Mortality beyond emergency threshold in a silent crisis– results from a population-based mortality survey in Ouaka prefecture, Central African Republic, 2020

Conflict and Health, Volume 15, No. 1, Year 2021

Interpretation

Background: The Central African Republic (CAR) suffers a protracted conflict and has the second lowest human development index in the world. Available mortality estimates vary and differ in methodology. We undertook a retrospective mortality study in the Ouaka prefecture to obtain reliable mortality data. Methods: We conducted a population-based two-stage cluster survey from 9 March to 9 April, 2020 in Ouaka prefecture. We aimed to include 64 clusters of 12 households for a required sample size of 3636 persons. We assigned clusters to communes proportional to population size and then used systematic random sampling to identify cluster starting points from a dataset of buildings in each commune. In addition to the mortality survey questions, we included an open question on challenges faced by the household. Results: We completed 50 clusters with 591 participating households including 4000 household members on the interview day. The median household size was 7 (interquartile range (IQR): 4—9). The median age was 12 (IQR: 5—27). The birth rate was 59.0/1000 population (95% confidence interval (95%-CI): 51.7—67.4). The crude and under-five mortality rates (CMR & U5MR) were 1.33 (95%-CI: 1.09—1.61) and 1.87 (95%-CI: 1.37–2.54) deaths/10,000 persons/day, respectively. The most common specified causes of death were malaria/fever (16.0%; 95%-CI: 11.0–22.7), violence (13.2%; 95%-CI: 6.3–25.5), diarrhoea/vomiting (10.6%; 95%-CI: 6.2–17.5), and respiratory infections (8.4%; 95%-CI: 4.6–14.8). The maternal mortality ratio (MMR) was 2525/100,000 live births (95%-CI: 825—5794). Challenges reported by households included health problems and access to healthcare, high number of deaths, lack of potable water, insufficient means of subsistence, food insecurity and violence. Conclusions: The CMR, U5MR and MMR exceed previous estimates, and the CMR exceeds the humanitarian emergency threshold. Violence is a major threat to life, and to physical and mental wellbeing. Other causes of death speak to poor living conditions and poor access to healthcare and preventive measures, corroborated by the challenges reported by households. Many areas of CAR face similar challenges to Ouaka. If these results were generalisable across CAR, the country would suffer one of the highest mortality rates in the world, a reminder that the longstanding “silent crisis” continues. We conducted a population-based two-stage cluster survey across Ouaka prefecture between 9 March and 9 April, 2020 in cooperation with the Ministry of Health and of the Population and the Central African Institute for Statistics and Economic and Social Studies (Institut Centrafricaine des Statistique et des Etudes Economique et Sociales (ICASEES)). The recall period commenced on the 26 May, 2019 (Mother’s Day). We chose this date as it is a well-known celebration in CAR, it fit with our requirement that the recall period cover part of the rainy and the dry seasons, and it was close in time to the end of Ramadan in 2019 (4 June) and to the Catholic feast day of the Ascension (30 May), factors which might further aid recall amongst participants. Using ENA software for SMART 2011 [33], we estimated a required sample size of 765 households and 3636 persons based on an estimated CMR of 0.72/10,000/day (double the UN estimate available at the time of protocol development [18]), precision of 0.25/10,000/day, design effect (DEFF) of 2, average household size of 5 [21], a non-response rate of 5%, and an average recall period of 265 days. We aimed to sample 64 clusters of 12 households each. The small cluster size was chosen to ensure it was feasible to complete clusters given the expected travel time to some areas, and to minimise time spent stationary in one location due to security risks. We used two-stage cluster sampling. Cluster sampling is frequently used in settings such as CAR where limited resources, logistical challenges such as poor roads, and security concerns make simple or systematic random sampling methods unfeasible [34–36]. First, we allocated clusters amongst the 16 communes of Ouaka proportional to population size according to population estimates for 2019 from ICASEES [25]. Secondly, we selected buildings as cluster starting points within each commune using systematic random sampling from a dataset of buildings. The sampling frame was created using a dataset of geographical building footprints (based on CAR Ecopia Building Footprint layer,©2019 Digital Globe, Inc) by commune [37, 38]. Buildings in settlements of less than ten buildings were excluded for feasibility reasons. From the cluster starting point, we selected subsequent households in a sequential manner by selecting the next closest building to the right until 12 households were included. We skipped buildings which were not households. In multi-household buildings, we selected one household randomly. We defined a household as a group who slept under the same roof the previous night, or if a group was spread across several huts, who ate together the previous night. In eligible and consenting households, we included all persons who