Community-based health financing and child stunting in rural Rwanda

American Journal of Public Health, Volume 106, No. 1, Year 2016

Ukuhunyushwa

Objectives. We analyzed the likelihood of rural children (aged 6-24 months) being stunted according to whether they were enrolled in Mutuelles, a community-based health-financing program providing health insurance to rural populations and granting them access to health care, including nutrition services. Methods. We retrieved health facility data from the District Health System Strengthening Tool and calculated the percentage of rural health centers that provided nutrition-related services required by Mutuelles’ minimum service package. We used data from the 2010 Rwanda Demographic and Health Survey and performed multilevel logistic analysis to control for clustering effects and sociodemographic characteristics. The final sample was 1061 children. Results. Among 384 rural health centers, more than 90% conducted nutrition-related campaigns and malnutrition screening for children. Regardless of poverty status, the risk of being stunted was significantly lower (odds ratio = 0.60; 95% credible interval = 0.41, 0.83) for Mutuelles enrollees. This finding was robust to various model specifications (adjusted for Mutuelles enrollment, poverty status, other variables) or estimation methods (fixed and random effects). Conclusions. This study provides evidence of the effectiveness of Mutuelles in improving child nutrition status and supported the hypothesis about the role of Mutuelles in expanding medical and nutritional care coverage for children. Our target population was rural children in Rwanda, who had a much higher prevalence of stunting than urban children in 2010 (47% vs 27%).25 We focused on rural children aged 6 to 24 months for the following reasons. First, it has been suggested that stunting is hard to reverse for children older than 2 years, and the first 2 years of life is the “window of opportunity” for interventions against stunting.2,31–34 Second, about 93% of rural children younger than 6 months had exclusive breastfeeding, and the prevalence of stunting for this age group was much lower (18%) than for the children aged 6 to 24 months (42%).25 We used data from the RDHS 2010 to conduct an individual-level analysis. The RDHS is a nationally representative, population-based survey conducted every 5 years to measure indicators of population health and nutrition, with a special emphasis on mothers and on children younger than 5 years. The RDHS also collects information on households’ and mothers’ sociodemographic characteristics, health insurance status, and utilization of health care. It has been widely used to provide national and regional evidence to policymakers in Rwanda.27,35,36 The RDHS 2010 collected individual information through a 2-stage sampling process. Villages, or primary sampling units, were selected at the first stage. Households in the selected villages were chosen at the second stage.37 We used the DHHST to study the availability of nutritional services included in the MSP across rural health centers in Rwanda. The DHHST is an ongoing Web-based database system built by the Rwandan Ministry of Health in 2009 for monitoring and strategic planning on strengthening health systems.36 The DHHST requires rural health centers to report to the database on an annual basis. The data provide information on (1) medical services (including nutritional care), (2) capacity building (infrastructure and staffing), and (3) revenues and expenditures. The RDHS 2010 measured the heights of children younger than 5 years from a randomly selected 50% subsample of households. Among the 3474 rural children younger than 5 years with a valid height measure, 1087 were aged 6 to 24 months. To identify the link between children’s Mutuelles membership and their nutritional status, we included only children who were either enrolled in the Mutuelles program or had no insurance. We excluded 23 rural children who reported other types of insurance. The final sample size was 1061 children: 838 enrolled in Mutuelles and 223 uninsured. In the DHHST data, of 389 rural health centers in Rwanda, 384 reported service provision in 2010. For individual-level analysis, we constructed a dichotomous variable to indicate whether a child in our study population was stunted, defined as height-for-age 2 standard deviations below the median of the international reference population recommended by the World Health Organization in 2006.38 Facility-level analysis identified a list of services that were included in the MSP of the Mutuelles and available in rural health centers. We treated each variable as binary and assigned a value of 0 or 1. The DHHST surveys included 30 questions on nutrition services in 2010: 10 promotional services, 7 preventive services, and 13 curative services. Online Panel B (available as a supplement to the online version of this article at http://www.ajph.org) lists these nutrition services and how they were delivered by community health workers. For individual-level analysis, we constructed a dichotomous variable to indicate whether a child was enrolled in Mutuelles or uninsured. According to the Mutuelles legislation enacted in 2008,35 enrollees were entitled to access the nutritional care (listed in online Panel B) when the services were available in rural health centers. Sociodemographic variables included a child’s gender, maternal characteristics (mother’s age, education, and height), and household wealth status. We constructed 2 dummy variables that accounted for a mother’s completion of primary school and age older than 30 