Spatio-temporal child and maternal mortality patterns and associations with health interventions and health systems performance in sub-Saharan Africa

In 2000, the international community adopted eight Millennium Development Goals (MDGs). Three were devoted to health (reducing child mortality; improving maternal health; and combating HIV/AIDS, malaria, and other diseases). The MDGs facilitated the mobilisation of funds, implementation and assessment of cost-effective health interventions to reverse the high burden of diseases, particularly in low-and middle-income countries (LMICs). Indeed, under-5 and maternal mortality rates were unacceptably high, the pandemic of HIV/AIDS in full blow and the prevalence and incidence of malaria and other infectious diseases was rampant. The increase of financial and technical support in sub-Saharan Africa led to the scaling up of maternal and child health interventions. Furthermore, technical support enabled synergised efforts, regular monitoring of the progress and impact assessment. The “Countdown to 2015” initiative was set up to track the progress towards MDG 4 (reduce by two-thirds, between 1990 and 2015, the under-five mortality rate (U5MR) and MDG 5 (reduce by three-quarters, between 1990 and 2015, the maternal mortality ratio (MMR)). A set of 20 priority health interventions targeting life stages (from pregnancy to childhood) was promoted in several high burdened low-and middle-income countries to achieve the MDGs 4 and 5 by 2015. The fundament of these selected set of health interventions was the concept of the “continuum of care” which integrates the life cycle and place of provision of health care. At the end of 2015, the coverage of maternal and child health interventions such as skilled birth attendance, antenatal care visit, family planning, post-natal care, exclusive breastfeeding, micronutrients, supplementation, immunization, use of insecticide-treated nets (ITNs) had increased significantly. Consequently, U5MR and MMR declined substantially. However, most sub-Saharan Africa countries did not achieve MDG 4 and MDG 5. Several weaknesses of their health systems hindered the optimal implementation of the cost effective-interventions. Furthermore, a lack of reliable data prevent efficient tracking of the progress.
In LMICs, U5MR and MMR are most often derived from household surveys conducted every 3 to 5 years. Data from health services are not as reliable because they are based on attendances of health facilities and exclude events occurring in the community. Data from demographic and health surveillance systems (DHSS) are reliable but they only cover the HDSS area. Besides, most data analysis are limited to national averages ignoring local heterogeneities.
Like most sub-Saharan countries, the scaling up of the priority health interventions, health system reforms (removal of user fees, subsidization, strengthening of infrastructure and equipment) improvements of access to water, sanitation and education significantly improved health indicators during the MDGs era in Burkina Faso. However, the country failed to achieve MDGs 4 and 5.
In 2015, the international community set up new global objectives namely the Sustainable Development Goals (SDGs). The ambitious SDGs 3.1 and 3.2 related to mother and under-5 aimed at reducing the global maternal mortality ratio to less than 70 per 100,000 live births and the under-5 mortality rate to at least as low as 25 per 1,000 live births by 2030, respectively. Taking advantages of lessons learned from the MDG era, there is a need to increase the pace of annual reduction in U5MR and MMR mortality rates to achieve child and women related SDGs by 2030. As mortality rates are reducing and clustering, the subnational scale estimates become most important to optimize the health intervention’s impact.
To this end, local factors driving mortalities such as climatic and environmental factors, economic, educational individual as well as community and family level risks should be taken into account. Furthermore, health system performance and health interventions coverages also influence the distribution of mortality rate within the country and their incorporation in the statistical analysis could benefit decision-making.
Goal and objectives:
The U5MR and MMR are extremely high in Burkina Faso. We aimed to contribute to accelerating their reduction by advanced statistical analysis providing evidence for decision-making. The overreaching goal was to assess the spatio-temporal mortality patterns and associations with health system performance and interventions in Burkina Faso. The specific objectives were (1) to assess the spatial distribution of child mortality and its associations with child, maternal and household health interventions in Burkina Faso; (2) to assess the spatial distribution of child mortality and its associations with child main causes of mortality; (3) to assess the association between malaria-related health service readiness and malaria mortality in under-5 years old in Burkina Faso; (4) to assess temporal changes in the association of malaria-related health service readiness and malaria mortality in under-5 years old between 2012 and 2014 in; and (5) assess the effect of maternal, socio-economic, education and health system factors on maternal mortality across sub-Saharan Africa.
Methods: In chapter 2, we fit Bayesian geostatistical Weibull proportional hazards survival models with spatially varying coefficients. Sixteen maternal, child and household health interventions were assessed to quantify their effect on under-5 survival at the national as well at the subnational scale (administrative regions). In chapter 3, we applied the same method to assess the associations between under-5 mortality and childhood diseases. The analyses were adjusted for health interventions, climatic and environmental confounders. In both chapters, we assumed spatially structured covariate effects at the regional level. That is, the effects of health interventions or diseases are more alike in regions close to each other than those far away. Conditionally autoregressive (CAR) models modeled the spatial structure of the effects.
