Article Text

Original research
Spatial variation and associated factors of inadequate counselling regarding pregnancy danger signs during antenatal care visits among pregnant women in Ethiopia: a Geographically Weighted Regression Model
  1. Meron Asmamaw Alemayehu,
  2. Nebiyu Mekonnen Derseh,
  3. Tigabu Kidie Tesfie,
  4. Habtamu Wagnew Abuhay,
  5. Getaneh Awoke Yismaw,
  6. Muluken Chanie Agimas
  1. Department of Epidemiology and Biostatistics, Institute of Public Health, College of Medicine and Health Sciences, University of Gondar, Gondar, Ethiopia
  1. Correspondence to Ms Meron Asmamaw Alemayehu; merryalem101{at}gmail.com

Abstract

Introduction Inadequate counselling of pregnant women regarding pregnancy danger signs contributes to a delay in deciding to seek care, which causes up to 77% of all maternal deaths in developing countries. However, its spatial variation and region-specific predictors have not been studied in Ethiopia. Hence, the current study aimed to model its predictors using geographically weighted regression analysis.

Methods The 2019 Ethiopian Mini Demographic and Health Survey data were used. A total weighted sample of 2922 women from 283 clusters was included in the final analysis. The analysis was performed using ArcGIS Pro, STATA V.14.2 and SaTScan V.10.1 software. The spatial variation of inadequate counselling was examined using hotspot analysis. Ordinary least squares regression was used to identify factors for geographical variations. Geographically weighted regression was used to explore the spatial heterogeneity of selected variables to predict inadequate counselling.

Results Significant hotspots of inadequate counselling regarding pregnancy danger signs were found in Gambella region, the border between Amhara and Afar regions, Somali region and parts of Oromia region. Antenatal care provided by health extension workers, late first antenatal care initiation and antenatal care follow-up at health centres were spatially varying predictors. The geographically weighted regression model explained about 66% of the variation in the model.

Conclusion Inadequate counselling service regarding pregnancy danger signs in Ethiopia varies across regions and there exists within country inequality in the service provision and utilisation. Prioritisation and extra efforts should be made by concerned actors for those underprivileged areas and communities (as shown in the maps), and health extension workers, as they are found in the study.

  • Maternal medicine
  • OBSTETRICS
  • Antenatal

Data availability statement

Data are available in a public, open access repository. The dataset used for this study is publicly available at the MEASURE DHS programme website (https://www.dhsprogram.com/data).

http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

STRENGTHS AND LIMITATIONS OF THIS STUDY

  • This study used the latest, large-scale and nationally representative data.

  • A robust spatial modelling technique was used.

  • A few important explanatory variables were not included due to unavailability.

  • Geographical coordinates of clusters were displaced up to 10 km to keep anonymity of study subjects and this must be considered when interpreting the results.

Introduction

Pregnancy danger signs (PDSs) are warning signs that women might encounter during pregnancy and they signify a life-threatening complication. Counselling on those signs is to be conducted during the antenatal care (ANC) service of a pregnant woman. It involves telling a pregnant woman to seek immediate care if she encounters signs such as vaginal bleeding, severe headache, vaginal gush of fluid, blurred vision, severe abdominal pain, fever and/or convulsion.1

Life-threatening pregnancy complications can effectively be prevented with earlier identification of the mentioned danger signs and timely access to quality emergency obstetric care (EmOC). However, each day in the world, 830 women die during pregnancy, childbirth and post partum. About 99% of these deaths occur in low-income countries. The 2016 Ethiopian Demographic and Health Survey shows a maternal mortality ratio of 412 maternal deaths per 100 000 live births. This represents 25% of all deaths among reproductive-age women in the country.2

Poor knowledge of pregnant women regarding PDSs, which is a direct consequence of inadequate counselling during ANC visits, is the predominant contributor to two of three delays that cause massive maternal mortality in low-income and middle-income countries. It contributes to delay in deciding to seek care and delay in reaching to seek care on time. The former delay is reported to contribute to about 30%–77% of all maternal deaths in low-income and middle-income countries.3 Interventional studies confirmed that good knowledge regarding PDSs influences the use of obstetric services such as ANC and can ultimately reduce maternal mortality and morbidity.4 5

In Ethiopia, several community-based cross-sectional studies were conducted to assess the proportion of pregnant women who were consulted about the danger signs of pregnancy. The studies reported prevalence ranging from 7.9% to 58.8%. Media exposure, type of health facility, maternal educational level, timing of first ANC, educational status of husband/partner and type of health professional who provided the ANC service were among significant explanatory variables that affected counselling service regarding PDSs.1 6 7

Before the 2019 Mini Ethiopian Demographic and Health Survey (mEDHS), data on the outcome variable of this study (ie, inadequate counselling regarding the danger signs of pregnancy) had never been collected. As a result, as far as we are concerned, no study determined the spatial variation of inadequate counselling service regarding PDSs, or predictors of such spatial variation among pregnant women in Ethiopia. This implies that the ‘where’ dimension of inadequate counselling service has never been specifically identified in the country. Failure to consider a geographical variation of inadequate counselling service regarding PDSs may result in wastage of clinical time and public health resources, which are already overstretched. It could also cause socioeconomic inequity in public health resource allocation and policy design. Hence, it is economical to identify high-risk geographical areas for the development of targeted programmes.8

Region-level estimates of inadequate counselling regarding PDSs and their region-specific predictors are invaluable information for policymakers and healthcare professionals. It would help them to design locally sensitive interventions and also to evaluate past intervention strategies. To fill the information gap and provide this vital information that promotes evidenced-based decision, the current study aimed to assess predictors of inadequate counselling regarding PDSs hotspots using a geographically weighted regression (GWR) model.

Methodology

Study area

The study was conducted in Ethiopia. Ethiopia is strategically located in the Horn of Africa, 3′ and 14.8′′ latitude 33′ and 48′ longitude bordering Somalia, Sudan, Djibouti, Kenya and Eritrea with a total border length of 5311 km. The country occupies an area of 1.1 million km2 with an altitude that ranges from the highest peak at Ras Dashen (4620 m above sea level) down to the Dallol depression, about 148 m below sea level. Even though the country now is subdivided into 12 regions, the border for the newly added regions (Sidama, south) is not yet officially announced and consequently, the shape files are not available. Hence, the mEDHS keeps using the former regional structure which was nine regions and two administrative cities. The country is also further subdivided into zonal and woreda administrations.

Study design, data sources and sampling

This study is a secondary analysis of 2019 Ethiopian mini Demographic and Health Survey (mEDHS). The data collection period of the 2019 mEDHS was between 21 March 2019 and 28 June 2019. As mentioned earlier, there were nine regions and two administrative cities by the time when the mEDHS was conducted. In the 2019 mEDHS, the sampling frame was obtained from the 2019 Ethiopia Population and Housing Census (EPHC). The survey used a complete list of the 149 093 enumeration areas (EAs) where one EA on average contains 131 households.

Two-stage stratified sampling technique was used in the 2019 mEDHS. Initially, the nine regions and the two city administrations in the country were stratified into urban and rural areas. As a result, 21 sampling strata were created. Next, samples of EAs were selected from the 21 stratum in 2 stages. In the first stage, 305 EAs were selected with the random sampling technique from the 149 093 EAs that were created during the 2019 EPHC. About 93 of the selected clusters were from urban strata while 212 were from rural strata. Consequently, the list of households in all selected clusters was recorded. Listing of households in large clusters that contain more than 300 households was done after segmenting the EA and selecting only one segment, to minimise the time that it takes to list households. Hence, in this survey, the cluster is either an EA or a segment of EA. In the second stage, a systematic random sampling technique was used to select 30 households from each cluster. To select the 30 households, the list of households recorded in each selected EA or segment of EA was used as a sampling frame. The questionnaires were administered to women aged between 15 and 49 years and who either were permanent residents or visitors who spent at least a night in the selected household. Further details of the methodologies used in the survey can be accessed in the official 2019 mEDHS report.9

The current study included women aged between 15 and 49, who had a live birth in the last 5 years preceding the 2019 survey, from 283 clusters, and who received ANC from a skilled birth attendant. Women who responded ‘I do not know’ to the question ‘Were you told about the PDSs were excluded from the study (n=13).

To collect the geographical coordinates of each survey cluster, global positioning system (GPS) receivers were used. The GPS reading was made at the centre of each cluster. The GPS data were collected by ensuring it was away from tall buildings and tree canopies to facilitate the receiving of adequate satellite signals. The coordinate points for every survey cluster were displaced randomly 2 km for urban clusters. For rural clusters, the coordinate points were displaced 5 km for about 99% of rural clusters while for the remaining 1% of clusters, it was displaced a maximum of 10 km. The displacement was restricted to the country’s second administrative level so that the coordinate points would always stay within the country.10 The administrative polygons, used in the study to generate the map of every spatial analysis output, were taken from the Ethiopian Central Statistical Agency).

Study variables

The DHS Birth Record section questionnaire and dataset were used for the current study. It was collected based on the DHS programme’s standard questionnaire. The dependent variable (ie, counselled on danger signs) was one of the newly added variables in the 2019 mEDHS. It is defined as a pregnant women received a counselling service regarding the PDSs (vaginal bleeding, severe headache, vaginal gush of fluid, blurred vision, severe abdominal pain, fever, convulsion and any others) and to seek immediate care when she experience one of them. It was coded as a dichotomous variable with possible values of ‘1’ inadequate counselling regarding PDSs (if the woman had not been counselled on any of the mentioned danger signs and ‘0’ adequate counselling regarding danger signs (if the woman had been counselled on the danger signs).

The predictor variables were selected after a careful literature review. Predictors included in the exploratory regression are maternal age (grouped), maternal education level, maternal marital status, wealth quintile, type of ANC provider, timing of first ANC visit, total number of ANC visits, place of ANC visit, exposure to mass media, region and residence. Media exposure status was created from the frequency of watching television and listening to the radio, and if a woman has at least one yes, she is considered as having media exposure. The wealth index was computed and provided with the dataset by the ‘DHS programme’ using the Gini coefficient and was used as it is grouped into five categories poorest, poorer, middle, richest and richer. An exploratory regression tool in ArcGIS Pro was used to select explanatory variables that fulfil the assumptions of ordinary least squares (OLS) models.

Data source and extraction

After permission was obtained through a formal online request, the data were retrieved from the DHS programme’s official website and database which is (https://www.dhsprogram.com/data). A formal and separate request was submitted to access the GPS data. The data of the 2019 mEDHS are open to all DHS-registered researchers. After downloading the data, the response variable, potential predictor variables and coordinate data were extracted separately and merged for the final analysis.

Data management and analysis

Careful data cleaning, recoding and weighting were undertaken prior to any descriptive and spatial analysis. Descriptive and summary statistics were performed using STATA V.14.2 (Stata). ArcGIS Pro and SaTScan V.10.1 were used for the spatial analysis. Prior to the spatial analysis, the weighted proportion of the dependent variable and candidate predictor variables were computed on STATA software and exported to a CSV file format, to facilitate easier capture by the ArcGIS Pro. A detailed explanation of the sample weighting procedure is presented in the 2019 EMDHS final report.9

Descriptive analysis

Participant characteristics were described using frequency with per cent. Pearson’s χ2 tests were used to assess differences in short birth interval frequencies between places (urban/rural) and regions of residence.

Spatial analysis

Spatial autocorrelation analysis

The distribution of inadequate counselling service regarding danger signs of pregnancy complications was examined using the spatial autocorrelation (Global Moran’s I) statistics to identify whether it is clustered, dispersed or random. Moran’s I is a statistic that produces a single output number between −1 and +1. A Global Moran’s I value approaching +1 suggests that it was spatially clustered. A Global Moran’s I value approaching −1 indicates a dispersed spatial distribution of inadequate counselling regarding danger signs, while a Global Moran’s I value of 0 indicates a dispersed spatial distribution. The presence of a significant spatial autocorrelation (p<0.05) is confirmed by a statistically significant Global Moran’s I test.

Incremental spatial autocorrelation by distance

Spatial autocorrelation for a series of distances was measured using incremental spatial autocorrelation. The corresponding line graph and their z-scores were created. Z-scores indicate the intensity of spatial clustering of inadequate counselling service. Statistically significant peak z-scores reflect distances where geographical processes promoting clustering of inadequate counselling service regarding PDSs are most pronounced. The peak distance was used as one of input values for the hotspot analysis.

Hotspot analysis

The degree of clustering which could either be high or low was calculated using the Getis-Ord General G statistics in a hot spot analysis. The statistical significance of clustering was confirmed using the Z-score with a 95% CI and a p<0.05. Statistical output with a high Gi* shows the no counselling hotspots (clustering of high values), while a low Gi* the inadequate counselling cold spots (clustering of low values). The weighted proportion of inadequate counselling regarding danger signs in each of the 283 cluster was taken as an input for this hotspot analysis.