years. Previous studies found that a mother’s height was significantly associated with her child’s stunting status.39 Following Özaltin et al.,39 we constructed a categorical variable to indicate the mother’s height (< 150.0 cm, 150.0–154.9 cm, 155.0–159.9 cm, or ≥ 160.0 cm). The RDHS 2010 had a wealth quintile variable that summarized a household’s assets (e.g., motorcycle), housing construction (e.g., floor), water source, and sanitation. Because about 45% of Rwanda’s rural population lived below the national poverty line (defined as US $0.45 per day per adult) in 2010,24 we regrouped the households into 2 groups: below the poverty line (households in the lowest and next-to-lowest wealth quintile) and above the poverty line (all others). We constructed a dummy variable to indicate whether a child was from a household living below the poverty line. Note that information about a household’s water and sanitation was reflected in its wealth quintile, and we did not construct separate variables to represent these factors. To determine whether Mutuelles enrollment was associated with the nutritional status of children living below the poverty line versus those living above the poverty line, we constructed an interaction variable between Mutuelles enrollment and poverty indicators. To control for district-level heterogeneity of health systems at the individual-level analysis, we used the reported number of community health workers in the DHSST to construct a variable indicating the number of district-level community health workers per capita in the 27 rural districts. As described in online Panel B, community health workers play a key role in implementing nutritional services in rural areas. We used a multilevel logistic regression model in individual-level analysis to estimate the association between children’s Mutuelles status and their likelihood of being stunted, controlling for child’s gender, poverty status, maternal factors (age, education, and height), and district- and village-level clustering effects. The multilevel model enabled us to control for clustering effects and correct for standard errors in the higher levels so as to obtain coefficients with more efficiency.40 Because of the process by which the RDHS sampled villages at the first stage, children from the same village could be more similar to each other than to children from other villages. As a result of decentralized health care delivery at the district level in Rwanda, children in the same district were more likely to share similarities to each other than to children of other districts.41 Ignoring these clustering effects could lead to underestimation of standard errors and assumption of statistical significance where it does not exist.42 To control for clustering effects, we included random effects at the village and the district level. More details on modeling are presented in the online Supplementary Methods (available as a supplement to the online version of this article at http://www.ajph.org). We applied a standard 2-step procedure recommended for discrete-outcomes multilevel models and started with a first-order marginal quasi-likelihood approach to get crude estimates, followed by a Bayesian Markov chain Monte Carlo approach in the second stage to improve the approximations.40,43 Details can be found in the online Supplementary Methods (see also Figures A and B, available as a supplement to the online version of this article at http://www.ajph.org). We report odd ratios and credible intervals. The credible intervals derived from the Markov chain Monte Carlo method indicate that, with 100 000 simulations, the true estimate will lie in the credible interval with a probability of 95% (Table A, available as a supplement to the online version of this article at http://www.ajph.org).43,44 We used the Stata version 14 command runmlwin (StataCorp LP, College Station, TX) to perform the multilevel statistical analysis.45–47 Because the inclusion of different covariates and the use of different estimation methods may alter the association between the exposure and the outcome variable, we examined the sensitivity of the results to model specifications. We estimated the main association between Mutuelles status and stunting (with or without covariates) using the logistic model with multilevel random effects (model 1 to model 3 in Table B, available as a supplement to the online version of this article at http://www.ajph.org). We then estimated the association between Mutuelles status and stunting among children in households below versus above the poverty line by adding the interaction variable between Mutuelles and poverty indicators (model 4 in Table B). We also tested the sensitivity of results to estimation methods by conducting an analysis with a fixed-effects model including the 27 district indicators in regression analysis. A fixed-effects model generates less biased estimates, whereas a random-effects model generates more efficient estimates.48

I-AI Digest

Study Justification:

– The study aimed to analyze the impact of a community-based health-financing program, called Mutuelles, on the likelihood of rural children in Rwanda being stunted.
– The prevalence of stunting in rural children was much higher than in urban children, highlighting the need for targeted interventions.
– The first 2 years of life are crucial for interventions against stunting, and it is suggested that stunting is hard to reverse for children older than 2 years.
– The study used nationally representative data from the Rwanda Demographic and Health Survey (RDHS) 2010, which is widely used to inform policymakers in Rwanda.

Highlights:

– More than 90% of rural health centers in Rwanda provided nutrition-related campaigns and malnutrition screening for children.
– Regardless of poverty status, children enrolled in Mutuelles had a significantly lower risk of being stunted.
– The findings were robust to various model specifications and estimation methods, supporting the effectiveness of Mutuelles in improving child nutrition status.

Recommendations:

– Expand the coverage and accessibility of Mutuelles to ensure more rural children have access to health care, including nutrition services.
– Strengthen the implementation of nutrition-related campaigns and malnutrition screening in rural health centers to further improve child nutrition status.
– Continue monitoring and evaluating the impact of Mutuelles on child stunting to inform future interventions and policies.

Key Role Players:

– Rwandan Ministry of Health: Responsible for overseeing the implementation of Mutuelles and coordinating with rural health centers.
– Community Health Workers: Play a key role in implementing nutritional services in rural areas, as identified in the District Health System Strengthening Tool (DHHST) data.
– District Health Systems: Responsible for ensuring the availability and quality of health services, including nutrition services, in rural health centers.

Cost Items for Planning Recommendations:

– Expansion of Mutuelles coverage: Budget for enrolling more rural children in the program and providing them with health insurance.
– Strengthening nutrition services in rural health centers: Budget for training and capacity building of health workers, procurement of necessary equipment and supplies, and monitoring and evaluation activities.
– Monitoring and evaluation: Budget for data collection, analysis, and reporting to assess the impact of Mutuelles on child stunting and inform future interventions.

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 design includes a large sample size (1061 children) and uses data from a nationally representative survey. The analysis controls for clustering effects and sociodemographic characteristics. The findings are consistent across different model specifications and estimation methods. However, the abstract could be improved by providing more information on the statistical significance of the results and the magnitude of the effect. Additionally, it would be helpful to include information on potential limitations of the study, such as any confounding factors that were not accounted for. Overall, the evidence is strong, but providing more details and addressing potential limitations would further strengthen the abstract.

Abstract

Our target population was rural children in Rwanda, who had a much higher prevalence of stunting than urban children in 2010 (47% vs 27%).25 We focused on rural children aged 6 to 24 months for the following reasons. First, it has been suggested that stunting is hard to reverse for children older than 2 years, and the first 2 years of life is the “window of opportunity” for interventions against stunting.2,31–34 Second, about 93% of rural children younger than 6 months had exclusive breastfeeding, and the prevalence of stunting for this age group was much lower (18%) than for the children aged 6 to 24 months (42%).25 We used data from the RDHS 2010 to conduct an individual-level analysis. The RDHS is a nationally representative, population-based survey conducted every 5 years to measure indicators of population health and nutrition, with a special emphasis on mothers and on children younger than 5 years. The RDHS also collects information on households’ and mothers’ sociodemographic characteristics, health insurance status, and utilization of health care. It has been widely used to provide national and regional evidence to policymakers in Rwanda.27,35,36 The RDHS 2010 collected individual information through a 2-stage sampling process. Villages, or primary sampling units, were selected at the first stage. Households in the selected villages were chosen at the second stage.37 We used the DHHST to study the availability of nutritional services included in the MSP across rural health centers in Rwanda. The DHHST is an ongoing Web-based database system built by the Rwandan Ministry of Health in 2009 for monitoring and strategic planning on strengthening health systems.36 The DHHST requires rural health centers to report to the database on an annual basis. The data provide information on (1) medical services (including nutritional care), (2) capacity building (infrastructure and staffing), and (3) revenues and expenditures. The RDHS 2010 measured the heights of children younger than 5 years from a randomly selected 50% subsample of households. Among the 3474 rural children younger than 5 years with a valid height measure, 1087 were aged 6 to 24 months. To identify the link between children’s Mutuelles membership and their nutritional status, we included only children who were either enrolled in the Mutuelles program or had no insurance. We excluded 23 rural children who reported other types of insurance. The final sample size was 1061 children: 838 enrolled in Mutuelles and 223 uninsured. In the DHHST data, of 389 rural health centers in Rwanda, 384 reported service provision in 2010. For individual-level analysis, we constructed a dichotomous variable to indicate whether a child in our study population was stunted, defined as height-for-age 2 standard deviations below the median of the international reference population recommended by the World Health Organization in 2006.38 Facility-level analysis identified a list of services that were included in the MSP of the Mutuelles and available in rural health centers. We treated each variable as binary and assigned a value of 0 or 1. The DHHST surveys included 30 questions on nutrition services in 2010: 10 promotional services, 7 preventive services, and 13 curative services. Online Panel B (available as a supplement to the online version of this article at http://www.ajph.org) lists these nutrition services and how they were delivered by community health workers. For individual-level analysis, we constructed a dichotomous variable to indicate whether a child was enrolled in Mutuelles or uninsured. According to the Mutuelles legislation enacted in 2008,35 enrollees were entitled to access the nutritional care (listed in online Panel B) when the services were available in rural health centers. Sociodemographic variables included a child’s gender, maternal characteristics (mother’s age, education, and height), and household wealth status. We constructed 2 dummy variables that accounted for a mother’s completion of primary school and age older than 30 years. Previous studies found that a mother’s height was significantly associated with her child’s stunting status.39 Following Özaltin et al.,39 we constructed a categorical variable to indicate the mother’s height (< 150.0 cm, 150.0–154.9 cm, 155.0–159.9 cm, or ≥ 160.0 cm). The RDHS 2010 had a wealth quintile variable that summarized a household’s assets (e.g., motorcycle), housing construction (e.g., floor), water source, and sanitation. Because about 45% of Rwanda’s rural population lived below the national poverty line (defined as US $0.45 per day per adult) in 2010,24 we regrouped the households into 2 groups: below the poverty line (households in the lowest and next-to-lowest wealth quintile) and above the poverty line (all others). We constructed a dummy variable to indicate whether a child was from a household living below the poverty line. Note that information about a household’s water and sanitation was reflected in its wealth quintile, and we did not construct separate variables to represent these factors. To determine whether Mutuelles enrollment was associated with the nutritional status of children living below the poverty line versus those living above the poverty line, we constructed an interaction variable between Mutuelles enrollment and poverty indicators. To control for district-level heterogeneity of health systems at the individual-level analysis, we used the reported number of community health workers in the DHSST to construct a variable indicating the number of district-level community health workers per capita in the 27 rural districts. As described in online Panel B, community health workers play a key role in implementing nutritional services in rural areas. We used a multilevel logistic regression model in individual-level analysis to estimate the association between children’s Mutuelles status and their likelihood of being stunted, controlling for child’s gender, poverty status, maternal factors (age, education, and height), and district- and village-level clustering effects. The multilevel model enabled us to control for clustering effects and correct for standard errors in the higher levels so as to obtain coefficients with more efficiency.40 Because of the process by which the RDHS sampled villages at the first stage, children from the same village could be more similar to each other than to children from other villages. As a result of decentralized health care delivery at the district level in Rwanda, children in the same district were more likely to share similarities to each other than to children of other districts.41 Ignoring these clustering effects could lead to underestimation of standard errors and assumption of statistical significance where it does not exist.42 To control for clustering effects, we included random effects at the village and the district level. More details on modeling are presented in the online Supplementary Methods (available as a supplement to the online version of this article at http://www.ajph.org). We applied a standard 2-step procedure recommended for discrete-outcomes multilevel models and started with a first-order marginal quasi-likelihood approach to get crude estimates, followed by a Bayesian Markov chain Monte Carlo approach in the second stage to improve the approximations.40,43 Details can be found in the online Supplementary Methods (see also Figures A and B, available as a supplement to the online version of this article at http://www.ajph.org). We report odd ratios and credible intervals. The credible intervals derived from the Markov chain Monte Carlo method indicate that, with 100 000 simulations, the true estimate will lie in the credible interval with a probability of 95% (Table A, available as a supplement to the online version of this article at http://www.ajph.org).43,44 We used the Stata version 14 command runmlwin (StataCorp LP, College Station, TX) to perform the multilevel statistical analysis.45–47 Because the inclusion of different covariates and the use of different estimation methods may alter the association between the exposure and the outcome variable, we examined the sensitivity of the results to model specifications. We estimated the main association between Mutuelles status and stunting (with or without covariates) using the logistic model with multilevel random effects (model 1 to model 3 in Table B, available as a supplement to the online version of this article at http://www.ajph.org). We then estimated the association between Mutuelles status and stunting among children in households below versus above the poverty line by adding the interaction variable between Mutuelles and poverty indicators (model 4 in Table B). We also tested the sensitivity of results to estimation methods by conducting an analysis with a fixed-effects model including the 27 district indicators in regression analysis. A fixed-effects model generates less biased estimates, whereas a random-effects model generates more efficient estimates.48

Izinto zokwakha

N/A

Emisha

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

1. Community-based health financing: Implementing community-based health financing programs, such as Mutuelles in rural Rwanda, can provide health insurance to rural populations and grant them access to health care, including nutrition services. This can help improve access to maternal health services and reduce the risk of child stunting.

2. Strengthening health systems: Investing in the strengthening of health systems, particularly in rural areas, can improve the availability and quality of maternal health services. This can include improving infrastructure, staffing, and capacity building in rural health centers.

3. Promoting nutrition services: Conducting nutrition-related campaigns and providing malnutrition screening for children in rural areas can help identify and address nutritional deficiencies early on. This can be done through community health workers and other healthcare providers.

4. Increasing awareness and education: Implementing educational programs and awareness campaigns about the importance of maternal health and nutrition can help improve knowledge and understanding among rural communities. This can empower individuals to seek appropriate care and make informed decisions regarding their health.

5. Integrating maternal health services: Integrating maternal health services with other healthcare services, such as family planning and child health services, can improve access and continuity of care for women and children. This can be done through the establishment of integrated health centers or through mobile health clinics.

6. Leveraging technology: Utilizing technology, such as telemedicine and mobile health applications, can help overcome geographical barriers and improve access to maternal health services in remote areas. This can enable women to receive virtual consultations, access health information, and receive reminders for prenatal and postnatal care.

7. Empowering women and communities: Promoting women’s empowerment and community engagement can help improve access to maternal health services. This can involve training and supporting community health workers, involving women in decision-making processes, and addressing cultural and social barriers that may hinder access to care.

It is important to note that the effectiveness of these innovations may vary depending on the context and specific needs of the population. Therefore, further research and evaluation are necessary to determine the most appropriate and effective strategies for improving access to maternal health in different settings.

AI Innovations Description

The recommendation that can be developed into an innovation to improve access to maternal health is the implementation of a community-based health financing program, similar to the Mutuelles program in rural Rwanda. This program provides health insurance to rural populations, granting them access to health care, including nutrition services. The study mentioned in the description found that children enrolled in the Mutuelles program had a significantly lower risk of being stunted, regardless of poverty status.

To implement this recommendation, the following steps can be taken:

1. Conduct a feasibility study: Assess the current healthcare infrastructure and resources in the target area to determine if a community-based health financing program is viable. Identify potential challenges and opportunities for implementation.

2. Develop a program framework: Design a program that includes health insurance coverage for maternal health services, including prenatal care, delivery, and postnatal care. Consider incorporating nutrition services to address the link between maternal health and child stunting.

3. Collaborate with stakeholders: Engage with local communities, healthcare providers, and government agencies to gain support and ensure the program aligns with existing healthcare initiatives. Seek partnerships with organizations experienced in implementing community-based health financing programs.

4. Establish financing mechanisms: Identify sustainable funding sources for the program, such as government subsidies, community contributions, or partnerships with private sector entities. Develop a transparent and accountable financial management system.

5. Train healthcare providers: Provide training to healthcare providers on maternal health services, nutrition counseling, and the importance of early intervention to prevent child stunting. Ensure that providers are equipped with the necessary skills and knowledge to deliver quality care.

6. Raise awareness and promote enrollment: Conduct community outreach activities to raise awareness about the program and its benefits. Emphasize the importance of maternal health and the impact it has on child development. Implement strategies to encourage enrollment, such as offering incentives or simplifying the enrollment process.

7. Monitor and evaluate the program: Establish a monitoring and evaluation system to track the program’s impact on maternal health outcomes and access to care. Collect data on key indicators, such as maternal mortality rates, prenatal care coverage, and child stunting rates. Use this data to make informed decisions and continuously improve the program.

By implementing a community-based health financing program, similar to the Mutuelles program in rural Rwanda, access to maternal health services can be improved, leading to better health outcomes for both mothers and children.

AI Innovations Methodology

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

1. Expand the Mutuelles program: The study found that children enrolled in the Mutuelles community-based health-financing program had a significantly lower risk of being stunted. Expanding the program to reach more rural populations could improve access to health care and nutrition services for pregnant women, ultimately reducing the risk of maternal and child health complications.

2. Strengthen nutrition services in rural health centers: The study found that over 90% of rural health centers conducted nutrition-related campaigns and malnutrition screening for children. Further investment in training and resources for nutrition services in these centers could improve access to maternal health services, including nutrition counseling and support.

3. Increase awareness and utilization of health insurance: The Mutuelles program provides health insurance to rural populations, but there may be barriers to enrollment and utilization. Implementing targeted awareness campaigns and simplifying the enrollment process could help increase the number of pregnant women and mothers accessing maternal health services.

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

1. Define the target population: Determine the specific population that will be impacted by the recommendations. This could include pregnant women and mothers in rural areas of Rwanda.

2. Collect baseline data: Gather data on the current access to maternal health services, including the percentage of pregnant women enrolled in the Mutuelles program, the availability of nutrition services in rural health centers, and the utilization rates of these services.

3. Develop a simulation model: Create a mathematical model that incorporates the various factors influencing access to maternal health services, such as enrollment in the Mutuelles program, availability of nutrition services, and awareness and utilization of health insurance. The model should also consider other relevant factors, such as socioeconomic status and geographic location.

4. Input data and run simulations: Input the baseline data into the simulation model and run multiple simulations to estimate the potential impact of the recommendations. Vary the input parameters to assess different scenarios and their potential outcomes.

5. Analyze results: Analyze the simulation results to determine the potential impact of the recommendations on improving access to maternal health services. This could include quantifying the increase in enrollment in the Mutuelles program, the improvement in availability and utilization of nutrition services, and the overall improvement in maternal health outcomes.

6. Validate and refine the model: Validate the simulation model by comparing the simulated results with real-world data. Refine the model based on feedback and additional data to improve its accuracy and reliability.

7. Communicate findings and recommendations: Present the findings of the simulation analysis to relevant stakeholders, such as policymakers, healthcare providers, and community organizations. Use the results to advocate for the implementation of the recommended interventions and to guide decision-making processes.

It is important to note that the methodology described above is a general framework and may need to be adapted based on the specific context and available data. Additionally, the accuracy of the simulation results will depend on the quality and reliability of the input data and the assumptions made in the model.

Umbhali

Dima Health

Statistics:

Citations: 23

Identifiers:

DOI: 10.2105/AJPH.2015.302913

ISSN: 00900036

Research Areas:

Community Interventions, Environmental, Food Security, Genetics and Genomics, Health System and Policy, Maternal Access, Maternal and Child Health, Quality of Care, Social Determinants, Workforce

Study Countries:

Rwanda

Study Design:

Case-Control Study, Cross Sectional Study

Study Approach:

Quantitative

Yabelana ngalokhu: