Determinants and spatial distribution of institutional delivery in Ethiopia: evidence from Ethiopian Mini Demographic and Health Surveys 2019

Archives of Public Health, Volume 80, No. 1, Year 2022

Interpretation

Background: Over the past few decades, maternal and child mortality had drawn the attention of governments and policymakers. Institutional delivery has been among the implementations needed to reduce maternal and child mortality. The fact that the problem was persisted intensified studies to research for more factors. Thus, the current study was intended for further analyses of EMDHS to identify the magnitude, spatial patterns, and predictors of institutional delivery. Methods: A cross-sectional survey data from EMDHS 2019 was analyzed involving 5488 reproductive-age women regarding institutional deliveries. We presented descriptive statistics using mean, standard deviations, and proportions. To check the nature of the distribution of institutional delivery, we applied the global Moran’s I statistics. Getis-Ord Gi statistics was applied to detect spatial locations, and we applied spatial interpolation to predict unknown locations of institutional delivery using the Ordinary Kriging method. Kulldorff’s SatScan was also applied to identify the specific local clustering nature of institutional delivery using the Bernoulli method. We applied multilevel binary logistic regression for the scrutiny of individual and community-level factors. We applied P < 0.25 to include variables in the model and P < 0.05 to declare associations. AOR with 95% CI was used to describe variables. Results: The prevalence of institution/facility delivery was 2666.45(48.58%) in the survey. The average number of children was 4.03 ± 2.47, and most women in this survey were in the age range of the 25-29 years (31.84%) and 30–34 years (21.61%). Women who learned primary education (AOR = 1.52; 95% CI 1.20–1.95), secondary education (AOR = 1.77; 95% CI 1.03–3.07), and higher education (AOR = 5.41; 95% 1.91–15.25), while those who can read and write sentences (AOR = 1.94; 95% 1.28–2.94), Rich (AOR = 2.40 95% CI 1.82–3.16), and those followed 1–2 ANC (AOR = 2.08; 95% CI 1.57–2.76), 3 ANCs (AOR = 3.24; 95% CI 2.51–418), and ≥ 4 ANCs (AOR = 4.91; 95% CI 3.93–6.15) had higher odds of delivering at health institutions. Conclusion: The institutional delivery was unsatisfactory in Ethiopia, and there were various factors associated differently across the different regions. Pastoralist regions showed high home delivery than institutions which invites further interventions specific to those regions. Factors like age, highest education level achieved, preceding birth interval, literacy status, wealth status, birth order, regions, and rural residences were all affected institutional delivery so that interventions considering awareness, access, and availability of the services are vital. Ethiopia is the country located at (3o-14oN, 33o – 48°E). The country had undertaken four standard Demographic Health Surveys (EDHS). The country started EDHS in the year 2000 and conducted every five years since then. There were also two Ethiopian Mini Demographic Health Surveys (EMDHS) conducted in 2014 and 2019. EMDHS usually conduct between the standard EDHS (two to three years) after the EDHS conducted. The 2019 EMDHS is the second nationwide mini survey conducted in the country. In Ethiopian DHS, data has been collected using a two-level multistage stratified cluster sampling to pick eligible respondents from rural and urban areas. For the current analysis, we used Ethiopia Mini Demographic Health Survey (EMDHS) 2019 data. All nine regions and two city administrations were involved in the data collection. The regions were further categorized as agrarian (Benishangul-Gumuz Amhara, Southern Nations, Nationalities, and People Gambela, Oromia, Harari, Region (SNNPR), and Tigray), pastoralists (Afar and Somali), and city administrations (Addis Ababa and Dire-Dawa) contextually. We retrieved the data from the DHS website: (www.dhsprogram.com) after the measure program allowed us to download the datasets. The weighted sample became 5488 women who had live births in the last five years before the survey. They conducted the interview on the permanent residents and visitors who stayed the day before the survey in the residences, and it was a face-to-face manner [30]. The outcome variable for this study was the health institutions/facilities delivery, which was coded as “0” if the women gave birth at home and “1” if the women gave birth at a health facility. Institutions/facilities delivery was stated as the births at health institution/facility within five years afore the survey. Maternal education, maternal age, religion, ANC follow-up, sex of household head, literacy, the total number of children, birth order, preceding birth interval, the timing of 1st ANC visit, wealth index, and marital status were the variables. Region and place of residence. Before conducting the descriptive data analysis, we weighted the data to adjust the non-proportional allocation of samples to strata and regions. Then, descriptive statistics were presented using weighted and unweighted frequencies, mean ± (standard deviations), and percentage, while all analyses were performed using STATA version 15 (STATA Corporation. IC., TX, USA). The mean-variance inflation factor also was checked to be 3.53, which was in the acceptable range. For spatial analysis, we used ArcGIS 10.7 that determined the clustering, dispersion, and random distribution nature of the institutional delivery. Moran’s I output lies between (− 1 to + 1). The values close to − 1 indicated dispersed institutional delivery, and those closes to + 1 indicated clustering distribution. After discovering significant global autocorrelation, we tested the local Getis Ord statistics to identify the areas with high and low institutional deliveries [31]. For statistical optimization of the weight, the Ordinary Kriging spatial interpolation method was applied, and enabled us to make the prediction of institutional delivery for un-sampled areas of the country. SaTScan Version 9.6 software was used for the local cluster detection. A circular window that moves systematically throughout the study area was used to identify a significant clustering of institutional delivery. We presented the results of primary and secondary observed clusters using log-likelihood (LL) and p-value < 0.05. Since the data from country representative surveys are usually clustered or have hierarchical structure, we applied multilevel analysis. We went through four consecutive models building strategies to identify felicitous predictors of institutional delivery in the country. Model 0 is an empty/null (the intercept only model) existed before addition of the predictors. Model 1 (fixed effect model) included all individual-level variables that were initially significant at p-value of < 0.25 to determine the level of variance explained by the model. Model 2 (random effect model) included cluster-level (community -level) variables and model 3 (the mixed effect model) was the final model in which both the individual and community level variables introduced to judge final model performance. The log of the probability of the institutional delivery was modeled using multilevel binary logistic regression as: logπij1-πij=β0+β0Xij+β2Zij+uij; where, i and j are the level 1 (individual) and level 2 (community) units; X and Z refer to individual and community-level variables, in sequence. Πij is the probability of the institutional delivery for the ith mother in the jth community. We resolute random effect using Intra-community Correlation (ICC), ICC=σ2aσ2a+σ2b; where, σ2a is the community level variance and σ2b indicates individual level variance. The variance (σ2b) is equal to π23 which is the fixed value. Likelihood Ratio (LR) test for model comparison and deviance (−2LL) for the goodness of fit check were calculated, while Median Odds Ratio (MOR) and Proportional Change in Variance (PCV) were also estimated [32]. Finally, the mixed effect model, which included both fixed and random effect variables were fitted. To include variable in the model p-value < 0.25 and to declare association p-value< 0.05 were used. AOR with 95% CI was also used to articulate the results.

AI Digest

Study Justification:

– Maternal and child mortality are significant concerns for governments and policymakers.
– Institutional delivery is crucial for reducing maternal and child mortality.
– Further analysis of the Ethiopian Mini Demographic and Health Surveys (EMDHS) is needed to identify the factors influencing institutional delivery.

Highlights:

– The prevalence of institution/facility delivery in Ethiopia was 48.58%.
– Factors such as maternal education, literacy status, wealth status, and ANC follow-up were associated with higher odds of delivering at health institutions.
– Different regions showed variations in institutional delivery, with pastoralist regions having higher rates of home delivery.

Recommendations:

– Interventions should be tailored to address the specific needs of pastoralist regions, where home delivery rates are high.
– Awareness, access, and availability of services should be improved to encourage institutional delivery.
– Efforts should be made to increase education levels, particularly among women, to promote institutional delivery.

Key Role Players:

– Government health departments and ministries
– Non-governmental organizations (NGOs) working in maternal and child health
– Community health workers and volunteers
– Health facility staff and managers
– Researchers and academics in the field of maternal and child health

Cost Items for Planning Recommendations:

– Awareness campaigns and educational materials
– Training programs for healthcare providers and community health workers
– Infrastructure development and improvement of health facilities
– Transportation and logistics for reaching remote areas
– Monitoring and evaluation systems
– Research and data collection activities

Strength of Evidence

The strength of evidence for this abstract is 7 out of 10.
The evidence in the abstract is rated 7 because it provides a detailed description of the methods used and presents statistical results. However, it lacks information on the representativeness of the sample and potential biases. To improve the evidence, the abstract could include information on the sampling methodology, response rate, and any limitations of the study. Additionally, providing information on the generalizability of the findings to the larger population would enhance the strength of the evidence.

Abstract

Ethiopia is the country located at (3o-14oN, 33o – 48°E). The country had undertaken four standard Demographic Health Surveys (EDHS). The country started EDHS in the year 2000 and conducted every five years since then. There were also two Ethiopian Mini Demographic Health Surveys (EMDHS) conducted in 2014 and 2019. EMDHS usually conduct between the standard EDHS (two to three years) after the EDHS conducted. The 2019 EMDHS is the second nationwide mini survey conducted in the country. In Ethiopian DHS, data has been collected using a two-level multistage stratified cluster sampling to pick eligible respondents from rural and urban areas. For the current analysis, we used Ethiopia Mini Demographic Health Survey (EMDHS) 2019 data. All nine regions and two city administrations were involved in the data collection. The regions were further categorized as agrarian (Benishangul-Gumuz Amhara, Southern Nations, Nationalities, and People Gambela, Oromia, Harari, Region (SNNPR), and Tigray), pastoralists (Afar and Somali), and city administrations (Addis Ababa and Dire-Dawa) contextually. We retrieved the data from the DHS website: (www.dhsprogram.com) after the measure program allowed us to download the datasets. The weighted sample became 5488 women who had live births in the last five years before the survey. They conducted the interview on the permanent residents and visitors who stayed the day before the survey in the residences, and it was a face-to-face manner [30]. The outcome variable for this study was the health institutions/facilities delivery, which was coded as “0” if the women gave birth at home and “1” if the women gave birth at a health facility. Institutions/facilities delivery was stated as the births at health institution/facility within five years afore the survey. Maternal education, maternal age, religion, ANC follow-up, sex of household head, literacy, the total number of children, birth order, preceding birth interval, the timing of 1st ANC visit, wealth index, and marital status were the variables. Region and place of residence. Before conducting the descriptive data analysis, we weighted the data to adjust the non-proportional allocation of samples to strata and regions. Then, descriptive statistics were presented using weighted and unweighted frequencies, mean ± (standard deviations), and percentage, while all analyses were performed using STATA version 15 (STATA Corporation. IC., TX, USA). The mean-variance inflation factor also was checked to be 3.53, which was in the acceptable range. For spatial analysis, we used ArcGIS 10.7 that determined the clustering, dispersion, and random distribution nature of the institutional delivery. Moran’s I output lies between (− 1 to + 1). The values close to − 1 indicated dispersed institutional delivery, and those closes to + 1 indicated clustering distribution. After discovering significant global autocorrelation, we tested the local Getis Ord statistics to identify the areas with high and low institutional deliveries [31]. For statistical optimization of the weight, the Ordinary Kriging spatial interpolation method was applied, and enabled us to make the prediction of institutional delivery for un-sampled areas of the country. SaTScan Version 9.6 software was used for the local cluster detection. A circular window that moves systematically throughout the study area was used to identify a significant clustering of institutional delivery. We presented the results of primary and secondary observed clusters using log-likelihood (LL) and p-value < 0.05. Since the data from country representative surveys are usually clustered or have hierarchical structure, we applied multilevel analysis. We went through four consecutive models building strategies to identify felicitous predictors of institutional delivery in the country. Model 0 is an empty/null (the intercept only model) existed before addition of the predictors. Model 1 (fixed effect model) included all individual-level variables that were initially significant at p-value of < 0.25 to determine the level of variance explained by the model. Model 2 (random effect model) included cluster-level (community -level) variables and model 3 (the mixed effect model) was the final model in which both the individual and community level variables introduced to judge final model performance. The log of the probability of the institutional delivery was modeled using multilevel binary logistic regression as: logπij1-πij=β0+β0Xij+β2Zij+uij; where, i and j are the level 1 (individual) and level 2 (community) units; X and Z refer to individual and community-level variables, in sequence. Πij is the probability of the institutional delivery for the ith mother in the jth community. We resolute random effect using Intra-community Correlation (ICC), ICC=σ2aσ2a+σ2b; where, σ2a is the community level variance and σ2b indicates individual level variance. The variance (σ2b) is equal to π23 which is the fixed value. Likelihood Ratio (LR) test for model comparison and deviance (−2LL) for the goodness of fit check were calculated, while Median Odds Ratio (MOR) and Proportional Change in Variance (PCV) were also estimated [32]. Finally, the mixed effect model, which included both fixed and random effect variables were fitted. To include variable in the model p-value < 0.25 and to declare association p-value< 0.05 were used. AOR with 95% CI was also used to articulate the results.

Materials

N/A

Innovations

Based on the provided information, here are some potential innovations that can be used to improve access to maternal health:

1. Mobile Health (mHealth) Solutions: Develop mobile applications or text messaging services to provide pregnant women with information on prenatal care, nutrition, and the importance of institutional delivery. These platforms can also be used to schedule appointments and send reminders for antenatal care visits.

2. Telemedicine: Implement telemedicine services to connect pregnant women in remote or underserved areas with healthcare providers. This allows for virtual consultations, remote monitoring of maternal health, and timely access to medical advice.

3. Community Health Workers (CHWs): Train and deploy CHWs to provide education and support to pregnant women in their communities. CHWs can conduct home visits, provide counseling on the benefits of institutional delivery, and assist with referrals to healthcare facilities.

4. Maternal Waiting Homes: Establish maternal waiting homes near healthcare facilities to accommodate pregnant women who live far away. These homes provide a safe and comfortable place for women to stay during the final weeks of pregnancy, ensuring they are close to the facility when labor begins.

5. Transportation Support: Develop transportation initiatives to address the challenges of accessing healthcare facilities. This can include providing subsidized transportation vouchers or partnering with local transportation services to ensure pregnant women have reliable and affordable transportation options.

6. Quality Improvement Initiatives: Implement quality improvement programs in healthcare facilities to enhance the overall experience of pregnant women. This can involve training healthcare providers on respectful maternity care, improving infrastructure and equipment, and ensuring the availability of essential supplies and medications.

7. Financial Incentives: Introduce financial incentives, such as conditional cash transfers or maternity vouchers, to encourage pregnant women to seek institutional delivery. These incentives can help offset the costs associated with transportation, facility fees, and other expenses.

8. Public-Private Partnerships: Foster collaborations between the public and private sectors to expand access to maternal health services. This can involve leveraging the resources and expertise of private healthcare providers, pharmaceutical companies, and technology companies to improve service delivery and reach more women.

It is important to note that the implementation of these innovations should be context-specific and tailored to the needs and challenges of the Ethiopian healthcare system.

AI Innovations Description

Based on the description provided, the recommendation to improve access to maternal health in Ethiopia is to implement targeted interventions that address the identified factors associated with institutional delivery. These interventions should focus on increasing awareness, improving access, and ensuring the availability of maternal health services.

Specific recommendations based on the findings of the study include:

1. Education: Promote and prioritize education for women, particularly at the primary, secondary, and higher education levels. This can be achieved through awareness campaigns, scholarships, and incentives to encourage women to pursue education.

2. Antenatal Care (ANC) follow-up: Encourage women to attend ANC visits, with a particular emphasis on ensuring that women receive at least 3 or more ANC visits. This can be achieved through community outreach programs, mobile clinics, and education on the importance of ANC.

3. Literacy and awareness: Improve literacy rates among women and promote awareness of the benefits of institutional delivery. This can be done through community-based education programs, media campaigns, and the involvement of community leaders and influencers.

4. Wealth status: Address the disparities in institutional delivery by targeting interventions towards women from lower socio-economic backgrounds. This can include providing financial support for transportation to health facilities, reducing or eliminating out-of-pocket expenses for maternal health services, and implementing social protection programs.

5. Regional and rural disparities: Develop region-specific interventions that address the unique challenges faced by pastoralist regions and rural areas. This can include mobile health clinics, community health workers, and infrastructure improvements to ensure access to health facilities.

6. Age and birth order: Provide targeted support and education for younger women and those with higher birth orders to ensure they have access to maternal health services. This can include peer education programs, mentorship, and counseling services.

7. Strengthen health systems: Invest in improving the capacity and quality of health facilities, particularly in underserved areas. This can include training healthcare providers, improving infrastructure, and ensuring the availability of essential equipment and supplies.

By implementing these recommendations, it is expected that access to maternal health services will improve, leading to a reduction in maternal and child mortality rates in Ethiopia.

AI Innovations Methodology

Based on the information provided, here are some potential recommendations to improve access to maternal health:

1. Increase awareness and education: Implement programs to educate women and communities about the importance of institutional delivery and the benefits of accessing maternal health services. This can include campaigns, workshops, and community outreach programs.

2. Improve access to healthcare facilities: Increase the number of healthcare facilities, especially in rural and remote areas, to ensure that women have access to quality maternal health services. This can involve building new facilities, upgrading existing ones, and providing necessary equipment and supplies.

3. Strengthen antenatal care services: Enhance antenatal care services to ensure that women receive adequate prenatal care and are informed about the importance of institutional delivery. This can include increasing the number of antenatal care visits, providing comprehensive screenings and tests, and offering counseling and support.

4. Address socio-economic barriers: Implement measures to address socio-economic barriers that prevent women from accessing maternal health services. This can include providing financial assistance, reducing out-of-pocket expenses, and improving transportation infrastructure to facilitate access to healthcare facilities.

To simulate the impact of these recommendations on improving access to maternal health, a methodology could be as follows:

1. Data collection: Collect data on the current state of maternal health access, including the prevalence of institutional delivery, socio-economic factors, and geographical distribution of healthcare facilities.

2. Spatial analysis: Use spatial analysis techniques, such as Moran’s I statistics and Getis-Ord Gi statistics, to identify spatial patterns and clustering of institutional delivery. This will help identify areas with low access to maternal health services.

3. Spatial interpolation: Apply spatial interpolation methods, such as Ordinary Kriging, to predict the unknown locations of institutional delivery. This will help estimate the level of access in areas where data is missing or incomplete.

4. Cluster detection: Use SaTScan software to identify specific local clusters of institutional delivery. This will help identify areas where interventions can be targeted to improve access.

5. Multilevel regression analysis: Conduct multilevel binary logistic regression analysis to identify the factors that influence institutional delivery. This will help determine the individual and community-level factors that need to be addressed in interventions.

6. Model comparison and evaluation: Compare different models using likelihood ratio tests and deviance to assess the goodness of fit. Calculate the median odds ratio and proportional change in variance to evaluate the impact of the interventions on improving access to maternal health.

7. Interpretation and recommendations: Analyze the results and interpret the findings to provide recommendations for improving access to maternal health. These recommendations should be based on the identified factors and patterns, taking into account the specific context of Ethiopia.

It is important to note that this methodology is a general framework and may need to be adapted based on the specific research objectives and available data.

Author

Dima Health

Statistics:

Citations: 1

Identifiers:

DOI: 10.1186/s13690-022-00825-2

ISSN: 07787367

Research Areas:

Community Interventions, Health System and Policy, Maternal Access, Maternal and Child Health, Quality of Care, Sexual and Reproductive Health, Social Determinants

Study Countries:

Ethiopia

Study Design:

Case-Control Study, Cohort Study, Cross Sectional Study, Grounded Theory

Study Approach:

Quantitative

Participants Gender:

Female