Spatial scan statistical analysis

Bernouii-based model using SaTScan V.10.1 software was used to test the occurrence of statistically significant spatial clusters with both high and low distribution of inadequate counselling regarding PDSs. To identify statistically significant clusters, SaTScan uses a scanning window that moves across the study area. To allow both small and large clusters to be detected and to ignore clusters that contained more than the maximum limit with the circular shape of a window, the default maximum spatial cluster size of 50% population was used as an upper limit. For this model, women who received inadequate counselling regarding PDSs were taken as cases and women who received adequate counselling service were taken as controls. Most likely clusters were identified using high log-likelihood ratio tests and significant p value based on Monte Carlo replication.

Spatial regression

Ordinary least squares regression

OLS is a global statistical model used to represent the relationship between an outcome and exploratory variables. Hence, to identify predictors of the observed spatial variations of inadequate counselling service regarding danger signs, this study used a spatial regression modelling technique called OLS. OLS regression gives reliable results only if the regression model satisfies all of the assumptions. The coefficients of predictor variables in a properly run OLS model should have either a positive or negative sign and be statistically significant. In addition, there should not be any multicollinearity or redundancy among explanatory variables. Variance inflation factor (VIF) <7.5 was used to declare this assumption as fulfilled. In addition, the model should be unbiased (heteroscedasticity or non-stationarity). The residuals should be normally distributed (random) and reveal no spatial patterns. The OLS regression is often used to select appropriate predictors, considering their relationship with inadequate counselling service regarding danger signs of pregnancy, for the GWR model. The Koenker BP statistics of the OLS model were used to decide whether performing GWR is appropriate or not. A significant Koenker test (p<0.05) indicates the relationship between inadequate counselling service regarding PDSs and exploratory variables is different across regions of the country/non-stationary.

Geographically weighted regression

An explanatory variable that is an independent predictor in one geographical area may not necessarily be a strong predictor in another cluster. The GWR model makes it easier to identify this type of cluster variation (non-stationarity). The underlying idea of GWR is that parameters may be estimated anywhere in the study area given a dependent variable and a set of one or more independent variables that have been measured at places whose location is known. GWR creates an equation for each cluster and uses data from nearby features to calibrate its equation. This makes the GWR coefficient take different values for each geographical area. As a result, a hotspot of coefficients, generated by the GWR model, provides means of geographical area (cluster) sensitive interventions. The equation for a typical GWR version of the OLS regression model would be:

Embedded Image

Where:

Yi=samples of outcome y.

(ui vi)=the coordinate points (longitude, latitude).

βk (ui vi) (k=0, 1, … p,)=are p unknown functions of geographical locations (ui vi).

xik=explanatory variables at location (ui vi).

i=1, 2, …n.

εi=residuals with zero mean and homogeneous variance σ2.

The corrected Akaike information criteria (AICc) and adjusted R2 were used to compare the final OLS model and GWR model. The model with the highest adjusted R2 value and the lowest AIC value was declared the best-fitting model for the data.

Patient and public involvement

None.

Result

The current study included 2922 women from 283 clusters. The overall proportion of inadequate counselling regarding PDSs in Ethiopia was 43.3% (95% CI 41.5% to 45.1%). The weighted proportion of inadequate counselling regarding PDSs was presented for variables ‘residence’ and ‘region’. Nearly, 871 (42%) of pregnant women who received inadequate counselling regarding danger signs were residents in rural parts of the country. The majority of women (70.4%) with inadequate counselling regarding danger signs of pregnancy were from the Somalia region. In contrast, 86% of women who received adequate counselling were from Addis Ababa (online supplemental table 1).

Spatial autocorrelation of inadequate counselling regarding PDSs

The spatial distribution of inadequate counselling regarding PDSs in Ethiopia was clustered (Moran’s I=0.23, p<0.01). The result shows that the observed Moran’s Index value was greater than the expected Index (−0.0035), and the p value was <0.01, which is statistically significant. Given the Z-score of 8.922821 indicates that there is a less than 1% likelihood that this clustered pattern could be the result of random chance (online supplemental figure 1).

Incremental spatial autocorrelation by distance

To determine the spatial clustering of inadequate counselling regarding PDSs, global spatial statistics were estimated using Moran’s I value. As shown in the figure below, statistically significant z-scores indicate at 242 km distances where spatial processes promoting clustering are most pronounced. The incremental spatial autocorrelation indicates that a total of 10 distance bands were detected with a beginning distance of approximately 155 192 m. The spatial distribution of inadequate counselling regarding PDSs among pregnant women in Ethiopia was found to be non-random with a Global Maran’s I was 0.23 and a p value of 0.0001. With a z-score of 8.92, there is a less than 1% likelihood that this high-clustered pattern could be the result of random chance (online supplemental figure 2).

Hotspot analysis of inadequate counselling regarding danger signs

The Getis-Ord GI* statistical analysis was performed to identify the hotspot and cold spot areas with inadequate counselling regarding PDSs. The dark and light green colour points indicate significant cold spot clusters with inadequate counselling regarding PDSs. According to the analysis, these areas are found in observed in the Northern part of the Gambella, the border between the Amhara and the Afar regions, the Dire Dawa city administration, the Northern Somali region and the northeastern part of the Oromia region. In contrast, the central and the southern parts of the Amhara region, the border between Addis Ababa and the Oromia region, the Northwestern part and the central Oromia region and the central and northeastern SNNP region are among the significant hotspot areas. This group is indicated by light and dark red colour points and are clusters with high proportions of inadequate counselling regarding PDSs (online supplemental figure 3).

Spatial scan statistical analysis

Purely spatial analysis was run to scan for clusters with both high and low rates of inadequate counselling regarding PDSs. A total of 11 most likely significant clusters were detected. Three of those were primary clusters with a high rate of inadequate counselling, while two clusters were secondary clusters with a high rate. The rest six clusters were areas with a low rate of inadequate counselling regarding PDSs. As illustrated in figure 1, the proportion of inadequate counselling regarding PDSs is higher among pregnant women inside red and yellow spatial scan circular windows than in pregnant women outside of the window. Significant primary high rate clusters were found in the Oromia region East Shoa zone centred at 8.651588 N, 39.118340 E with 65.24 km radius (relative risk, RR=1.54, log likelihood ratio, LLR=46.64, with p<0.001); Benishangul Gumuz region around Asosa centred at 10.171930 N, 34.585369 E with 43.79 km radius (RR=1.49, LLR=16.16, with p<0.001); and central Southern region centred at 6.540286 N, 36.627468 E with 164.21 km radius (RR=1.28, LLR=9.29, with p<0.01). Furthermore, the two significant secondary high rate clusters were observed in the Tigray region, somewhere between western and central Tigray zones centred at 14.100610 N, 38.304649 E with 385.65 km radius (RR=1.19, LLR=11.03, with p<0.01). And Eastern Tigray zone centred at 14.379220 N, 39.606689 E with 392.25 km radius (RR=1.18, LLR=9.54, with p=0.01). The red circular windows implies clusters with an LLR value between 15 and 47 with a pe<0.001. Clusters with yellow window represents an LLR value between 20 and 25 with a p<0.01 while the green windows represents clusters with an LLR value between 6 and 15 with a p value of 0.05 (figure 1).

Figure 1

Significant primary and secondary windows of inadequate counselling regarding pregnancy danger signs among pregnant women in Ethiopia, mEDHS 2019, Shape file source: (CSA 2021/2013 EC; https://africaopendata.org/dataset/ethiopia-shapefiles); Map output: Own analysis using ArcGIS Pro Software. CSA, Central Statistical Agency; LLR, log likelihood ratio; mEDHS, Mini Ethiopian Demographic and Health Survey. EC, Ethiopian Calendar

Factors affecting the spatial variation of short birth interval

The OLS model explained about 58.0% of the variation in inadequate counselling regarding PDSs (adjusted R2=0.58). The joint Wald Statistic indicated the overall model significance (p<0.01). As indicated by the VIF values, there was no multicollinearity (redundancy) among explanatory variables (VIF<7.5). The Koenker test was statistically significant, and thus, showed the relationship between the exploratory variables and the outcome variable was non-stationary (heterogeneous) across the study area. Hence, the need for a GWR analysis to account for the geographically varying effect of explanatory variables was ensured. The spatial autocorrelation (Moran’s I) test also revealed that residuals were not spatially autocorrelated (residuals were random) (p=0.6). The model residuals were not normally distributed as the Jarque-Bera statistics was significant with p<0.01. Robust probability was considered to select statistically significant variables for the GWR analysis (online supplemental table 2).

Independent predictors associated with inadequate counselling regarding PDSs were pregnant women whose ANC was provided by health extension workers, pregnant women whose first ANC was between the seventh and eighth month of their pregnancy, and pregnant women whose ANC follow-up was at governmental health centres (online supplemental table 3).

Geographically weighted regression

Unlike the OLS regression which is a global model, GWR is a local model that improves a model’s performance only if the relationship between the explanatory variables and the outcome variable (ie, inadequate counselling regarding PDSs) is non-stationary (heterogeneous) across geographical areas. As there was non-stationarity in our study, the adjusted R2 value gained from the OLS model (58%) was increased to 66% (adjusted R2=0.66). The benefit of moving from a global model (OLS) to a local regression model (GWR) was also assessed by comparing the AICc values of the two models. If the AICc value of the two models (OLS and GWR) differ by more than 3, the model with the lower AICc is declared to be better. In this study, the AICc value reduced from 1591.59 in the OLS model (online supplemental table 1) to 1560.32 in the GWR model (table 1), justifying that the GWR explains the spatial heterogeneity and improved the model better than the OLS model (table 1).

Table 1

Geographically weighted regression (GWR) and ordinary least square (OLS) model comparison for inadequate counselling regarding danger signs of pregnancy in Ethiopia, mEDHS 2019

Pregnant women whose ANC was provided by health extension workers had a significant positive relationship with inadequate counselling service regarding PDSs. Particularly, in the Southern Amhara region, Benishangul Gumuz region, Gambela region and the western part of Oromia region, as the proportion of ANC provided by HEW increases, inadequate counselling service regarding PDSs increases. The geographical locations with dark and light red-coloured points indicate areas where the coefficient of the variable ANC provided by HEW highly predicts the inadequate counselling service regarding PDSs (figure 2).

Figure 2

GWR coefficients of pregnant women whose ANC was provided by HEW for predicting inadequate counselling service in Ethiopia, mEDHS 2019, shape file source: (CSA 2021/2013 EC; https://africaopendata.org/dataset/ethiopia-shapefiles); Map output: Own analysis using ArcGIS Pro Software. ANC, antenatal care; CSA, Central Statistical Agency; GWR, geographically weighted regression; HEW, health extension worker; mEDHS, Mini Ethiopian Demographic and Health Survey. EC, Ethiopian Calendar

Pregnant women who started their first ANC follow-up between the seventh & eighth month of their pregnancy had a significant positive relationship with inadequate counselling service regarding PDSs. Figure 3 illustrates that in the northeastern part of the SNNP region, southern Afar region, Gambela region around the borders of Amhara region, and part of the Oromia region, inadequate counselling regarding PDSs was best predicted with the variable late ANC initiation by the pregnant women themselves (figure 3).

Figure 3

GWR coefficients of pregnant women with first ANC initiated between the seventh and eighth month of pregnancy for predicting inadequate counselling service in Ethiopia, mEDHS 2019, Shape file source: (CSA 2021/2013 EC; https://africaopendata.org/dataset/ethiopia-shapefiles); Map output: Own analysis using ArcGIS Pro Software. ANC, antenatal care; CSA, Central Statistical Agency; GWR, geographically weighted regression; mEDHS, Mini Ethiopian Demographic and Health Survey. EC, Ethiopian Calendar

The coefficients of pregnant women whose ANC follow-up was at a governmental health centre are demonstrated in figure 4. This variable has a strong positive association with inadequate counselling regarding PDSs in rural zones of the Somali region, a small part of the Oromia region that borders Dire Dawa city administration, and some part of the SNNP region that borders the Oromia region as well (figure 4).

Figure 4

GWR coefficients of pregnant women whose ANC follow-up was at a government health centre for predicting inadequate counselling service in Ethiopia, mEDHS 2019, Shape file source: (CSA 2021/2013 EC; https://africaopendata.org/dataset/ethiopia-shapefiles); Map output: Own analysis using ArcGIS Pro Software. ANC, antenatal care; CSA, Central Statistical Agency; GWR, geographically weighted regression; mEDHS, Mini Ethiopian Demographic and Health Survey. EC, Ethiopian Calendar

Discussion

Inadequate counselling of pregnant women regarding PDSs results in a lack of knowledge about when to seek EmOC during their pregnancy. It is a predominant contributor to two of the three delays that caused massive maternal mortality in low-income and middle-income countries, delay in deciding to seek care and delay in reaching to seek care on time.11 12

To assess the spatial distribution of inadequate counselling regarding PDSs, spatial data analysis methods such as spatial autocorrelation, hotspot and spatial scan analysis was executed. The spatial autocorrelation statistic confirmed that the distribution of inadequate counselling was clustered in a part of geographical areas (Moran’s I=0.23, p<0.01). The hotspot analysis identified areas with a low and high distribution of inadequate counselling regarding PDSs, whereas the spatial scan statistical analysis identified the most likely significant clusters, which had high inadequate counselling service. Overall, a statistically significant spatial variation of inadequate counselling service was found across regions of Ethiopia. This could be due to an intracountry variation in the availability of health facilities, health professionals, basic infrastructure, as well as health literacy level of pregnant women. According to empirical studies conducted in the country, there exists substantial socioeconomic inequalities in maternal health service provision and utilisation, particularly inequalities driven by differences in places of residence. It was found that pregnant women in less urbanised regions are less likely to receive the standard counselling service regarding PDSs.13 14

The SaTScan analysis identified three primary and two secondary significant clusters across the study area, implying that the proportion of inadequate counselling service regarding PDSs is higher among pregnant women residing inside the spatial scan circular window than in pregnant women outside of the circular window. Three significant primary clusters with high rates were detected in the Oromia region East Shoa zone, the Benishangul Gumuz region around Asosa and the central South Nation Nationalities and Peoples (SNNP) region. Likewise, the two significant secondary clusters with high rates were detected in the eastern zone of the Tigray region, as well as somewhere between the western and central Tigray zones. This clustering of inadequate counselling service in the mentioned locations could first be because some of these areas are located in the rural part of the regions and the rest are populated by nomadic pastoral populations who essentially do not have a permanent living area. This might prevent pregnant women from obtaining quality healthcare due to longer distances from health facilities, poor quality roads, lack of transport and lack of media exposure/information. In addition, proximate determinants such as attitude, values, perceptions, past experiences and health literacy have been identified as the driving factors affecting the utilisation of counselling services regarding PDSs in those geographical areas.15–17

There was a strong positive correlation between pregnant women whose ANC was provided by HEWs and hotspots of inadequate counselling regarding PDSs in the Southern Amhara region, the Benishangul Gumuz region, the Gambela region and some parts of the western part of the Oromia region. HEWs were recruited by the government of Ethiopia under the Health Services Extension Programme, which aimed to expand primary healthcare services for the rural population. In the country, the specific criteria to become HEW are being females at least 18 years old, having at least a 10th-grade education and speaking the local language. On completion of training after recruitment, a couple of HEWs are assigned as paid government employees to health posts (ie, the smallest healthcare delivery point in the country) in one kebeles (the smallest administrative unit). They work directly with individual households18 and according to a study conducted in 2019,19 HEWs provided ANC for 83% of pregnant women residing in rural areas.

Having said that, despite HEWs being considered skilled birth attendants, particularly in rural settings, ANC services provided by them were associated with inadequate counselling regarding PDSs. This could be explained by the fact that health professionals’ performance is affected more by skills obtained after training and motivation, which includes being appreciated by the supervisors, as well as having a stable income and opportunities for increased education and personal growth.20 Unfortunately, HEWs are known to be underpaid, understaffed and underappreciated by other healthcare professionals and the healthcare system of the country as well. This is worth further among HEWs employed in rural and underprivileged regions such as Gambella and Benishangul Gumuz who often operate under poor working conditions such as local conflicts. This might demotivate HEWs from sticking consistently to the ANC protocols during each visit and might have contributed to missing important services like counselling about PDSs. Moreover, since HEWs have a lower educational background and are relatively less trained than other skilled birth attendants such as midwives and doctors, the chance of them providing a better and more comprehensive ANC counselling service that incorporates details of PDSs will be lower.19 21 Hence, HEWs who work in the mentioned region of Ethiopia should be given priority for further ongoing, intensive and up-to-date training opportunities regarding the recommended ANC guidelines.

Pregnant women with a late first ANC (between the seventh and eighth month of pregnancy) were associated with increased inadequate counselling service regarding PDSs in a rural part of the SNNP region, the southern Afar region, some parts of the Gambela region where it borders the Amhara region and rural part of Oromia region. The alternative explanation for this relationship could be related to the setting under which the identified areas are classified, that is, rural, regional borders and pastoral areas. It is reasonable to assume that these types of areas have lesser access to health facilities because of poor health infrastructures and their nomadic lifestyles than urban areas. In addition, women living in rural and pastoral areas are associated with lower education and health literacy levels, not to mention their strong adherence to cultural and religious values. This would affect not only the women’s health-seeking behaviour and autonomy in making health-related decisions, including how soon to start ANC follow-up. It also affects their ability to afford the cost of medical healthcare in private health facilities that might be found around their residence.8 17 22 In addition, we believe this finding suggests a qualitative approach to identifying additional local factors that might affect pregnant women’s maternal health service utilisation practice.

