Spatial clusters distribution and modelling of health care autonomy among reproductive‐age women in Ethiopia: spatial and mixed‐effect logistic regression analysis

BMC Health Services Research, Volume 21, No. 1, Year 2021

Ukuhunyushwa

Background: While millions of women in many African countries have little autonomy in health care decision-making, in most low and middle-income countries, including Ethiopia, it has been poorly studied. Hence, it is important to have evidence on the factors associated with women’s health care decision making autonomy and the spatial distribution across the country. Therefore, this study aimed to investigate the spatial clusters distribution and modelling of health care autonomy among reproductive-age women in Ethiopia. Methods: We used the 2016 Ethiopian Demographic and Health Survey (EDHS) data for this study. The data were weighted for design and representativeness using strata, weighting variable, and primary sampling unit to get a reliable estimate. A total weighted sample of 10,223 married reproductive-age women were included in this study. For the spatial analysis, Arc-GIS version 10.6 was used to explore the spatial distribution of women health care decision making and spatial scan statistical analysis to identify hotspot areas. Considering the hierarchical nature of EDHS data, a generalized linear mixed-effect model (mixed-effect logistic regression) was fitted to identify significant determinants of women’s health care decision making autonomy. The Intra-Class Correlation (ICC) were estimated in the null model to estimate the clustering effect. For model comparison, deviance (-2LLR), Akakie Information Criteria (AIC), and Bayesian Information Criteria (BIC) parameters were used to choose the best-fitted model. Variables with a p-value < 0.2 in the bivariable analysis were considered for the multivariable analysis. In the multivariable mixed-effect logistic regression analysis, the Adjusted Odds Ratio (AOR) with 95% Confidence Interval (CI) were reported to declare the strength and significance of the association between women’s decision-making autonomy and independent variables. Results: In this study, about 81.6% (95% CI: 80.6%, 82.2%) of women have autonomy in making health care decisions. The spatial distribution of women’s autonomy in making health decisions in Ethiopia was non-random (global Moran’s I = 0.0675, p < 0.001). The significant hotspot areas of poor women’s autonomy in making health care decisions were found in north Somali, Afar, south Oromia, southwest Somali, Harari, and east Southern Nations Nationalities and Peoples (SNNP) regions. In the mixed-effect logistic regression analysis; being urban (AOR = 1.59, 95% CI: 1.04, 2.45), having secondary education (AOR = 1.60, 95% CI: 1.06, 2.41), having an occupation (AOR = 1.19, 95% CI: 1.01, 1.40) and being from the richest household (AOR = 2.14, 95% CI: 1.45, 3.14) were significantly associated with women autonomy in deciding for health care. Conclusions: The spatial distribution of women’s autonomy in making the decision for health care was non-random in Ethiopia. Maternal education, residence, household wealth status, region, and maternal occupation were found to influence women’s autonomy. Public health interventions targeting the hotspot areas of poor women autonomy through enhancing maternal occupation and employment is needed to improve women empowerment in making decisions for health care. This study was based on the most recent Demographic and Health Survey (DHS) of Ethiopia (EDHS 2016). Ethiopia is situated in the horn of Africa and it has nine regions (Afar, Amhara, Benishangul-Gumuz, Gambela, Harari, Oromia, Somali, Southern Nations, Nationalities, and People’s (SNNP) Region and Tigray) and two administrative cities (Addis Ababa and Dire-Dawa) (Fig. 1). A two-stage stratified sampling technique was employed to select the study participants using the 2007 Population and Housing Census (PHC) as a sampling frame. Overall a total of 21 sampling strata have been created. Around 645 Enumeration Areas (EAs) (202 in the urban area) were selected in the first stage and on average 28 households per EA were chosen in the second stage. All currently married reproductive-age women in Ethiopia were the source of population, whereas, all currently married women in the selected EAs were the study population. For this study, the Individual Record data (IR) set was used. A total of 10,223 currently married reproductive age women were included. The detailed sampling procedure and methodology were presented in the full EDHS 2016 report [33]. The map of regions of Ethiopia (Figure was generated using ArcGIS version 10.6 statistical software) The outcome variable for this study was women’s health decision making autonomy. In EDHS 2016 the question was asked as “person who usually decides on the respondent’s health care?“. The response for this question was respondent alone coded as “1”, jointly with their partner coded as “2”, and partner alone coded as “3”. Then we recode women who take health care decision alone or with their partner were coded 1 while respondents, where their partner alone decides for health care, were coded 0. Where “0” represents a woman with no health care decision making autonomy and “1” represents a woman with health care decision making autonomy. The independent variables considered were maternal age (15–24/25–34/35–49 years), residence (rural/urban), region, maternal occupational status (working/not working), women’s educational status (no/primary/secondary/higher), husband’s educational status (no/primary/secondary/higher), religion (Muslim/orthodox/protestant/other), frequency of watching television (not at all/less than once a week/at least once a week), frequency of listening radio (not at all/less than once a week/at least once a week), frequency of reding newspaper (not at all/less than once a week/at least once a week), household wealth status (poorest/poorer/middle/richer/richest), and covered by health insurance (no/yes). All the statistics reported in this paper were adjusted for complex survey designs to get a reliable estimate and to draw valid conclusions. STATA version 14, ArcGIS version 10.6, and SaTScan version 9.6 statistical software were used for analysis. The global spatial autocorrelation using Moran’s index was done to assess whether women’s autonomy in health care decision making was random or non-random. Moran’s index is a spatial statistic that measures the spatial autocorrelation of women’s autonomy in deciding on health care [34]. It produces a single output by taking the entire data sets (proportion of women autonomy in deciding for health care at the cluster level, latitude, longitude, and cluster ID). The global Moran’s I statistical analysis produces the Moran’s I value, Z-score, and p-values. Moran’s I value ranges from − 1 to 1 [35]. A value close to 1 shows a strong positive spatial autocorrelation of women’s autonomy in making decisions for health care and a value close to -1 shows a strong negative spatial autocorrelation (opposition between enumeration areas concerning the prevalence of women’s health care decision making autonomy). Moran’s, I value close to 0, indicates the spatial distribution of women autonomy in deciding for health care randomly distributed (independence between EAs). A statistically significant Moran’s I (p < 0.05) indicates the spatial distribution for women’s autonomy in deciding for health care is non-random. To predict the prevalence of health care decision-making autonomy in the un-sampled areas based on sampled EAs, the spatial interpolation method is used. In unobserved areas of Ethiopia, the Kriging spatial interpolation approach was used to predict the prevalence of women’s autonomy of health care decision-making. Kriging is used by forecasting it at unsampled locations (enumeration areas) to generate smooth maps of the outcome (women’s health care decision-making autonomy) and is an optimal interpolation based on regression against observed values of the surrounding data points, weighted according to the spatial covariance values. The Spatial Scan Statistical (SaTScan) analysis was done to identify significant hot spot areas of women’s autonomy in deciding on health care. The Bernoulli model was employed to identify the statistically significant spatial clusters of health care decision making autonomy of women using Kuldorff’s SaTScan version 9.6 statistical software. Since the elliptical window is inactive in the SaTScan software, we used a circular scanning window that moves across the study area to identify significant hot spot areas. The numbers of cases in each location had Bernoulli distribution and the model required data for cases, controls, and geographic coordinates. Women who are not participating in health care decision making were considered as cases and those who participated in making health care decisions as controls. The default maximum spatial cluster size of < 50% of the population was used as an upper limit, which allowed both small and large clusters to be detected and ignored clusters that contained more than the maximum limit. Selecting the cluster size of 50% of the total population is the default option for the maximum scanning window size and is often used to search the most likely clusters with a higher value of the test significance. Kuldorff’s indicated that a window-sized up to 50% of the population at risk can reduce negative clusters, highly sensitive, avoid missing clusters, and more likely to contain the true significant clusters than the small scanning window. For each potential cluster, a likelihood ratio test statistic and the p-value were used to determine if the number of observed women who are not participating in health care decision making within the potential cluster was significantly higher than expected or not. The scanning window with maximum likelihood was the most likely performing cluster, and the p-value was assigned to each cluster using Monte Carlo hypothesis testing by comparing the rank of the maximum likelihood from the real data with the maximum likelihood from the random datasets. The primary and secondary clusters were identified and assigned p-values and ranked based on their likelihood ratio test, based on 999 Monte Carlo replications [36]. The data source for this study was EDHS data. Standard models such as the logistic regression model are not appropriate these models are used for data that has a flat structure but EDHS has hierarchical nature (data collected at individual and community level). This implies that there is a need to take into account the between cluster variability by using advanced models such as mixed-effect binary logistic regression analysis. Therefore, a mixed effect logistic regression model (both fixed and random effect) was fitted. By fitting the standard logistic regression and mixed-effects logistic regression models, deviance (-2LLR), Akakie Information Criteria (AIC), and Bayesian Information Criteria (BIC) were used as a model comparison parameter. The Intra-class Correlation Coefficient (ICC), Likelihood Ratio (LR) test, and Median Odds Ratio (MOR) were done to assess the clustering effect and for assessing model fitness. Variables with a p-value < 0.2 in the bi-variable analysis were considered in the multivariable mixed-effect logistic regression model. The Adjusted Odds Ratios (AOR) with a 95% Confidence Interval (CI) and p-value ≤ 0.05 in the multivariable model were used to declare statistically significant factors associated with women’s autonomy in making health care decisions. Permission to get access to the data was obtained from the measure DHS program online request from http://www.dhsprogram.com.website and the data used were publicly available with no personal identifier.

I-AI Digest

Study Justification:

– The study aimed to investigate the spatial clusters distribution and modeling of health care autonomy among reproductive-age women in Ethiopia.
– It addressed the gap in knowledge regarding women’s health care decision-making autonomy in African countries, including Ethiopia.
– The study provided evidence on the factors associated with women’s health care decision-making autonomy and its spatial distribution across the country.

Highlights:

– 81.6% of women in Ethiopia have autonomy in making health care decisions.
– The spatial distribution of women’s autonomy in health care decision-making was non-random, with significant hotspot areas identified in north Somali, Afar, south Oromia, southwest Somali, Harari, and east Southern Nations Nationalities and Peoples (SNNP) regions.
– Factors significantly associated with women’s autonomy included urban residence, secondary education, occupation, and being from the richest household.
– Maternal education, residence, household wealth status, region, and maternal occupation were found to influence women’s autonomy.

Recommendations:

– Public health interventions should target the hotspot areas of poor women’s autonomy, particularly in north Somali, Afar, south Oromia, southwest Somali, Harari, and east SNNP regions.
– Enhancing maternal occupation and employment can contribute to improving women’s empowerment in making decisions for health care.

Key Role Players:

– Ministry of Health, Ethiopia
– Non-governmental organizations (NGOs) working in the field of women’s health and empowerment
– Community health workers
– Health care providers
– Researchers and academics specializing in women’s health and empowerment

Cost Items for Planning Recommendations:

– Training programs for community health workers and health care providers
– Awareness campaigns and educational materials targeting women and communities
– Support for income-generating activities and job opportunities for women
– Monitoring and evaluation activities to assess the impact of interventions
– Research funding for further studies and evaluations

Amandla Obufakazi

The strength of evidence for this abstract is 8 out of 10.
The evidence in the abstract is strong, but there are some areas for improvement. The study used a large sample size and a nationally representative dataset, which enhances the generalizability of the findings. The authors employed spatial analysis techniques to identify hotspot areas and assess the spatial distribution of women’s health care decision-making autonomy. They also used mixed-effect logistic regression to account for the hierarchical nature of the data. However, the abstract could be improved by providing more details on the statistical methods used, such as the specific tests and measures used for model comparison and assessing clustering effect. Additionally, it would be helpful to include information on potential limitations of the study and recommendations for future research.

Abstract

This study was based on the most recent Demographic and Health Survey (DHS) of Ethiopia (EDHS 2016). Ethiopia is situated in the horn of Africa and it has nine regions (Afar, Amhara, Benishangul-Gumuz, Gambela, Harari, Oromia, Somali, Southern Nations, Nationalities, and People’s (SNNP) Region and Tigray) and two administrative cities (Addis Ababa and Dire-Dawa) (Fig. 1). A two-stage stratified sampling technique was employed to select the study participants using the 2007 Population and Housing Census (PHC) as a sampling frame. Overall a total of 21 sampling strata have been created. Around 645 Enumeration Areas (EAs) (202 in the urban area) were selected in the first stage and on average 28 households per EA were chosen in the second stage. All currently married reproductive-age women in Ethiopia were the source of population, whereas, all currently married women in the selected EAs were the study population. For this study, the Individual Record data (IR) set was used. A total of 10,223 currently married reproductive age women were included. The detailed sampling procedure and methodology were presented in the full EDHS 2016 report [33]. The map of regions of Ethiopia (Figure was generated using ArcGIS version 10.6 statistical software) The outcome variable for this study was women’s health decision making autonomy. In EDHS 2016 the question was asked as “person who usually decides on the respondent’s health care?“. The response for this question was respondent alone coded as “1”, jointly with their partner coded as “2”, and partner alone coded as “3”. Then we recode women who take health care decision alone or with their partner were coded 1 while respondents, where their partner alone decides for health care, were coded 0. Where “0” represents a woman with no health care decision making autonomy and “1” represents a woman with health care decision making autonomy. The independent variables considered were maternal age (15–24/25–34/35–49 years), residence (rural/urban), region, maternal occupational status (working/not working), women’s educational status (no/primary/secondary/higher), husband’s educational status (no/primary/secondary/higher), religion (Muslim/orthodox/protestant/other), frequency of watching television (not at all/less than once a week/at least once a week), frequency of listening radio (not at all/less than once a week/at least once a week), frequency of reding newspaper (not at all/less than once a week/at least once a week), household wealth status (poorest/poorer/middle/richer/richest), and covered by health insurance (no/yes). All the statistics reported in this paper were adjusted for complex survey designs to get a reliable estimate and to draw valid conclusions. STATA version 14, ArcGIS version 10.6, and SaTScan version 9.6 statistical software were used for analysis. The global spatial autocorrelation using Moran’s index was done to assess whether women’s autonomy in health care decision making was random or non-random. Moran’s index is a spatial statistic that measures the spatial autocorrelation of women’s autonomy in deciding on health care [34]. It produces a single output by taking the entire data sets (proportion of women autonomy in deciding for health care at the cluster level, latitude, longitude, and cluster ID). The global Moran’s I statistical analysis produces the Moran’s I value, Z-score, and p-values. Moran’s I value ranges from − 1 to 1 [35]. A value close to 1 shows a strong positive spatial autocorrelation of women’s autonomy in making decisions for health care and a value close to -1 shows a strong negative spatial autocorrelation (opposition between enumeration areas concerning the prevalence of women’s health care decision making autonomy). Moran’s, I value close to 0, indicates the spatial distribution of women autonomy in deciding for health care randomly distributed (independence between EAs). A statistically significant Moran’s I (p < 0.05) indicates the spatial distribution for women’s autonomy in deciding for health care is non-random. To predict the prevalence of health care decision-making autonomy in the un-sampled areas based on sampled EAs, the spatial interpolation method is used. In unobserved areas of Ethiopia, the Kriging spatial interpolation approach was used to predict the prevalence of women’s autonomy of health care decision-making. Kriging is used by forecasting it at unsampled locations (enumeration areas) to generate smooth maps of the outcome (women’s health care decision-making autonomy) and is an optimal interpolation based on regression against observed values of the surrounding data points, weighted according to the spatial covariance values. The Spatial Scan Statistical (SaTScan) analysis was done to identify significant hot spot areas of women’s autonomy in deciding on health care. The Bernoulli model was employed to identify the statistically significant spatial clusters of health care decision making autonomy of women using Kuldorff’s SaTScan version 9.6 statistical software. Since the elliptical window is inactive in the SaTScan software, we used a circular scanning window that moves across the study area to identify significant hot spot areas. The numbers of cases in each location had Bernoulli distribution and the model required data for cases, controls, and geographic coordinates. Women who are not participating in health care decision making were considered as cases and those who participated in making health care decisions as controls. The default maximum spatial cluster size of < 50% of the population was used as an upper limit, which allowed both small and large clusters to be detected and ignored clusters that contained more than the maximum limit. Selecting the cluster size of 50% of the total population is the default option for the maximum scanning window size and is often used to search the most likely clusters with a higher value of the test significance. Kuldorff’s indicated that a window-sized up to 50% of the population at risk can reduce negative clusters, highly sensitive, avoid missing clusters, and more likely to contain the true significant clusters than the small scanning window. For each potential cluster, a likelihood ratio test statistic and the p-value were used to determine if the number of observed women who are not participating in health care decision making within the potential cluster was significantly higher than expected or not. The scanning window with maximum likelihood was the most likely performing cluster, and the p-value was assigned to each cluster using Monte Carlo hypothesis testing by comparing the rank of the maximum likelihood from the real data with the maximum likelihood from the random datasets. The primary and secondary clusters were identified and assigned p-values and ranked based on their likelihood ratio test, based on 999 Monte Carlo replications [36]. The data source for this study was EDHS data. Standard models such as the logistic regression model are not appropriate these models are used for data that has a flat structure but EDHS has hierarchical nature (data collected at individual and community level). This implies that there is a need to take into account the between cluster variability by using advanced models such as mixed-effect binary logistic regression analysis. Therefore, a mixed effect logistic regression model (both fixed and random effect) was fitted. By fitting the standard logistic regression and mixed-effects logistic regression models, deviance (-2LLR), Akakie Information Criteria (AIC), and Bayesian Information Criteria (BIC) were used as a model comparison parameter. The Intra-class Correlation Coefficient (ICC), Likelihood Ratio (LR) test, and Median Odds Ratio (MOR) were done to assess the clustering effect and for assessing model fitness. Variables with a p-value < 0.2 in the bi-variable analysis were considered in the multivariable mixed-effect logistic regression model. The Adjusted Odds Ratios (AOR) with a 95% Confidence Interval (CI) and p-value ≤ 0.05 in the multivariable model were used to declare statistically significant factors associated with women’s autonomy in making health care decisions. Permission to get access to the data was obtained from the measure DHS program online request from http://www.dhsprogram.com.website and the data used were publicly available with no personal identifier.

Izinto zokwakha

N/A

Emisha

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

1. Mobile Health (mHealth) Solutions: Develop mobile applications or text messaging services to provide pregnant women and new mothers with important health information, reminders for prenatal and postnatal care appointments, and access to telemedicine consultations.

2. Community Health Workers: Train and deploy community health workers to provide education, support, and basic healthcare services to pregnant women and new mothers in remote or underserved areas. These workers can also help identify high-risk pregnancies and refer women to appropriate healthcare facilities.

3. Telemedicine: Establish telemedicine networks to connect healthcare providers in urban areas with pregnant women and new mothers in rural or remote areas. This would allow for remote consultations, monitoring, and follow-up care, reducing the need for women to travel long distances for healthcare services.

4. Maternal Health Vouchers: Implement voucher programs that provide pregnant women with financial assistance to cover the costs of prenatal care, delivery, and postnatal care. This would help reduce financial barriers to accessing maternal healthcare services.

5. Maternal Waiting Homes: Set up maternal waiting homes near healthcare facilities in rural areas, where pregnant women can stay in the weeks leading up to their due date. This would ensure that women have timely access to skilled birth attendants and emergency obstetric care.

6. Transportation Solutions: Improve transportation infrastructure and services to ensure that pregnant women can easily access healthcare facilities. This could include providing affordable transportation options or establishing emergency transportation systems for women in labor.

7. Maternal Health Education Programs: Develop comprehensive maternal health education programs that target women, their families, and communities. These programs should focus on raising awareness about the importance of prenatal and postnatal care, nutrition, hygiene, and family planning.

8. Public-Private Partnerships: Foster collaborations between the government, private sector, and non-profit organizations to improve access to maternal healthcare. This could involve leveraging private sector resources and expertise to expand healthcare infrastructure and services in underserved areas.

9. Maternal Health Financing: Implement innovative financing mechanisms, such as microinsurance or community-based health financing schemes, to make maternal healthcare more affordable and accessible for low-income women.

10. Quality Improvement Initiatives: Implement quality improvement initiatives in healthcare facilities to ensure that pregnant women receive safe, respectful, and evidence-based care. This could involve training healthcare providers, improving infrastructure and equipment, and strengthening referral systems.

These innovations, along with a comprehensive approach that addresses social, cultural, and economic barriers, have the potential to improve access to maternal health and reduce maternal mortality rates in Ethiopia.

