Spatial variation and predictors of missing birth preparedness and complication readiness (BPCR) messages in Ethiopia

Abstract

Background

The Birth Preparedness and Complication Readiness (BPCR) message is one of the prenatal care packages targeted at reducing maternal and neonatal mortality by avoiding unnecessary delays during labor and delivery. There is limited evidence in Ethiopia that has looked at the spatial variation of missing BPCR messages and potential predictors. Hence, this study aimed to identify spatial predictors missing BPCR messages at the national level.

Methods

The study was based on analysis of 2016 Ethiopia Demographic Health Survey data, using a weighted sample of 4771 women. Arc-GIS version 10.7 and SaTScan version 9.6 statistical software were used for the spatial analysis. To explore spatial variation and locate spatial clusters of missing BPCR messages, the Global Moran’s I statistic and Bernoulli-based spatial scan (SaTScan) analysis were carried out, respectively. Hotspot (Getis-OrdGi*) analysis was conducted to identify Hotspots and Cold spotsof missing BPCR messages. Finally, spatial regression were carried out via ordinary least squares and geographically weighted regression to identify predictors of hotspots for missing BPCR messages.

Results

The overall prevalence of missing BPCR messages in Ethiopia was found to be 44.0% (95%CI: 42.6, 45.4%), with significant spatial variation across regions (Moran’s I = 0.218, p-value<0.001) and seven most likely significant SaTScan clusters. The vast majority of Somali, central Afar, and Gambella regions were identified as statistically significant hotspots. Living in the poorest wealth quintile, having only one ANC visit, lack of access to listening to the radio, facing difficulty in accessing money, not having a mobile phone, and being not covered by health insurance were identified as significant spatial predictors of missing BPCR messages.

Conclusion

The level of missing BPCR messages during pregnancy was found to be high in Ethiopia, with significant local variation. As a result, policymakers at the national level and local planners should develop strategies and initiatives that enhance women’s economic capacities, health-seeking behavior, and media exposure. Furthermore, the regional authorities should focus on strategies that promote universal health coverage through enrolling citizens in health insurance schemes.

Introduction

Despite a decline in the global maternal mortality ratio (MMR) from 339 in 2000 to 223 in 2020, it remains unacceptably high, with nearly 287,000 women dying during and after pregnancy and childbirth [1, 2]. Almost 95% of all maternal deaths occurred in Low and Middle-Income Countries (LMICs) in 2020, and most could have been prevented [3]. Sub-Saharan Africa (SSA) and Southern Asia accounted for approximately 70 percent (202,000) of the estimated global maternal deaths in 2020 [1, 2]. Between 2000 and 2016, Ethiopia reduced maternal death from 871 to 412 per 100,000 live births and child mortality from 166 to 67 per 1,000 live births, still, both remain much too high [4, 5].

Delays in preparing for and responding to potential direct obstetric complications during pregnancy and childbirth by pregnant women, families, and healthcare providers have been recognized as the key barrier to reducing maternal morbidity and mortality in SSA [6]. Thus, interventions such as the Birth preparedness and complication readiness (BPCR) package are vital for addressing such delays in developing countries like Ethiopia, where obstetric services are often limited and/or underutilized [7].

BPCR is one of the WHO recommendations for newborn and maternal health promotion activities that are delivered and executed through an antenatal care (ANC) service [8]. It is defined as a comprehensive strategy to increase the adoption and effectiveness of essential maternal and newborn health care by reducing three delays in accessing basic maternal health services (delays in deciding, reaching, and receiving care) [9]. It has a set of indicators that were intended to be implemented by pregnant women, families(husband/partner), communities, health facilities, and health care providers [10]. Those items of care that are intended to be implemented by all pregnant women are: deciding on a birthplace, maintaining the necessary supplies for childbirth, figuring out or establishing emergency transportation, setting aside money or an emergency fund, identifying blood donor/s, determining companions during labor and childbirth which are vital in easing access to care [10, 11].

Despite having a considerable impact on maternal mortality reduction, BPCR messages were not widely implemented in Ethiopia. A systematic review and meta-analysis conducted in Ethiopia revealed that BPCR uptake was 25.2% [11]. On the other hand, a recent study from nationally representative data showed that more than half (56.02%) of women received at least one BPCR message [12]. Small-scale studies conducted in the country showed that there was a positive association between the receipt of BPCR messages and educational status, occupation, being in a higher wealth quintiles, receipt of ANC visits, previous history of health facility delivery, history of adverse birth outcomes, number of living children (parity), knowledge of key danger signs during pregnancy and postpartum, and male partner involvement in BPCR plan [13–18].

Despite numerous studies on the uptake of BPCR messages across the country, there was a dearth of evidence that has investigated the spatial distribution that shows regions with poor uptake of BPCR messages. Hence, the current study carried out both spatial analytic techniques (a hot spot and a geographically weighted regression (GWR)). First, a hotspot analysis was carried out to identify the regions and areas with low BPCR message adoption, and the study will next delve into identifying potential variables that lead to regional differences in poor service uptake by employing a GWR. Conducting a hotspot analysis could offer a data-driven approach to understanding and addressing the challenges hindering the dissemination and adoption of key BPCR messages. By identifying specific geographic regions with low BPCR message adoption, resources and interventions can be allocated to areas of highest need and groups that are especially vulnerable to maternal complications of pregnancy. The findings from the current study will allow program planners, policymakers, and healthcare providers to customize their efforts to address the factors behind poor service uptake in each hotspot, thereby increasing the likelihood of successful intervention. By targeting interventions in hotspots with low BPCR message uptake, maternal and neonatal health outcomes can be significantly improved via working on safer pregnancies, and timely access to healthcare services.

Methods and materials

Study design and setting

The current study was based on the 2016 Ethiopian Demographic and Health Survey (EDHS) which was conducted by the Central Statistics Agency (CSA) from January 18, 2016, to June 27, 2016 [5]. Ethiopia is located in the Horn of Africa (3°–15°N latitude and 33°–48°E longitude). The country is divided into nine administrative regions (Tigray, Afar, Amhara, Oromia, Somali, Benishangul-Gumuz, Southern Nation Nationality and People’s Region (SNNPR), Gambella, and Harari) and two self-administrative cities (Addis Ababa and Dire Dawa) [5]. A secondary analysis of the women’s (IR) file of 2016 EDHS was carried out.

Populations

The source populations were all women aged 15 to 49 who gave birth within the five years preceding the survey, whereas the study populations were women who had complete information on the uptake of BPCR messages during their ANC visit. The EDHS has numerous datasets, for males, women, children, birth, and households. The women’s (IR file) was used for the current study, and a total weighted sample of 4,771 women was considered in the analysis.

Sampling procedures and data collection tool

The Ethiopia Population and Housing Census (PHC), carried out in 2007 by the Ethiopia CSA, served as the sampling frame for the 2016 EDHS. The samples were chosen using a stratified, two-stage cluster sampling procedure in which data were hierarchical. In the first stage, 645 enumeration areas (EAs) were selected (202 in urban and 443 in rural areas), with a probability proportional to EA size and independent selection in each sampling stratum. In each of the selected EAs, a household listing operation was carried out, and the consequent lists of households served as a sampling frame for the selection of households in the second stage. In the second stage, a fixed number of 28 houses per cluster were chosen with equal probability from the sample frame [5].

The Woman’s Questionnaire, which includes socio-demographic and economic information, obstetric characteristics, and maternal health service utilization, was used to collect data from all eligible women aged 15 to 49. The location data (geographic coordinates) of each survey cluster were collected using Global Positioning System (GPS) receivers. Each cluster’s GPS reading was taken in the center. To protect respondents’ privacy, GPS latitude/longitude positions for all survey groups were randomly displaced. The highest displacement for urban clusters was two kilometers (km), whereas, for 99% of rural clusters, it was five kilometers (km). The remaining 1% of rural clusters have been relocated up to a distance of ten km [19].

Measurement of variables of the study

The outcome variable was missing BPCR messages and was measured based on the six key WHO-recommended messages delivered to a mother during her last pregnancy. The key messages were about deciding the place of birth, collecting the necessary supplies for childbirth, arranging emergency transportation, setting money aside for emergency conditions, identifying companions during labor and childbirth, and securing potential blood donors. Information on these six key BPCR messages was obtained from the response to the question: During any of your antenatl visist, were you told about your Place of birth? Were you told about supplies needed for birth? Were you told about emergency transportation? The responses were recorded as Yes (= 1) or No (= 0). A single woman may be advised about any of the six messages many times throughout the same pregnancy, but each response was recorded as a single message. Based on the responses, a composite index of BPCR was created, which had a minimum and maximum value of 0 and 6 respectively. The women with ’0’ values were regarded as ’missing BPCR messages,’ whereas those with one or more messages were considered to have received BPCR [12, 20].

Explanatory variables. Potential determinants of BPCR were extracted from the data set after reviewing related and current literature [10–12, 18, 21] (Table 1).

[Figure omitted. See PDF.]

Autonomy in decision-making was measured using replies to questions about who makes the ultimate decision for the family on big property purchases, visits to relatives, and health care. (i) respondent alone, (ii) respondent and husband/partner, (iii) husband/partner alone, (iv) someone else, and (v) others were the response categories. For each question, replies (i) or (ii) earned a score of 1, indicating good decision-making capacity, while the remaining responses were assigned a value of 0, indicating limited capacity. The responses on each of the three aspects were added together to get a total score ranging from 0 to 3. Finally, a composite score had been divided into two unique groups: low and high for "0 to 2" and "3" scores, respectively [22, 23].

Data management and statistical analysis

STATA version 14.1 software was used for data extraction, cleaning, coding, and analysis, while ArcGIS 10.7 and SaTScan were used for spatial analysis. The data were weighted by applying a weighting factor to minimize under- or over-representation of the data in the surveys due to differential selection among strata. Using the svyset command, the data was further structured as survey data. The weighting procedure was thoroughly discussed in the 2016 EDHS report [5]. The weighted proportions of a response and explanatory variables have been estimated in STATA and exported to Microsoft Excel 2016 before being loaded into ArcGIS 10.7 for further analysis. To describe the background characteristics of respondents, descriptive statistics such as frequency and percentage were used.

Spatial analysis

Spatial autocorrelation.

To figure out if the spatial distribution of not receiving BPCR messages in Ethiopia was dispersed, clustered, or randomly distributed, the global spatial autocorrelation (Global Moran’s I) was carried out [24]. The Global Moran’s I statistic is used to quantify spatial autocorrelation by taking the whole data set and producing a single output value ranging from -1 to +1. Moran’s I value at -1, +1, and 0 indicates the likelihood of missing BPCR messages being dispersed, clustered, and randomly distributed, respectively. The Z-score determines the statistically significant difference in clustering, and the p-value determines the significance. Accordingly, Moran’s, I value was statistically significant (p<0.05), showing the presence of spatial autocorrelation and it showed the need for hot spot analysis [25].

Incremental spatial autocorrelation.

The peak distance from the spatial incremental autocorrelation model identifies the place where the spatial dependency of missing BPCR messages was most evident and that highest peak distance value has been selected as the threshold distance for hotspot analysis (S1 File) [26].

Hot spot analysis (Gettis-Ord Gi* statistics).

The local spatial analysis was performed using Getis-Ord Gi* statistics to identify important hot spot and cold spot areas for missing BPCR messages. The hotspot analysis compares the local mean rate (the rates for a cluster and its nearest neighbouring clusters) to the global mean rate (the rates for all clusters). The z-score was determined to confirm the statistical significance of clustering, and the p-value was calculated to establish the significance at a p-value<0.05 with 95% CI. Cold spots were announced if the z-score is less than -1.96, and hotspot locations were declared if the z-score was greater than +1.96 [24, 27–29]. In addition, Incremental Spatial Autocorrelation was examined to determine a suitable threshold for finding spatial processes that favour clustering [30].

Spatial interpolation.

Based on reported or observed values from selected enumeration areas (EAs), spatial interpolation was used to anticipate the proportion of missing BPCR messages in unsampled EAs. For doing this, the standard Kriging interpolation method was used instead of other interpolation approaches since it is an optimal interpolator with a minimum mean error (ME) and root mean square error (RMSE) [31, 32].

Spatial scan statistical analysis

Using Kuldorff’s SaTScan version 9.6 statistical software, the Bernoulli-based model was employed to detect the statistically significant spatial clusters with missing BPCR messages. A Bernoulli-based model was used in which events at particular places were analyzed, whether women received (Control) or missed (Case) BPCR messages. Scan statistics scanned the space gradually to determine the number of observed and expected observations inside the window at each cluster. The scanning window with the maximum likelihood was the most likely and high-performing cluster for being a case, and the level of significance was determined at a p-value <0.05.

Spatial regression analysis

Spatial regression incorporates both global (ordinary list squares) and local(geographically weighted regression) analysis approaches [33, 34].

Ordinary least squares (OLS) regression

The Ordinary Least Squares (OLS) model is a global statistical model used to evaluate and explain the relationship between outcome and explanatory variables [35]. It served as a tool for diagnosis as well as to pick the best predictors for the Geographically Weighted Regression (GWR) model. Findings from OLS regression are only reliable if the regression model meets all of the assumptions necessary by this method. In a properly fitted OLS model, coefficients of predictors should be statistically significant and have either a positive or negative sign. Furthermore, there should be no multicollinearity among explanatory variables [34, 36]. Multicollinearity was examined using the Variance Inflation Factor (VIF), and those variables with VIF > 10 were deemed as multicollinear (redundant) and were removed from the model. Furthermore, the Koenker Bp statistic had been utilized to determine whether the model could be used to do a spatially weighted regression analysis. The Koenker statistics were significant (p-value = 0.0198) in the current study, and GWR analysis was necessary to analyze the distribution of relevant predictors.

Geographically weighted regression (GWR)

It is evident that a strong predictor in one cluster is not always a strong predictor in another. The application of GWR can identify this form of cluster variation (non-stationarity) of predictors [37]. Unlike OLS which fits a single linear regression equation to all of the data of all features (in this instance, the clusters), GWR uses data from neighboring features, so the GWR coefficient has distinct values for each cluster [38]. The six checks indicated for spatial regression analysis were also carried out [37]. Variables with p-values less than 0.05 are chosen and discussed based on their coefficients. Finally, the model with the lowest AICc score and the highest adjusted R-squared value was determined to be the ideal fit for the data.

Ethical consideration and consent to participate

Written permission was received from ICF International to access both the DHS and GPS datasets following registration at the DHS Program office with possible justification. The accessed data were only used for the registered research and were not shared with anyone other than the co-authors. The DHS also declared that informed consent and assent was obtained from study participants or their legal guardian during the primary data collection [5]. Furthermore, the Institutional Review Board of Wachemo University College of Medicine and Health Sciences declared that no formal ethics approval was required in this case because it is secondary data, but to ensure the ethical compliance of the research process with the national and international standards.

Results

Socio-demographic characteristics of respondents

A total weighted sample of 4771 women was considered in the current study. The mean (±SD) age of women was 28.79(±6.61) years, with more than half (52.3%) of them belonging to the age group 25–34 years. One-third (33.7%) of study participants were from the Oromia region. The vast majority of respondents (81.7%) were from rural areas of the country, and more than half (54.1%) had no formal education. The proportion of women who missed BPCR messages was higher among those living in the poorest wealth quintile (51.0%), and received only one ANC visit(65.8%). Furthermore, 44.7% of women living in rural areas did not receive BPCR messages (Table 2).

[Figure omitted. See PDF.]

The level of missing BPCR messages

During the 2016 EDHS, 44.0% (95% CI: 42.6, 45.4%) of the surveyed women reported that they missed all BPCR messages during their last pregnancy. The proportion of women who missed the messages was highest in the Afar and Somali regions, with 78.1% and 67.2%, respectively. The top three messages not received by the majority of women were about arranging a potential blood donor (92.7%), identifying a companion during childbirth (90.8%), and securing emergency transportation (84.1%) (Table 3).

[Figure omitted. See PDF.]

Spatial autocorrelation of not receiving BPCR messages

The spatial autocorrelation analysis revealed that there was a considerable spatial variation in not receiving BPCR messages across the country (Global Moran’s I value 0.218, p-value<0.001). The clustered patterns (on the right sides) suggest that missed BPCR messages happened at a high rate throughout the study area. The Z-score of 7.212 implies that there is less than a 1% possibility that this clustered pattern is due to random chance (Fig 1).

[Figure omitted. See PDF.]

The hot spot (Getis-Ord Gi*) analysis for missing BPCR messages

Hot spot analysis enables the detection of either high or low statistically significant coverage regions with the absence of BPCR messages. The northern, southern, and western Somali, central Afar, and Gambella regions were found as statistically significant hotspots (red dots) for not getting BPCR messages. While considerable cold spot areas were found in Tigray, Amhara, and Addis Ababa (Fig 2).

[Figure omitted. See PDF.]

Interpolation using GIS mapping. Based on EDHS-2016 sampled data, areas with a relatively high proportion of missing BPCR messages (red) were predicted in Somali, Afar, and some parts of Oromia regions. In contrast, the predicted proportion of high receipt of BPCR messages (green shaded) encompasses the entire Tigray, Western Amhara, and central Addis Ababa regions (Fig 3).

[Figure omitted. See PDF.]

Spatial scan statistical (SaTScan) analysis. The SaTScan spatial analysis detected a total of seven statistically significant SaTScan cluster areas with a high proportion of women missing BPCR messages, which means that the prevalence is higher inside the SaTScan circular window compared to outside the SaTScan window. The most likely primary SaTScan cluster of areas was detected in the Somali region at geographical coordinates of (6.745502 N, 44.259011 E) and with (LLR = 44.68, RR = 1.70, p < 0.001). This revealed that women within the spatial window had 1.70 times higher odds of missing BPCR messages as compared to their counterparts outside the spatial window (Table 4).

[Figure omitted. See PDF.]

Predictors of missing BPCR messages

The global ordinary least square (OLS) analysis result.

OLS is a global regression model that uses a single equation to estimate the relationship between the dependent and independent variables, and it assumes the coefficients of each variable are homogenous across the study area. By meeting all of the fundamental assumptions, the initially employed spatial analysis, OLS model was fitted for the candidate explanatory variables. The model has been tuned to detect multicollinearity among the independent variables, with a mean VIF of less than 10. The adjusted R2 indicated that 24.2% (adjusted R2 = 0.242) of the variation in missing BPCR messages was explained by those six explanatory variables in the model. The robust probabilities for the intercept and all the included explanatory showed coefficient significance (p < 0.01) for the explanatory variables. Jarque-Bera statistics with p-values greater than 0.05 suggest that the model prediction using OLS was not biased (the assumption of residual normality was met). The Koenker statistics in the model, on the other hand, exhibited a statistically significant p-value (p = 0.0198), indicating that the regression model is inconsistent across the study area (as geographic position changes, so does the association of variables. This implies that the GWR model was deemed more suited for estimating model parameters. Consequently, being in the poorest wealth quintile, having only one ANC visit, having no access to listening to the radio, having difficulty accessing money to get medical care, not having a mobile phone, and not being covered by any health insurance scheme were the spatial determinants of hot spot areas for missing BPCR messages (Table 5).

[Figure omitted. See PDF.]

Geographically weighted regression (GWR)

Although the OLS model assumes that the association between each covariate and the outcome of interest is stationary across the study area, significant Koenker (BP) Statistics (p<0.01) show that this assumption is violated. To deal with this breach of the global (OLS) model’s stationarity assumption, the local (GWR) model was fitted to provide credible estimates. The GWR analysis showed a considerable improvement over the global model (OLS). The AICc value decreased from 181.81 in the OLS model to 161.21 in the GWR model. Likewise, the adjusted R² in OLS ascended to 0.332 in GWR, indicating that the local model improved the capacity for predicting hotspots of missing BPCR messages. When the AICc values of two models (OLS versus GWR) differ by more than three, the model with the lower AICc is considered superior. Overall, GWR analysis outperformed the model developed with OLS in the current study.

The predicted coefficient of being in the poorest wealth quintile varies across the regions, revealing an uneven association between poverty and the proportion of missing BPCR messages. As the proportion of women living in the poorest wealth quintile increased, so did the proportion of missing BPCR messages in most parts of Afar and Somali, as well as in some parts of Gambella and Benishangul Gumuz. In those regions, the predicted coefficient ranges from 0.67 to 1.00 units. The northern part of the country, on the other hand, had the lowest coefficient for women in the poorest wealth quintile (northern, southern, and central Amhara region, and the majority of Tigray region). The red dotted area denotes a strong influence (high predicted coefficient) of the explanatory variables on missing BPCR messages (Fig 4).

[Figure omitted. See PDF.]

Besides, receiving the lowest number of ANC visits (only one visit) was associated with the spatial variation of missing BPCR messages. As the proportion of women who received only one ANC visit increased, the percentage of missing BPCR messages among pregnant women in various parts of Afar, western, southern, and northern Somali, and some parts of SNNPR and Diredawa. The weakest relationship was detected in the majority of parts of Tigray, and Amhara regions (Fig 5).

[Figure omitted. See PDF.]

The proportion of women who never listen to the radio was found to be strongly associated with an increased risk of missing BPCR messages, with the greatest effect observed in Somali (southern, northern, and western parts), Afar (eastern and southern parts), and Gambella (northern and southern parts). The weakest association with the predicted coefficient of 0.04 to 0.32 was found in the northern Tigray and some parts of the Amhara regions (Fig 6).

[Figure omitted. See PDF.]

Furthermore, not having a mobile phone was associated with a regional difference in not getting BPCR messages. Women in the northern, southern, and western parts of the Somali region and the eastern border of Afar who did not have mobile phones had a higher likelihood of missing BPCR messages (Fig 7).

[Figure omitted. See PDF.]

The predicted coefficient of difficulty in obtaining money for medical care on the level of missing BPCR messages varies by geographical region. As the proportion of women who faced difficulty in getting money increased, the failure to receive BPCR messages in the northern and southern portions of Somali, as well as the central and south-western parts of the Afar regions. Additionally, it encompasses some parts of Diredawa and Hareri, Oromia (southern), Gambella (northern), and the Benishangul Gumuz (southern) regions (Fig 8).

[Figure omitted. See PDF.]

Finally, not covered by any type of health insurance scheme has a positive relationship with missing BPCR messages. As the proportion of women who were not covered by any HI schemes increased, the prevalence of missing BPCR increased greatly in all regions except Amhara and Tigray (Fig 9).

[Figure omitted. See PDF.]

Discussion

The overall prevalence of missing BPCR messages in Ethiopia was found to be 44.0% (95% CI: 42.6, 45.4). Because there was a dearth of studies on the topic at the national level, it was difficult to compare the prevalence to other findings. The findings of spatial global Moran’s analysis found that the proportion of women who missed BPCR messages significantly varied across regions.

As per hot spot analysis, the vast majority of Somali, central Afar, and Gambella regions had a high clustering (hot spot) of not getting BPCR messages. Those regions were known for low uptake of maternal and child health services [39–41]. Prior small-scale studies conducted in those regions also supported this finding [42–44]. This clustering might be due to a variety of factors. To begin, those two regions are known to have fewer healthcare facilities, inadequate healthcare providers, and lesser availability of medical supplies and equipment to access maternal health services when compared to other regions [40, 45]. In addition, these regions frequently have remote and difficult-to-reach areas with limited transportation, cultural practices that may take precedence over modern medical care, and lower socioeconomic conditions, such as poverty and a lack of education, making it difficult for pregnant women to access healthcare services on time [40, 41, 46]. Furthermore, some populations in these regions live nomadic or semi-nomadic lives, which might make it difficult for pregnant women to get consistent healthcare services because their locations shift over time [40, 47]. All these can lead to a lack of proper follow-up during pregnancy which might lead to missing BPCR messages.

As per the local (GWR) model, being in the poorest wealth quintile, receiving only one ANC visit, never listening to the radio, facing difficulty in accessing money, not having a mobile phone, and not being covered by any health insurance scheme were identified as significant spatial predictors of missing BPCR messages in.

Living in a household with the poorest wealth quintile was identified as a significant predictor of hot spot for not receiving BPCR messages in most parts of Afar and Somali, as well as in some parts of Gambella and Benishangul Gumuz. In those regions, the predicted coefficient ranges from 0.67 to 1.00 units. This was supported by studies conducted in Nigeria [48], Ghana [49], and Ethiopia [14, 50, 51]. This association may be explained by the low utilization of maternal health services in those regions [39–41]. In addition, those women in the poorest wealth quintile are more likely to experience financial barriers in accessing maternal healthcare especially ANC, which is a vital entry point to get information about BPCR messages. In addition, poverty and economic challenges can cause extra pressures on women and families, by hampering access to media and the ability to make decisions [52]. All of this might influence their capacity to seek healthcare services as well as receive key BPCR messages. It is vital to emphasize that being in the poorest wealth quintile might lead to a cycle of inadequate service use and poor health outcomes. Improving the issue necessitates multifaceted approaches that address economic disparities, enhance public infrastructures, promote education, and engage communities in addressing cultural norms and beliefs [53].

Similarly, the frequency of ANC visits is a key spatial predictor of hotspots of missing BPCR messages across the region. The likelihood of missing BPCR messages increased in the vast majority of Afar, western, southern, and northern Somali, and some parts of SNNPR and Diredawa as the proportion of women who received only one ANC visit increased. Studies conducted in Tanzania [54], Egypt [55], Kenya [56], and Ethiopia [43, 51] found that having a low number of ANC visits was positively related to not receiving BPCR messages. This could be attributed to a variety of reasons. To begin, receiving BPCR messages entails a sequence of steps that must be followed to ensure a safe pregnancy and childbirth, and if a woman just attends one visit, she may not obtain comprehensive information about all of these elements. In addition, with changing circumstances, not getting adequate visits resulted in missing out on all of the WHO-recommended BPCR messages. Overall, because pregnancy is a dynamic process, obtaining only one ANC visit can result in missing BPCR messages due to a lack of comprehensive education, support, tailored updates, engagement, and follow-ups that subsequent visits offer [57, 58]. Thus, pregnant women need to attend all directed ANC visits to ensure they obtain the appropriate information and care for a safe pregnancy and childbirth.

Failure to listen to the radio was found to be strongly associated with an increased risk of missing BPCR messages in Somali (southern, northern, and western parts), Afar (eastern and southern parts), and Gambella (northern and southern parts). This was supported by studies conducted in Uganda [59], and Ethiopia [12, 60, 61]. As a result of not tuning in to the radio, women may remain unaware of and miss out on services available to them throughout pregnancy and childbirth, becoming alienated or feeling more lonely and less supported during their pregnancy experience [62]. All these could contribute to missing BPCR messages during pregnancy. Subsequently, a lack of radio access can limit women’s ability to make informed decisions and contribute to discrepancies in the uptake of key BPCR messages. Hence, a concerted effort is needed to enhance the accessibility of radio, to ensure that no woman is deprived of those key BPCR messages during pregnancy. Similarly, not having a mobile phone was associated with a regional difference in not getting BPCR messages in the northern, southern, and western parts of the Somali region and the eastern border of Afar. Studies conducted elsewhere [63–65] showed that there was a strong association between having a mobile phone and receipt of BPCR messages. This could be because women without mobile phones were less likely to receive health-related information, appointment reminders, health awareness campaigns, and peer support, resulting in BPCR messages being overlooked [66]. Overall, mobile phones provide a versatile and accessible platform for bridging gaps in maternal healthcare service usage through providing information, support, reminders, and access to healthcare providers, particularly in areas with limited healthcare infrastructure [67]. As a result, it is vital to prioritize digital literacy, network coverage, and cost to promote fair access for all women.

Finally, being not covered by health insurance schemes depicts a positive link with missing BPCR messages; as the proportion of women who were uninsured increased, the prevalence of missing BPCR increased significantly in all regions except Amhara and Tigray. Studies conducted in Rwanda [68], and Nigeria [69] showed high uptake of BPCR messages among those women who were insured. This could be because a lack of health insurance coverage can create hurdles to access and continuity of care, as well as engagement with health service delivery points, due to cost concerns and financial hardship [70, 71]. As a result, people may connect with healthcare providers less frequently, missing out on vital information about BPCR messages. On the other hand, women who were covered by health insurance were more likely to visit health facilities, and this trend made them more familiar with the healthcare system, which in turn boosted their health-seeking behavior during pregnancy [72].

The study has the following strengths. First, the findings from the current study were based on an analysis of nationally representative data, making the findings more generalizable. Also, the findings from the Hot Spot, SaTScan, and GWR analyses can assist government and program planners in designing geographically focused public health interventions to boost BPCR message delivery. Despite the strengths, the findings should be interpreted considering the following limitations. Firstly, to ensure the privacy of respondents, the geographical coordinates of clusters were displaced by up to 2km in urban areas, 5km for most clusters in rural areas, and 10km for 1% of clusters in rural areas; this may impact estimated cluster effects in the spatial regression. In addition, due to the cross-sectional nature of the data, the findings may be affected by recall- and social desirability biases and it is difficult to establish a temporal/causal relationship between the outcome and explanatory variables.

Conclusion

The level of missing BPCR messages during pregnancy was found to be high in Ethiopia with a significant spatial variation across regions. Living in the poorest wealth quintile, receiving only one ANC visit, not having access to listening to the radio, having difficulty accessing money to get medical care, not having a mobile phone, and being not covered by health insurance were identified as significant predictors of hot spots for missing BPCR messages. The level of missing BPCR messages during pregnancy was found to be high in Ethiopia, with significant local variation. As a result, policymakers at the national level and local planners should develop strategies and initiatives that enhance women’s economic capacities, health-seeking behavior, and media exposure. Furthermore, the regional authorities should focus on strategies that promote universal health coverage through enrolling citizens in health insurance schemes.

Supporting information

S1 File. Output for incremental spatial autocorrelation.

https://doi.org/10.1371/journal.pone.0295744.s001

Acknowledgments

We are grateful to ICF macro (Calverton, USA) for providing the 2016 DHS data of Ethiopia.

Citation: Habte A, Hailegebreal S, Melis T, Haile D (2023) Spatial variation and predictors of missing birth preparedness and complication readiness (BPCR) messages in Ethiopia. PLoS ONE 18(12): e0295744. https://doi.org/10.1371/journal.pone.0295744

About the Authors:

Aklilu Habte

Roles: Conceptualization, Data curation, Formal analysis, Methodology, Project administration, Software, Validation, Visualization, Writing – original draft, Writing – review & editing

E-mail: [email protected]

Affiliation: School of Public Health, College of Medicine and Health Sciences, Wachemo University, Hosanna, Ethiopia

ORICD: https://orcid.org/0000-0002-5719-4294

Samuel Hailegebreal

Roles: Methodology, Software, Writing – original draft

Affiliation: Department of Health Informatics, College of Medicine and Health Sciences, School of Public Health, Wachemo University, Hosaena, Ethiopia

Tamirat Melis

Roles: Methodology, Writing – review & editing

Affiliation: Department of Public Health, College of Medicine and Health Sciences, Wolkite University, Wolkite, Ethiopia

Dereje Haile

Roles: Methodology, Writing – review & editing

Affiliation: Department of Reproductive Health, College of Medicine and Health Sciences, School of Public Health, Wolaita Sodo University, Sodo, Ethiopia

[/RAW_REF_TEXT]

[/RAW_REF_TEXT]

[/RAW_REF_TEXT]

[/RAW_REF_TEXT]

[/RAW_REF_TEXT]

[/RAW_REF_TEXT]

[/RAW_REF_TEXT]

[/RAW_REF_TEXT]

[/RAW_REF_TEXT]

References

1. WHO, Maternal mortality,Key facts, 2023 https://www.who.int/news-room/fact-sheets/detail/maternal-mortality.

2. Organization W.H., Trends in maternal mortality 2000 to 2020: estimates by WHO, UNICEF, UNFPA, World Bank Group and UNDESA/Population Division: executive summary. 2023.

3. Zarocostas J., Global maternal mortality rates stagnating. The Lancet, 2023. 401(10377): p. 632. pmid:36842436

4. USAID, Ethiopia, Maternal, Neonatal and Child Health https://www.usaid.gov/ethiopia/global-health/maternal-neonatal-and-child-health#:~:text=Since%202000%2C%20Ethiopia%20has%20reduced,1%2C000%20are%20still%20too%20high.

5. CSA, I., Central Statistical Agency (CSA)[Ethiopia] and ICF. Ethiopia Demographic and Health Survey 2016. Addis Ababa, Ethiopia, and Rockville, Maryland, USA: CSA and ICF; 2016. 2017.

6. Solnes Miltenburg A., et al., Impact of birth preparedness and complication readiness interventions on birth with a skilled attendant: a systematic review. PloS one, 2015. 10(11): p. e0143382. pmid:26599677

7. Organization, W.H., Birth and emergency preparedness in antenatal care. Standards for Maternal and Neonatal Care. Geneva, World Health Organization, 2006: p. 1–6.

8. Jhpiego, Monitoring birth preparedness and complication readiness: tools and indicators for maternal and newborn health. 2004.

9. Sk M.I.K., et al., Praying until death: revisiting three delays model to contextualize the socio-cultural factors associated with maternal deaths in a region with high prevalence of eclampsia in India. BMC pregnancy and childbirth, 2019. 19: p. 1–11.

10. Ohenhen V., et al., Birth preparedness and complication readiness: Evaluating the “know-do” gap among women receiving antenatal care in Benin City, Nigeria. Journal of Biosocial Science, 2023: p. 1–15. pmid:36605008

11. Ketema D.B., et al., Effects of maternal education on birth preparedness and complication readiness among Ethiopian pregnant women: a systematic review and meta-analysis. BMC pregnancy and childbirth, 2020. 20(1): p. 1–9. pmid:32143581

12. Habte A., Tamene A., and Woldeyohannes D., The uptake of WHO-recommended birth preparedness and complication readiness messages during pregnancy and its determinants among Ethiopian women: A multilevel mixed-effect analyses of 2016 demographic health survey. Plos one, 2023. 18(3): p. e0282792. pmid:36952431

13. Markos D. and Bogale D., Birth preparedness and complication readiness among women of child bearing age group in Goba woreda, Oromia region, Ethiopia. BMC pregnancy and childbirth, 2014. 14: p. 1–9.

14. Debelew G.T., Afework M.F., and Yalew A.W., Factors affecting birth preparedness and complication readiness in Jimma Zone, Southwest Ethiopia: a multilevel analysis. Pan African Medical Journal, 2014. 19(1).

15. Bitew Y., Awoke W., and Chekol S., Birth preparedness and complication readiness practice and associated factors among pregnant women, Northwest Ethiopia. International scholarly research notices, 2016. 2016. pmid:27722201

16. Kaso M. and Addisse M., Birth preparedness and complication readiness in Robe Woreda, Arsi Zone, Oromia Region, Central Ethiopia: a cross-sectional study. Reproductive health, 2014. 11: p. 1–12.

17. Worku M., et al., Male involvement and associated factors in birth preparedness and complication readiness in Debre Berhan town, North East Ethiopia. The Pan African Medical Journal, 2020. 35. pmid:32499852

18. Debelie T.Z., et al., Birth preparedness and complication readiness practice and associated factors among pregnant women in Northwest Ethiopia: 2018. PLOS ONE, 2021. 16(4): p. e0249083.

19. ICF International. 2013. Incorporating Geographic Information into Demographic and Health Surveys: A Field Guide to GPS Data Collection. Calverton, Maryland, USA: ICF International https://dhsprogram.com/pubs/pdf/DHSM9/DHS_GPS_Manual_English_A4_24May2013_DHSM9.pdf.

20. Jaleta A.D., et al., Birth Preparedness, Complication Readiness, and Determinants Among Women Attending Antenatal Care from Ethiopia: A Cross-sectional Facility-Based Survey. European Journal of Clinical and Biomedical Sciences, 2021. 7(3): p. 37–43.

21. Begashaw B., et al., Assessment of birth preparedness and complication readiness among pregnant mothers attending ante Natal Care Service in Mizan-Tepi University Teaching Hospital, South West Ethiopia. Clinics Mother Child Health, 2017. 14.

22. Kassahun A. and Zewdie A., Decision-making autonomy in maternal health service use and associated factors among women in Mettu District, Southwest Ethiopia: a community-based cross-sectional study. BMJ open, 2022. 12(5): p. e059307. pmid:35501088

23. Asaolu I.O., et al., Measuring women’s Empowerment in sub-Saharan Africa: exploratory and confirmatory factor analyses of the demographic and health surveys. Frontiers in psychology, 2018. 9: p. 994. pmid:29971030

24. Chen Y., An analytical process of spatial autocorrelation functions based on Moran’s index. PLoS One, 2021. 16(4): p. e0249589. pmid:33852621

25. Ord J.K. and Getis A., Local spatial autocorrelation statistics: distributional issues and an application. Geographical analysis, 1995. 27(4): p. 286–306.

26. Lu J., Incremental autocorrelation calculation for streamed data using components. 2019, Google Patents.

27. Chen Y., New approaches for calculating Moran’s index of spatial autocorrelation. PloS one, 2013. 8(7): p. e68336. pmid:23874592

28. Melaku M.S., et al., Geographical variation and predictors of zero utilization for a standard maternal continuum of care among women in Ethiopia: a spatial and geographically weighted regression analysis. BMC Pregnancy and Childbirth, 2022. 22(1): p. 76. pmid:35090405

29. Wulder M. and Boots B., Local spatial autocorrelation characteristics of remotely sensed imagery assessed with the Getis statistic. International Journal of Remote Sensing, 1998. 19(11): p. 2223–2231.

30. Getis A. and Ord J.K., The analysis of spatial association by use of distance statistics. Geographical analysis, 1992. 24(3): p. 189–206.

31. Meng J., Raster data projection transformation based-on Kriging interpolation approximate grid algorithm. Alexandria Engineering Journal, 2021. 60(2): p. 2013–2019.

32. Van Beers W.C. and Kleijnen J.P.. Kriging interpolation in simulation: a survey. in Proceedings of the 2004 Winter Simulation Conference, 2004. 2004. IEEE.

33. Fortin M.-J., et al., Spatial statistics, spatial regression, and graph theory in ecology. Spatial Statistics, 2012. 1: p. 100–109.

34. Chi G. and Zhu J., Spatial regression models for demographic analysis. Population Research and Policy Review, 2008. 27: p. 17–42.

35. Shrestha, P.M., Comparison of ordinary least square regression, spatial autoregression, and geographically weighted regression for modeling forest structural attributes using a Geographical Information System (GIS). Remote Sensing (RS) Approach. Thesis. Canada: University of Calgary. http://people. ucalgary. ca/~ mcdermid/Docs/Theses/Shrestha_2006. pdf (Accessed October 30, 2012), 2006.

36. Beale C.M., et al., Regression analysis of spatial data. Ecology letters, 2010. 13(2): p. 246–264. pmid:20102373

37. Fotheringham A.S., Brunsdon C., and Charlton M., Geographically weighted regression: the analysis of spatially varying relationships. 2003: John Wiley & Sons.

38. Chi G. and Wang D., Small-area population forecasting: A geographically weighted regression approach. The frontiers of applied demography, 2017: p. 449–471.

39. Bobo F.T., Yesuf E.A., and Woldie M., Inequities in utilization of reproductive and maternal health services in Ethiopia. International journal for equity in health, 2017. 16(1): p. 1–8.

40. Wester K.C., et al., Best practices for addressing socio-cultural barriers to reproductive, maternal and neonatal health service utilization among women from pastoralist communities of Afar, Ethiopia. The Ethiopian Journal of Health Development, 2018. 32(Special Is).

41. Muchie K.F., Quality of antenatal care services and completion of four or more antenatal care visits in Ethiopia: a finding based on a demographic and health survey. BMC pregnancy and childbirth, 2017. 17(1): p. 1–7.

42. Abebaw N., et al., Assessments of Birth Preparedness and Complication Readiness Among Women of Childbirth in Samara Logia Town, Afar, North East Ethiopia. American Journal of Laboratory Medicine, 2020. 5(4): p. 95–101.

43. Musa A. and Amano A., Determinants of birth preparedness and complication readiness among pregnant woman attending antenatal care at Dilchora Referral Hospital, Dire Dawa City, East Ethiopia. Gynecol Obstet (Sunnyvale), 2016. 6(2): p. 356.

44. Regesu A.H., et al., Birth Preparedness and Complication Readiness and Associated Factors among Mothers Attending Antenatal Care at Chiro District Health Centers, West Hararge, Ethiopia. International Journal of Health Sciences and Research, 2019. 9(4): p. 187–95.

45. Eba K., et al., Mobile health service as an alternative modality for hard-to-reach pastoralist communities of Afar and Somali regions in Ethiopia. Pastoralism, 2023. 13(1): p. 1–12.

46. Birch I., Barriers affecting demand for reproductive, maternal, newborn and child health (RMNCH) services in Ethiopia. 2020.

47. Zepro N.B., et al., Lived Experiences and Perceptions of Childbirth among Pastoralist Women in North-Eastern Ethiopia: A Multimethod Qualitative Analysis to the WHO Health Systems Responsiveness Framework. International Journal of Environmental Research and Public Health, 2021. 18(23): p. 12518. pmid:34886246

48. Idowu A., et al., Birth preparedness and complication readiness among women attending antenatal clinics in Ogbomoso, South West, Nigeria. International Journal of Maternal and Child Health and AIDS (IJMA), 2015. 4(1): p. 47–56.

49. Saaka M. and Alhassan L., Prevalence and predictors of birth preparedness and complication readiness in the Kassena-Nankana district of Ghana: an analytical cross-sectional study. BMJ open, 2021. 11(3): p. e042906. pmid:33789849

50. Gurara M.K., et al., Maternity waiting homes as component of birth preparedness and complication readiness for rural women in hard-to-reach areas in Ethiopia. Reproductive health, 2021. 18(1): p. 1–14.

51. Girma D., Waleligne A., and Dejene H., Birth preparedness and complication readiness practice and associated factors among pregnant women in Central Ethiopia, 2021: A cross-sectional study. Plos one, 2022. 17(10): p. e0276496. pmid:36301854

52. Wuneh A.D., et al., Wealth-based equity in maternal, neonatal, and child health services utilization: a cross-sectional study from Ethiopia. International Journal for Equity in Health, 2019. 18: p. 1–9.

53. Mwiti F. and Goulding C., Strategies for community improvement to tackle poverty and gender issues: An ethnography of community based organizations (‘Chamas’) and women’s interventions in the Nairobi slums. European Journal of Operational Research, 2018. 268(3): p. 875–886.

54. Urassa D.P., Pembe A.B., and Mganga F., Birth preparedness and complication readiness among women in Mpwapwa district, Tanzania. Tanzania journal of health research, 2012. 14(1). pmid:26591746

55. Aziz M.M., El-Deen R.M.S., and Allithy M.A., Birth preparedness and complication readiness among antenatal care clients in Upper Egypt. Sexual & Reproductive Healthcare, 2020. 24: p. 100506. pmid:32193127

56. Omari P.K., Afrane Y.A., and Ouma P., Birth preparedness and complication readiness among women attending antenatal care clinic in health facilities within Bureti Sub County of Kericho County, Kenya. 2016.

57. Villar J., et al., WHO antenatal care randomised trial for the evaluation of a new model of routine antenatal care. Lancet, 2001. 357. pmid:11377642

58. Organization, W.H., WHO recommendations on antenatal care for a positive pregnancy experience. 2016: World Health Organization.

59. Asp G., et al., Associations between mass media exposure and birth preparedness among women in southwestern Uganda: a community-based survey. Global health action, 2014. 7(1): p. 22904. pmid:24433945

60. Andarge E., Nigussie A., and Wondafrash M., Factors associated with birth preparedness and complication readiness in Southern Ethiopia: a community based cross-sectional study. BMC pregnancy and childbirth, 2017. 17(1): p. 1–13.

61. Gedefa A.G., et al., Barriers to birth preparedness and complication readiness among pregnant women in rural Ethiopia: using a mixed study design, 2020. BMJ open, 2023. 13(4): p. e069565. pmid:37185635

62. Fatema K. and Lariscy J.T., Mass media exposure and maternal healthcare utilization in South Asia. SSM-Population Health, 2020. 11: p. 100614.

63. Masoi T.J. and Kibusi S.M., Improving pregnant women’s knowledge on danger signs and birth preparedness practices using an interactive mobile messaging alert system in Dodoma region, Tanzania: a controlled quasi experimental study. Reproductive health, 2019. 16: p. 1–10.

64. Santoso H.Y.D., et al., Android application model of “suami siaga plus” as an innovation in birth preparedness and complication readiness (BP/CR) intervention. Journal of family & reproductive health, 2017. 11(1): p. 30. pmid:29114266

65. Cheptum J., Omoni G., and Mirie W., Effectiveness of mobile phone text message reminder on birth preparedness in a rural community in Kenya. Tropical Journal of Obstetrics and Gynaecology, 2019. 36(1): p. 49–53.

66. Guy R., et al., How effective are short message service reminders at increasing clinic attendance? A meta‐analysis and systematic review. Health services research, 2012. 47(2): p. 614–632. pmid:22091980

67. Choudhury A., Asan O., and Choudhury M.M., Mobile health technology to improve maternal health awareness in tribal populations: mobile for mothers. Journal of the American Medical Informatics Association, 2021. 28(11): p. 2467–2474. pmid:34459478

68. Smeele P., et al., Birth preparedness and complication readiness among pregnant women admitted in a rural hospital in Rwanda. BMC pregnancy and childbirth, 2018. 18: p. 1–7.

69. Ekabua J.E., et al., Awareness of birth preparedness and complication readiness in southeastern Nigeria. International Scholarly Research Notices, 2011. 2011. pmid:21808742

70. Kesuma Z.M. and Chongsuvivatwong V., Utilization of the Local Government Health Insurance Scheme (JKA) for maternal health services among women living in underdeveloped areas of Aceh Province, Indonesia. Asia Pacific Journal of Public Health, 2015. 27(3): p. 348–359. pmid:24566605

71. Habte A., et al., Towards universal health coverage: The level and determinants of enrollment in the Community-Based Health Insurance (CBHI) scheme in Ethiopia: A systematic review and meta-analysis. PloS one, 2022. 17(8): p. e0272959. pmid:35980888

72. Seid A. and Ahmed M., Association between health insurance enrolment and maternal health care service utilization among women in Ethiopia. BMC Public Health, 2021. 21(1): p. 1–10.