In chapter 4 and chapter 5, we analyzed nationally representative health facility survey to assess the readiness of health facilities to perform malaria services. Specifically, in Chapter 4, we identified firstly from malaria and general service items of the service availability and readiness assessment (SARA) survey of 2014, the most important tracer items related to malaria deaths. The items are binary with the presence of the item corresponding to “1” and “0” if the item is absent. We fit Bayesian geostatistical variable selection using stochastic search and adopting a spike and slab prior distributions for the regression coefficients. The variables selection' were applied to two separated groups of health facilities namely peripheral health centres (low level) and medical centres (high level). Multiple correspondence analysis (MCA) was applied on the selected items of each level. The creation of the composite readiness score followed the approach proposed by Asselin to create a composite poverty index. The methodology ensures the monotone increasing or decreasing condition of the score for all indicators. The number of the factorial axes to be included is determined when a factorial axis has been selected for each indicator. The factorial axes with higher discrimination measure are those that are selected.
In Chapter 5, we combined the data of the SARA surveys of 2012 and 2014. We, again grouped the health facilities into 2 levels and followed the process of Chapter 4 to create the composite readiness score.
In both Chapters 4 and 5, we split the readiness score into 3 categorical ordered levels as readiness index. A geostatistical negative binomial model was fitted to assess the effect of the facility readiness index on malaria mortality adjusted for facility characteristics (type of health facility location and administrative status).
Finally, in chapter 6 we fit a negative binomial model to assess the association of maternal mortality rate with the change in the coverage of health intervention, socio-economic covariates, health financing and health human resources indicators within two rounds of demographic and health survey (DHS). We linked the mean count of maternal death and the covariates with the number of exposure years to death as offset via a log-linear regression equation.
Results: The results of chapter 2 showed uneven spatial distribution of the associations between U5MR and health interventions. At the national level, DPT3, immunization, and baby post-natal check within 24 hours after birth had the most important effect on U5MR (hazard ratio (HR)=0.89, 95% Bayesian credible interval (BCI): 0.86-0·98 and HR=0.89, 95% BCI: 0·86-0·92, respectively). At the subnational level, the most effective interventions were skilled birth attendance, and improved drinking water, followed by baby post-natal check within 24 hours after birth, vitamin A supplementation, antenatal care visit, and all-antigens immunization (including BCG, Polio3, DPT3, and measles immunization). Centre-Est, Sahel, and Sud-Ouest were the regions with the largest number of effective interventions. There was no intervention with a significant effect on child survival in the region of Hauts Bassins.
Concerning chapter 3, malaria positive parasitemia stands as the predominant childhood condition that affects the survival of under-5 in 6 regions out of 13. It was followed by low birth weight (4 regions) and severe anemia (3 regions). The regions of Centre and Centre-Est had the lowest under-five mortality rates and there was no association with none of the selected childhood diseases.
The results of chapter 4 and 5 showed that the composite readiness index captures more variability in the dataset than the first component. That is, in chapter 4, the composite score explained 30% of variability compared to 14% when used the first axis of MCA for medical centres. For peripheral health centres, the composite score explained 53% whereas the first axis explained 18%. Peripheral health centres with the higher readiness score were associated with a 59% of reduction of malaria mortality compared to the lowest level of readiness.
In Chapter 5, the readiness of malaria service increase from 2012 to 2014 for both health facilities levels. Peripheral health centres with higher readiness index were associated with a 52% of reduction of malaria mortality compared to the lowest level. For medical centres, the middle and highest level of readiness index were associated with 28 and 38% of reduction of mortality rate compared to the lowest readiness index group.
In chapter 6, our results revealed that the temporal trend of the decreasing of the MMR was associated with the increase of the coverage of skilled birth attendance, family planning and female education rate in 24 sub-Saharan countries.
Conclusion: The crucial implication of our results from Chapters 2 and 3, is the need of shifting from the nationally and uniformly allocation of resources to targeted subnational allocation. Indeed, our results show the administrative regions that lack the effectiveness of health interventions and regions with a high burden of diseases. Furthermore, we stressed the most important health interventions to be scaled up.
In Chapters 4 and 5, the clear effect of malaria related-service readiness to reduce malaria burden in under-5 years old suggests a need for a national policy of strengthening the health system, which is lacking. Importantly, all health programmes or projects should incorporate health system reinforcement as a core component. Undoubtedly, this policy will beneficiate to others health programme to reduce morbidity and mortality of other important diseases in under-5 as well in others population groups.
Our results in Chapter 6 suggest a need for multi-sectoral synergies at each country level to reduce optimize health interventions effects. Indeed, women empowerment (education), an alternative to health financing such as insurances, removal or subsidization of user’s fees related to maternal health services could increase the coverage of maternal health interventions that, in turn, will accelerate progress toward the attainment of SDG 3.1.