AI Innovations Description

The study mentioned in the description focuses on investigating the spatial clusters distribution and modeling of health care autonomy among reproductive-age women in Ethiopia. The goal is to understand the factors associated with women’s health care decision-making autonomy and identify hotspot areas with poor autonomy. The study used data from the 2016 Ethiopian Demographic and Health Survey (EDHS) and employed various statistical analyses, including spatial analysis and mixed-effect logistic regression.

The findings of the study revealed that approximately 81.6% of women in Ethiopia have autonomy in making health care decisions. The spatial distribution of women’s autonomy was found to be non-random, with significant hotspot areas of poor autonomy identified in certain regions of the country. Factors such as urban residence, secondary education, occupation, and being from a wealthier household were significantly associated with women’s autonomy in deciding for health care.

Based on these findings, the study suggests that public health interventions should target the hotspot areas with poor women’s autonomy. Enhancing maternal occupation and employment is recommended as a means to improve women’s empowerment in making decisions for health care. By addressing these factors, access to maternal health can be improved, leading to better health outcomes for women in Ethiopia.

It is important to note that the study utilized complex survey designs and statistical software such as STATA, ArcGIS, and SaTScan for data analysis. The methodology and detailed sampling procedure can be found in the full EDHS 2016 report. The study also used Moran’s index for global spatial autocorrelation analysis and Kriging spatial interpolation for predicting prevalence in unsampled areas. Additionally, the SaTScan analysis was conducted to identify significant hot spot areas.

The data used in the study were obtained from the EDHS, and permission to access the data was obtained from the Measure DHS program. The data used were publicly available and did not contain any personal identifiers.

AI Innovations Methodology

Based on the information provided, the study aimed to investigate the spatial clusters distribution and modeling of health care autonomy among reproductive-age women in Ethiopia. The methodology used in the study involved the following steps:

1. Data Source: The study used the 2016 Ethiopian Demographic and Health Survey (EDHS) data. The data were weighted for design and representativeness to ensure reliable estimates.

2. Sampling Technique: A two-stage stratified sampling technique was employed to select the study participants. Enumeration Areas (EAs) were selected in the first stage, and households were chosen in the second stage. A total of 10,223 currently married reproductive-age women were included in the study.

3. Outcome Variable: The outcome variable for the study was women’s health decision-making autonomy. The EDHS survey asked women who usually decides on their health care, and the responses were coded accordingly.

4. Independent Variables: Several independent variables were considered, including maternal age, residence, region, maternal occupational status, women’s educational status, husband’s educational status, religion, frequency of media exposure, household wealth status, and health insurance coverage.

5. Spatial Analysis: Arc-GIS version 10.6 was used to explore the spatial distribution of women’s health care decision-making. The global Moran’s I index was used to assess whether the spatial distribution was random or non-random. The Kriging spatial interpolation approach was used to predict the prevalence of health care decision-making autonomy in unsampled areas.

6. Spatial Scan Statistical Analysis: The SaTScan version 9.6 software was used to identify significant hotspot areas of women’s autonomy in deciding on health care. The Bernoulli model was employed, and a circular scanning window was used to detect clusters.

7. Statistical Analysis: A generalized linear mixed-effect model (mixed-effect logistic regression) was fitted to identify significant determinants of women’s health care decision-making autonomy. The model considered the hierarchical nature of the EDHS data. Model comparison parameters such as deviance, Akaike Information Criteria (AIC), and Bayesian Information Criteria (BIC) were used to select the best-fitted model.

8. Reporting Results: Adjusted Odds Ratios (AOR) with 95% Confidence Intervals (CI) were reported to declare the strength and significance of the association between women’s decision-making autonomy and independent variables.

In conclusion, the study utilized spatial analysis techniques, mixed-effect logistic regression modeling, and statistical tests to investigate the spatial clusters distribution and determinants of health care autonomy among reproductive-age women in Ethiopia. The findings highlighted significant hotspot areas and factors associated with women’s autonomy in making health care decisions.

Umbhali

Dima Health

Statistics:

Citations: 6

Identifiers:

DOI: 10.1186/s12913-020-06052-1

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, Cross Sectional Study

Study Approach:

Quantitative

Participants Gender:

Female

Yabelana ngalokhu: