplemental Material confirmed that the standard maximum likelihood of multiple logistic regression model failed, as indicated by coefficients of extreme magnitude, tremendously large standard errors, non-significant p-values (all ~1.0), and OR that are either effectively infinite or zero with infinite 95%CIs. This is a typical sign of complete or quasi-complete separation, a scenario where one or more predictor variables perfectly or near-perfectly predict the outcome as suggested by the contingency tables of Table 2 in Supplemental Material. The algorithm fails to converge to a finite solution, interpreting these findings invalid for inference or generalization. This demands the use of a penalized logistic regression approach, such as Firth’s bias-reduced logistic regression, to offer stable, interpretable parameter estimates. Therefore, the result of Firth’s penalized binary logistic regression estimates for knowledge of pregnant women about obstetric danger signs is displayed in Table 6.

Table 6 confirms that the Firth’s penalized binary logistic regression model successfully converged and notorious key factors of knowledge of pregnant women about obstetric danger signs after regulating for all other variables. The urban residence was a strong, statistically significant predictor (OR = 27.4, 95%CI: 3.0, 1.9e + 04, p = 0.002) of knowledge of pregnant women about obstetric danger signs. Similarly, the education Level 2 was also significant (OR = 13.6, 95%CI: 1.3, 653.3, p = 0.027), while higher education levels showed strong but non-significant associations with knowledge of pregnant women about obstetric danger signs. Additionally, the longer travel times were associated with substantially lower odds of knowledge of pregnant women about obstetric danger signs, even though the associations were marginally non-significant (p = 0.085 for Travel Time 3). Moreover, the parity showed a positive gradient but was not statistically significant in the adjusted model. The firth’s penalized binary logistic regression model demonstrated excellent overall fit (p < 0.001). The comparison of regularization methods are displayed in Table 7.

Table 7.

Comparison of cross-validated error and model complexity for regularized regression methods.

Method	Lambda (λ)	Cross-validation error	Non-zero coefficients
Ridge regression	0.0387	0.2951	10
Least Absolute Shrinkage and Selection Operator (LASSO)	0.0103	0.3254	9
Elastic net	0.0074	0.2378	10

Table 7 briefly shows the results of regularization regression methods to further address multicollinearity and potential overfitting; a comparison of penalized regression approaches was conducted via cross-validation. The Elastic Net model attained the lowermost prediction error (CV Error = 0.238), outperforming both Ridge (0.295) and Least Absolute Shrinkage and Selection Operator (LASSO) (0.325) regression. All coefficients persevered non-zero in the Ridge and Elastic Net regression models, whereas LASSO selected a model with nine non-zero coefficients, signifying variable selection.

The model comparison metrics in Table 5 of Supplemental Material, confirmed the failure of the standard binary logistic regression and the superiority of the penalized binary logistic regression method. The standard model reveals a positive Akaike information criterion (AIC = 26.5), whereas Firth’s model produces a noticeably lower, negative AIC (−33.8). Similarly, the Bayesian information criterion (BIC) is lower for the penalized binary logistic regression model (−7.6 vs. 52.7), further supporting its selection. These findings are further supported by the results of the Bayesian binary logistic regression, as displayed in Table 8.

Table 8.

Bayesian logistic regression results for predictors of correct knowledge (Posterior OR).

Predictor	Median OR	95%Credible Interval
(Intercept)	0.05	(0.002, 0.70)
Education: Level 2	15.32	(1.94, 144.44)
Education: Level 3	30.48	(1.97, 832.21)
Education: Level 4	8.52	(0.60, 102.40)
Parity: 2	0.44	(0.02, 6.61)
Parity: 3	13.80	(1.43, 252.31)
Parity: 4	9.88	(0.75, 118.70)
Residence: Urban	78.24	(9.65, 1042.41)
Travel time: 2	0.26	(0.03, 2.20)
Travel time: 3	0.03	(0.002, 0.34)

OR: Odds ratio.

Reference categories: Education = Level 1, Parity = 1, Residence = Rural, Travel Time = 1.

= Posterior estimates based on weakly informative priors. Model fit not displayed (Mean Posterior Predictive Distribution (mean_PPD = 0)).

The findings from the Bayesian binary logistic regression analysis in Table 8 validate and strengthen findings from the penalized likelihood model, providing a direct probabilistic explanation. The urban residence appears as the solidest predictor, with a median OR of 78.2 (95%Credible Interval (CrI): 9.7, 1042.4). Likewise, the higher education levels (2 & 3) and higher parity (3) show strong positive relations, with CrIs excluding 1 for education Level 2 and parity 3. Greatest conclusively, the longest travel time (Level 3) is related with a 97% reduction in the odds of knowledge of pregnant women about obstetric danger signs (OR = 0.03, 95%CrI: 0.002, 0.34). This interval excludes 1, providing vigorous evidence of a significant negative effect.

Discussions

This section dealt with discussions of the findings by comparing them with previous research works. Thus, this finding showed that many of the reproductive women who have been attending ANC at health centers were married 96 (94.1%). This was consistent with the findings in Dire Dawa Administrative Public Health facilities and in Debre Birhan city administration north Shewa Amhara region Ethiopia in which 491 (97.8%) and 563 (89.1%) of the reproductive women were married, respectively.^19,20 In terms of occupation, this finding briefly showed that the majority 71 (69.6%) of the reproductive women were housewives; it was similar to the study done in Dire Dawa and Debre Birhan in which about 330 (65.7%) and 413 (65.3%) of the reproductive women were housewives.^19,20 The majority, 63 (61.76%), of the reproductive women included in this study were rural residents, which was higher than the finding in Dire Dawa, which was 38%.¹⁹

This finding has shown that only about 31 (30.39%) of the pregnant women were traveling less than 1 h to get health facilities, which was smaller than the finding in Dire Dawa administrative public health facilities in 2017, which was 415 (82.5%).¹⁹ This might be due to the sample size taken due to the topographical difference or the area where the study was done.

This study showed that, in total, 71.6% of the pregnant women at the health centers in eastern Tigray mentioned vaginal bleeding at any time during pregnancy, which was higher than the finding in Harar, Ethiopia, which was 29.1%. However, this finding was almost similar to the findings in Angolela Tera District, Northern Ethiopia, which was 72.6%.⁴

In this study, 71 (77.2%) and 64 (69.6%) of the women mentioned vaginal bleeding and leaking of fluid from the birth canal, respectively, as key obstetric danger signs during pregnancy, which were smaller than the finding in Arba Minch town governmental institutions, South Ethiopia in 2016, which was 98.37% and 72.65%, respectively.⁸

According to this study, only 9.8% of the mothers were unable to mention an obstetric danger sign, which is less than the study done in Harar, Ethiopia, which was 30%.¹¹

To this end, among the women who reported that they had information about obstetric danger signs during pregnancy, those who could mention at least three obstetric danger signs were considered to have good knowledge. A small dominant number of pregnant women did not have good knowledge of obstetric danger signs during pregnancy. The most profound disparities were associated with residence and travel time, suggesting that access-related factors may be critical elements of knowledge in this population. These results justify further analysis to isolate the independent effects of these variables.

From the chi-square test of independence, age was not significantly associated with knowledge of obstetric danger signs (p-value = 0.801), which is consistent with the finding of the study in Bureti Sub-County of Kericho County, Kenya in 2015,¹⁵ and this finding was not similar to the finding of the study done in Illu Ababor, in which age was significantly associated with the knowledge of obstetric danger signs during pregnancy.¹⁴

In general, the chi-square test measures the association between two categorical factors without showing the direction and strength of the association. In this sense, the chi-square test cannot tell us the strength and direction of association between two categorical factors. However, it can show only the existence of a significant association between two categorical attributes. To overcome this issue, binary logistic regression can be used for dichotomous outcomes. Binary logistic regression is used to predict the likelihood of success of the dichotomous outcome conditional to different factors.

Therefore, these findings showed that knowledge of women toward obstetric danger signs was strongly dependent on educational status, parity, residence, and distance from the health facility in hours. However, it did not depend on age and occupation. Knowledge of reproductive women toward obstetric danger signs was higher among women who attended formal education as compared to illiterate women. In general, the findings showed that an increased maternal educational status was associated with increased women’s knowledge of obstetric danger signs during pregnancy.

The results indicated that mothers who gave birth as many times as possible experienced good knowledge of obstetric danger signs during pregnancy time. To this generalization, when mothers gave birth twice and above, they significantly experienced good knowledge about obstetric danger signs during pregnancy time.

The finding from simple binary logistic regression briefly showed that mothers who can read and write were 81.6 times more likely to have good knowledge than illiterate mothers (OR = 81.6; 95%CI: 18.002, 369.878). Similarly, the mothers who attended elementary school (OR = 187; 95%CI: 17.743, 1970.82) and secondary school and above (OR = 102; 95%CI: 9.105, 1142.606) were 187 times and 102 times more likely to have better knowledge than illiterate mothers, respectively. This finding is in line with the finding from the study that was done in Gedo Town Health Facilities in 2015 and in Angolela Tera District, Northern Ethiopia in 2019 in which the respondents who attended above grade 12 were about four times more likely knowledgeable about obstetrics danger signs than those respondents who were illiterate in education status (AOR = 4.421, 95%CI: 2.102, 9.743)¹³ and attending formal education (AOR: 4.01; 95%CI: 2.35, 6.75; p-value < 0.001).⁴

Furthermore, in terms of parity, the mothers who gave birth twice were 27.2 times more likely to have good knowledge than mothers who did not give birth before (OR = 27.2, 95%CI: 3.118, 237.28). Likewise, the mothers who gave birth three times or more were 68 times more likely to have good knowledge compared to mothers who did not give birth before (OR = 68, 95%CI: 7.487, 617.575). The result sought as mothers give birth as many times as possible, they experienced good knowledge of obstetric danger signs during pregnancy time. To this generalization, when mothers gave birth twice and above, they significantly experienced good knowledge about obstetric danger signs during pregnancy time. Similarly, the mothers who live in urban areas were 82.5 times more likely to have good knowledge of obstetric danger signs during pregnancy than the mothers who live in rural areas (OR = 82.5; 95%CI: 20.511, 331.835). Likewise, mothers who travel long distances are more likely to have poor knowledge than those who travel short distances. For instance, the mothers who travel less than 1 h were 268.25 times more likely to have good knowledge compared to mothers who travel more than 2 h. Moreover, the mothers who travel 1 up to 2 h were more likely to have good knowledge than the women who travel more than 2 h on foot. These findings were consistent with the findings of the study that was done in Angolela Tera District, Northern Ethiopia in 2019 which were urban residences (AOR: 2.01; 95%CI: 1.02, 5.65; p-value = 0.013), taken <20 min to reach health facility (AOR: 5.01; 95%CI: 2.76, 10.18; p-value < 0.001), ⩾2 history of pregnancies (AOR: 2.2; 95%CI: 1.2, 4.9; p-value = 0.04).⁴

To this end, knowledge of reproductive women toward obstetric danger signs was higher among women who attended formal education as compared to illiterate women. In general, the findings showed that an increased maternal educational status was associated with increased women’s knowledge of obstetric danger signs during pregnancy. This is like the findings of the recent study done in Angolela Tera District, Northern Ethiopia, in 2019, in which an increased maternal educational status was associated with increased women’s knowledge of obstetric danger signs during pregnancy.⁴

The research done in the Raya Kobo district of Ethiopia revealed that the mothers’ education was found to be significantly associated with knowledge of key danger signs during pregnancy and the postpartum period. Pregnant mothers who attended secondary education were more likely to be knowledgeable about danger signs during pregnancy than their illiterate counterparts (AOR: 3.63; 95%CI: 1.19, 11.07),¹⁵ which was in line with our finding.

As Dereje et al.¹⁴ stated, the reproductive mothers with higher education were 1.46 times and 1.24 times more likely to know obstetric danger signs during pregnancy childbirth than those who cannot read and write and with primary education (AOR = 1.46; 95%CI: 1.24, 1.91), respectively. Similarly, the pregnant mothers of around Illu Ababor who attended secondary education were 2.46 times more likely to know obstetric danger signs during postpartum than their counterparts (AOR = 2.36; 95%CI: 2.18, 2.72), which is consistent with our findings.

Overall, this survey study pinpoints thoughtful structural disparities as the major drivers of obstetric danger sign knowledge disparity of pregnant women in a postwar Eastern Tigray setting of Ethiopia. The stark urban–rural and access-based divides embodied by the solid, independent effects of residence and travel time recommend that maternal health literacy is fundamentally a function of geography and infrastructure. Whereas education remains a significant social determining factor, its effect appears secondary in this framework, potentially mediated by severe spatial barriers. These results underline that in fragile, postwar regions, information imbalances are not merely knowledge gaps but exhibitions of deeper, systemic access shortfalls.

Meticulously, the analysis climaxes a critical challenge in disparity research. The thorough failure of standard binary logistic regression due to quasi-complete separation, resolved merely through penalized binary logistic regression and Bayesian binary logistic regression approaches, elucidates how conventional tools can mislead when analyzing extreme imbalances. The convergence of results from Firth’s binary logistic regression and Bayesian binary logistic regression and their inference provides vigorous evidence that improving knowledge of pregnant women about obstetric danger signs will require more than conventional health education; it demands targeted, equity-focused strategies that explicitly bridge spatial and informational exclusion for the hardest-to-reach rural populations.

This huge discrepancy in model comparison’s AIC and BIC values is not typical of well-behaved models; the negative AIC for the penalized binary logistic regression model specifies an extremely good fit to the data after adjusting for separation; however, the positive AIC for the standard model imitates its pathological failure to converge. This comparison formally justifies the elimination of the standard maximum likelihood estimates and the use of Firth’s penalized binary logistic regression for valid conclusion in the presence of complete or quasi-complete separation.

The findings from this Firth’s penalized binary logistic regression model confirmed that access (residence and travel time) and education are primary correlates of knowledge of pregnant women about obstetric danger signs, with the model successfully overcoming the separation issues existing in the standard logistic regression.

The superior performance of the Elastic Net, which blends L1 (LASSO) and L2 (Ridge) penalties, proposes that it offers the optimal bias-variance trade-off for this dataset and is the chosen technique for obtaining final, stable coefficient estimates.

The findings of Bayesian logistic regression affirm that structural access factors (residence and travel time) and education are fundamental determinants of knowledge of pregnant women about obstetric danger signs in this population, with Bayesian implication measuring the extensive uncertainty about these large effect sizes in a small, separated dataset.

This health facility institution-based survey study has some limitations. First, the availability of literature and validated instruments specific to postwar settings for assessing ANC attendance and obstetric danger sign knowledge is limited. Second limitation of this study is that the questionnaire was not pilot-tested prior to data collection. Although internal consistency reliability was assessed using Cronbach’s alpha and found to be high (α = 0.859), future studies should conduct pretesting and formal validation procedures before field implementation. Finally, the use of purposive sampling weighted by ANC attendance may introduce selection bias and affect the proportionality of the sample.

Conclusions and recommendations

Conclusions

Based on the goal of this research, the findings came up with a few important and relevant conclusions.

This study was an attempt to examine the association between the knowledge of reproductive women toward obstetric danger signs and associated factors among mothers attending ANC at health centers in eastern Tigray. A total of 102 reproductive women were selected from the target population of reproductive women in eastern Tigray. After reviewing relevant literature from previous works, the knowledge level of reproductive women toward obstetric danger signs during pregnancy was considered as a dependent variable or outcome variable.

Then, the dichotomous outcome variable has two levels: “good knowledge as success” and “poor knowledge as failure.” The chi-square was applied to examine the association of the knowledge of pregnant women toward obstetric danger signs during their pregnancy time with some associated factors. This chi-square test for independence can only check the existence of association without showing the strength and direction of association between knowledge of pregnant women toward obstetric danger signs and associated factors. To overcome this, the binary logistic regression was applied to examine the strength and direction of the relationship between the knowledge of women toward obstetric danger signs during pregnancy and associated factors.

Based on the chi-square test of independence, the knowledge of pregnant women about obstetric danger signs during pregnancy was significantly associated with the educational status of the women, number of times pregnant (parity), residence, and distance from the health centers. However, there was no significant association between the knowledge of pregnant women toward obstetric danger signs during pregnancy and the age of the women as well as the occupation of the women at a 5% level of significance. This means the age and occupation of the women did not matter; the mothers had good knowledge or poor knowledge of obstetric danger signs during their pregnancy time.

The simple binary logistic regression showed a strong positive relationship between knowledge of pregnant women toward obstetric danger signs during pregnancy time and educational status, parity, residence, and distance from the health facility centers in hours relative to the respective reference levels (their counterparts). This means the knowledge of the reproductive women are more likely good among women attending formal educations compared to illiterate women. In this sense, as women’s education level increased, the knowledge of women toward obstetric danger signs also increased. Thus, this finding briefly showed that an increase in maternal educational status was associated with increased women’s knowledge of obstetric danger signs during pregnancy. This might be a result of the fact that education can provide appropriate information about pregnancy, and thus, educated women are better informed and make better choices than illiterate ones.

Similarly, knowledge of pregnant women regarding obstetric danger signs during pregnancy among women who have given a higher number of births is more likely good relative to the women who did not give birth before. This means that as mothers are experienced in giving birth, they increase their knowledge of obstetric danger signs during pregnancy. When women are exposed to previous pregnancies, they can get more information from health professionals due to their previous exposure to health institutions. Moreover, the women who live in urban areas are more likely to have good knowledge of obstetric danger signs during pregnancy than the women who live in rural areas. This could be because urban residents are exposed to better healthcare services and they have better access to relevant health information as compared to rural residents.

Furthermore, the women who travel a short distance to reach the health facility center are more likely to have good knowledge than the women who travel a long distance to reach the health facility center on foot. Therefore, the distance from the health facility center significantly affects the mothers’ knowledge of obstetric danger signs during their pregnancy. This means the women who travel a short distance to get to the health facility from the center are more advantageous in gaining knowledge and awareness toward obstetric danger signs as they get advice and guides from health experts, midwiferies, and physicians. This might be due to easy accessibility of different health facilities and better access to health information for respondents who lived near to health facilities as compared to those living in distant areas.

To this end, the cross-sectional survey reveals that disparity in knowledge of pregnant women about obstetric danger signs in post-conflict Eastern zone of Tigray, Ethiopia, is not random; however, it is scientifically structured by insightful disparities in access and education. Employing vigorous statistical methods to overcome analytical challenges, we found that being a rural resident and living far from a health facility center are the furthermost powerful barriers to knowledge of pregnant women, meritoriously excluding a substantial segment of the population from critical maternal health information. The educational level further provides a modifying influence; however, it does not overcome these ultimate spatial barriers. In the context of post-conflict recovery, these findings point out that equitable improvement in maternal health literacy will persist vague unless the core structural contributing factors of access are prioritized. Overall, the results can generalize that the knowledge of pregnant women about obstetric danger signs during their pregnancy time was strongly dependent on educational status, parity, residence, and distance from a health facility in hours while traveling on foot. However, it did not depend on age and occupation.

Recommendations

Regarding the findings and conclusions, this study set out to recommend certain points that are very critical if considered and implemented by health experts, midwives, physicians, local, national, and world health organizations, the government, and other concerned bodies properly. Therefore, the following recommendations have been forwarded.

As was clearly shown in the findings, discussions, and conclusions, the knowledge of the reproductive women toward obstetric danger signs during pregnancy was significantly associated with education, parity, residence, and the distance from the health facility center. Therefore, the government should provide a special opportunity for education for women by creating an opportunistic environment for girls.

Besides, the local, national, and world health organizations should work hard and set plans and strategies to enhance the knowledge of reproductive women about obstetric danger signs during pregnancy. Furthermore, the government should collaborate with local, national, and world health organizations to enhance the outreach of the health facility center for both rural and urban areas. This, in turn, can avoid the distant travel of the women to reach the health facility center.

On the other hand, health experts, midwives, and physicians should give proper guidance and counseling or advice from the very beginning up to the delivery time. They should let the women know about the obstetric danger signs in a timely and wise manner. They should ethically encourage pregnant women to follow their advice carefully. They should make the women aware of the negative impact of obstetric danger sign and their consequences during pregnancy.

To discourse the structural knowledge disparities of pregnant women about obstetric danger signs, maternal health strategy and policy in post-conflict areas must explicitly prioritize geographic equity. Substantial resources and interventions should be allocated using inverse-care principles, safeguarding that the furthermost marginalized remote populations are reached first. This requires fast-tracking the rehabilitation of rural maternal and community health infrastructure and rightfully authorizing funded outreach programs.

Programmatically, implementation should emphasize decentralized information delivery on obstetric danger signs. Strengthening community and maternal health employee linkages with transportation support and enticements for marginalized remote service is essential. Supplementing this with targeted maternal health campaigns in local languages and exploring community-based transport schemes can bridge the final mile in information and access.

Further investigation should assess and check the feasibility of these targeted interventions. Alongside, applied health research in similar fragile settings should adopt penalized binary logistic regression or Bayesian logistic regression approaches when analyzing separated data to ensure that findings are statistically valid and policy-relevant.

Finally, reproductive women should carefully follow the guides and advice of health experts, midwives, and physicians during antenatal healthcare to improve their knowledge of obstetric danger signs during pregnancy.

Supplemental Material

sj-docx-1-smo-10.1177_20503121261426069 – Supplemental material for Disparities in obstetric danger sign knowledge among pregnant women in postwar Eastern Tigray, Ethiopia: Robust logistic regression

Supplemental material, sj-docx-1-smo-10.1177_20503121261426069 for Disparities in obstetric danger sign knowledge among pregnant women in postwar Eastern Tigray, Ethiopia: Robust logistic regression by Yemane Hailu Fissuh, Yemane Asmelash Gebretensae, Tesfay Hailu Shifarre and Mebrihit Berihu Mesfun in SAGE Open Medicine

Footnotes

Acknowledgements

The authors would like to express heartfelt gratitude to Aksum University, Axum, Tigray, Ethiopia, for its funding grant for data collection.

ORCID iDs

Yemane Hailu Fissuh

Yemane Asmelash Gebretensae

Tesfay Hailu Shifarre

Ethical considerations

The Aksum University Internal Ethical Committee (AKUIEC), namely Institutional Research Ethics Review Committee (IRERC) with approval reference number AKU/IRERC/977/26, after carefully examining the protocol, determined that all required ethical concerns have been adequately addressed and fulfilled. The study did not undertake any human, animal, or environmental harmful concern. Therefore, the IRERC confirmed that the ethical approval was waived for this anonymous, non-experimental survey study. A formal supporting letter and ethical clearance approval letter was issued by the committee, affirming compliance with institutional ethical standards. It was confirmed that the IRERC also approved the dataset for public utilization without personal, household identification.

Consent to participate

The Aksum University Internal Ethical Committee (AKUIEC), namely Institutional Research Ethics Review Committee (IRERC) with approval reference number AKU/IRERC/977/26, confirmed waiver of the participants’ written informed consent due to the study’s voluntary, anonymized, and risk-free questionnaire design. It was also confirmed that the verbal full informed consent from the participants was asked in the designed questionnaire, and the verbal informed consent was obtained from all participants. Why the verbal consent was enough for this study is because the data were collected immediately by asking their consent for each participant. Written consent was waived because it is not applicable for harmless survey study. Therefore, the privacy of the participants was kept anonymous.

Author contributions

F.Y.H. conceptualized the study, developed the research proposal, wrote the research, conceptualized the methodology, obtained the data, analyzed, conceptualized, interpreted, organized, and refined the final research paper. F.Y.H. also prepared the article for publication, checked language and grammar errors, and finally acted as the corresponding author for submission. The corresponding author cross-checked many things, including the author’s consent, conflicts of interest, and research ethics. G.Y.A. conceptualized a proposal, conceptualized the methodology, organized the data set for analysis, edited, and refined it. Furthermore, S.T.H. collected data, organized it, edited it, and made it ready for analysis. M.M.B. conceptualized the research paper and contributed to the editing, refining, writing, and organization of the work.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was funded by an internal grant with grant project code AKUR/CNCSM/006/16 from Funding agency: AKU-CNCS-RCS of Aksum University, Axum, and Tigray, Ethiopia.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability statement

The datasets analyzed in this study are original, first-hand case study data collected by the authors. The data are included in the article, and any additional materials are available upon official request by contacting the first and corresponding author directly via email at yemaoroba@gmail.com or hailuye@mail.aku.edu.et. All inquiries related to the study or data access may also be directed to this address.

Supplemental material

Supplemental material for this article is available online.

References

Nurgi

Tachbele

Dibekulu

, et al. Knowledge, attitude and practice of obstetric danger signs during pregnancy in Debre Berhan, Ethiopia. Health Sci J 2017; 11(6): 533.