The current study also identified a positive relationship between inadequate counselling regarding PDSs and pregnant women whose ANC follow-up was at government health centres in rural zones of the Somali region, a small part of the Oromia region that borders Dire Dawa. The first compelling justification for this finding could be the low number of health workforce and insufficiency of trained health professionals in the health centres found in those rural areas. Despite achievements in establishing a considerable number of health centres in these areas, there is a huge problem of job dissatisfaction and disengagement among health professionals working in those areas. The healthcare system in these identified locations of the country is suffering from job resignations, a shortage of health personnel, and an increasing number of intentions to leave the area. This is essentially due to the pastoral and rural nature of the areas where the health centres are located. As a result, health professionals complain about not having better living conditions, lack of regular promotions, poor second career opportunities, poor working environment and poor salaries that do not compensate for the fact that they are living in unurbanised settings. Studies conducted in the areas reported the above factors to negatively affect the maternal and child healthcare provision such as ANC service provision for pregnant women living in the geographical areas.23 24

The evident limitation of this study is that explanatory variables that could have increased the model’s ability to explain a bigger variation were not included in the study because the data were not collected in the 2019 mEDHS. These variables are the husband’s/partner’s education level, the husband’s/partner’s occupational status, and maternal occupation status. In addition, the estimated cluster effects of the spatial analysis could be affected because the geographical coordinates of clusters were displaced by the survey programme for up to 2 km in urban areas, 5 km for most clusters in rural areas and 10 km for 1% of clusters in rural areas, to prevent identification of respondents or the community. Therefore, the interpretation and/or conclusion based on this study are conditional on these limitations. In contrast to the limitations, this study used the latest, large-scale and nationally representative data which improves its representativeness. In addition, it used a robust design and spatial modelling technique to closely assess the geographically varying nature of predictors.

Conclusion and recommendation

Inadequate counselling regarding PDSs in Ethiopia varies across regions and there exists within-country inequality both in the service utilisation and provision. Inadequate counselling hotspot areas in the country were found in the Northern part of Gambella, a border between the Amhara and the Afar regions, Dire Dawa city administration, the Northern Somali region and northeastern parts of the Oromia region. Pregnant women whose ANC was provided by HEW, whose ANC follow-up was at government health centres and pregnant women with a late first ANC (between the seventh and eighth month of pregnancy) were correlated with inadequate counselling service regarding PDSs. Accordingly, this study recommends:

  • Concerned government bodies (ie, Ministry of Health, regional health bureaus and zone health departments) as well as other stakeholders to give ongoing, intensive and up-to-date training for HEWs about the importance of counselling pregnant women regarding the PDSs according to the recommended ANC guideline, particularly in the geographical areas where this predictor was identified.

  • Concerned government bodies should put an extra level of effort and dedication into increasing the job satisfaction and engagement of health professionals who work at health centres in rural and pastoral areas of the Somali region.

  • Key health system actors, particularly, health professionals organise and implement health education programmes and awareness campaigns that step up the public knowledge regarding the importance of earlier ANC visits and seek obstetric care as soon as needed.

  • Researchers should keep conducting ongoing research with the latest EDHS data as it grasps the potential to inform concerned actors to evaluate the impact of their interventions on the progress of counselling service regarding PDSs.

Data availability statement

Data are available in a public, open access repository. The dataset used for this study is publicly available at the MEASURE DHS programme website (https://www.dhsprogram.com/data).

Ethics statements

Patient consent for publication

Ethics approval

A formal online request was submitted to the DHS programme’s official website (https://www.dhsprogram.com/data) and ethical clearance and permission to use the dataset were obtained. The data used in this study are publicly available, aggregated secondary data that has not any personal identifying information that can be linked to study households. The confidentiality of data was maintained anonymously.

Acknowledgments

We would like to thank the MEASURE DHS programme for providing us with the data for further analysis.

References

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Footnotes

  • X @DersehNebiyu

  • Contributors MAA conceived and designed the study; performed data extraction, data analysis and interpretation and drafted the manuscript. NMD, TKT, HWA, GAY and MCA assisted with the design, analysis and interpretation of data and the critical review of the manuscript. All authors approved the final manuscript. The Guarantor of the study is MAA.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.