Abiyot

Kassa

Buruh

, et al. Awareness of obstetric danger signs and its associated factors among pregnant women in public health institutions, Mekelle City, Tigray, Ethiopia. J Pregnancy Child Health 2015; 2: 167.

Ayele

Catherine

Abdulbasit

, et al. Women’s knowledge of obstetric danger signs in Ethiopia (WOMEN’s KODE): a systematic review and meta-analysis. Syst Rev 2019; 8: 63.

Woldeamanuel

Lemma

Zegeye

Knowledge of obstetric danger signs and its associated factors among pregnant women in Angolela Tera District, Northern Ethiopia. BMC Res Notes 2019; 12: 606.

Ethiopian Demographic and Health survey (EDSH). Maternal mortality. EDSH, 2016.

World Health Organization (WHO). Maternal mortality, 2018. WHO.

Gebrehiwot

Bahta

Haile

Awareness of danger signs of pregnancy and its associated factors among pregnant women who visit ANC in Mekelle Public Hospitals

Am J Adv Drug Delivery 2014; 2(2): 164–173.

Samuel

Genet

Assessment of knowledge on danger sign of pregnancy and associated factors among ANC attendant pregnant women in Arbaminch Town Governmental Institutions, Southern Ethiopia. Ann Med Health Sci Res 2018; 8: 64–69.

Kheamy

Bardisi

SM.

Knowledge and factors associated with obstetric danger signs among pregnant women in Al-Nawariah primary health care center, Makkahalmukarramah, (a cross-sectional study). Int J Adv Res 2017; 5(8): 381–393.

10.

Federal Demographic republic of Ethiopia (FDRE) Ministry of Health (MOH). National reproductive health strategy, 2006–2015. FDRE and MOH.

11.

Masresha

Knowledge attitude and practice of pregnant women to ward danger sign of pregnancy in Harar Towen Eastern Ethiopia. EC Gynaecology 2019; 8(8): 656–662.

12.

Nebiyu

Kifle

and Lamessa Knowledge of obstetric danger signs and associated factors among pregnant women in Erer district, Somali region. Ethiopia Women’s Health 2016; 16: 30.

13.

Tolera

GD.

Knowledge of obstetric danger signs and associated factors among pregnant women attending ANC service at Gedo Town Health Facilities. J Health Med Nurs 2016; 28: 50.

14.

Dereje

Muluneh

Kebebe

, et al. Knowledge of obstetric danger signs and associated factors among pregnant women attending antenatal care at selected health facilities in Illu Ababor zone, Oromia National Regional State, south-west Ethiopia. Int J Nurs Midwifery 2017; 9(3): 22–32.

15.

Omari

Phanice

Masimba

, et al. Knowledge of obstetric danger signs among pregnant women attending antenatal care clinic at health facilities within Bureti Sub-County of Kericho County, Kenya. Res Obstetr Gynecol 2018; 6(1): 16–21.

16.

Bililign

Mulatu

Knowledge of obstetric danger signs and associated factors among reproductive age women in Raya Kobo district of Ethiopia: a community based cross-sectional study. BMC Pregnancy Childbirth 2017; 17: 70.

17.

Hibstu

Siyoum

YD.

Knowledge of obstetric danger signs and associated factors among pregnant women attending antenatal care at health facilities of Yirgacheffe town, Gedeo zone, Southern Ethiopia. Arch Public Health 2017; 75: 35.

18.

Mpembeni

RNM

Kilewo

Leshabari

, et al. Use pattern of maternal health services and determinants of skilled care during delivery in Southern Tanzania implications for achievement of MDG-5. BMC Pregnancy Childbirth 2007; 7: 29.

19.

Abdurashid

Ishaq

Ayele

, et al. Level of awareness on danger signs during pregnancy and associated factors among pregnant mothers, Dire Dawa Administrative Public Health Facilities, Eastern Ethiopia. Clin Mother Child Health 2018; 15: 290.

20.

Nurgi

Assessment of knowledge, attitude and practice of obstetric danger signs during pregnancy among mothers in Debrebirhan city administration, North Shewa, Amhara Region, Ethiopia. Department of Nursing and Midwifery, Allied School of Health Sciences, College of Health Science, Addis Ababa University, 2014.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.09 MB