were members of the household during the recall period. A priori, we did not exclude any area. We limited the number of clusters that could be replaced to 25%. If the locality of a starting point was deemed inaccessible in advance because of security concerns, we replaced it with another starting point in the same commune as per the stage two sampling previously described. If, on the day of the survey, we could not reach the starting point, or if there was no settlement at the starting point, we replaced the cluster with the next closest village on the return route. If after a second visit a household was still absent, we replaced it. If we did not achieve the target of 12 households after visiting all households in a settlement, we continued in the next closest settlement. We undertook 4 days of training with the locally recruited survey team on the aim and objectives of the survey, methods, ethics, data protection and smartphone use. During the training, with the assistance of an interpreter, the correct and appropriate phrasing of the survey questions in Sango, the local language, were practiced. We conducted structured interviews (Additional file 1) with the head of household or a designate. The head of the household (or designate) could be any adult member of the household who could provide information on events in the household over the recall period. Households self-identified the head of the household. We started the survey with an open question about challenges the household faces (“What difficulties does your household and community face on a daily basis?”) both to build rapport and to document general difficulties in the community. We noted the responses or a summary on paper. Then, using KoBoToolbox [39] on smartphones, we asked a series of questions on household composition during the recall period. For all identified members of the household during the recall period we asked demographic information, and noted arrivals or departures. For women aged 10–49 years we also asked about pregnancy during the recall period, and the outcome of the pregnancy. For deaths, the reported cause and place of death, and health seeking behaviour in the 2 weeks prior to the death was recorded. For the cause of death, the household was asked an open question and allowed to respond freely. If the response corresponded with one of the pre-defined categories listed in the questionnaire, (see Additional file 1) we marked this. If not, we noted as free text the reported cause of death or any additional information provided. The epidemiologist (ER who is a medical practitioner) reviewed these responses, and if sufficient information was available, categorised the cause of death. The recall period ran from 26 May, 2019 to the interview date. For members who left/died or arrived/were born during the recall period, their person-time contribution was adjusted for the exact date of the event if known. Otherwise, the mid-month was used. Using Poisson regression, we calculated the CMR and U5MR as deaths/10,000 population/day, the MMR as maternal deaths/100,000 live-births, and the neonatal mortality rate as deaths in the first 28 days of life/1000 live-births. We categorised the outcome of pregnancy as live-born, early pregnancy loss (< 3 months gestation or before a woman was visibly pregnant), late pregnancy loss (> 3 months or after a woman was visibly pregnant, including stillbirths). While we did not specifically ask if a pregnancy loss was a spontaneous or induced abortion, if the household mentioned it was an induced abortion this was noted. We present descriptive analyses as proportions with 95% CIs for categorical variables and means and standard deviations (SD) or medians including interquartile ranges (IQR) for continuous variables. Where appropriate, we measured differences in proportions using Pearson χ2 test and present a p-value (p). All analyses were conducted accounting for the survey sample weights and the effect of clustering induced by the two-stage sampling method. We undertook quantitative data analysis using Stata version 15.1 [40]. We digitalised the qualitative data collected in response to the introductory questions on difficulties faced. We coded it using a content analysis approach in order to identify themes and patterns from the data [41, 42]. We describe the identified themes and include a selection of illustrative quotations. Of note, during the first 2 days of interviews the responses of each individual household were not documented. Of note, we planned to conduct six other surveys using the same methodology across CAR – four studies in the other prefectures where MSF is present (Ouham, Mambere-Kadei, Mbomou, and Haute Kotto); one in the capital region of Bangui; and one covering all other prefectures. We planned to merge and weight the results of each study to provide country-wide estimates. The studies were due to take place in a staggered fashion in the first half of 2020. Ouaka was the first prefecture to start and was almost complete when the first case of COVID-19/ Sars-CoV-2 was detected in CAR. Unfortunately, the other surveys, which had not yet started, could not proceed for a number of reasons including travel restrictions to and within CAR, redirection of resources towards the COVID-19 response, and uncertainty regarding the impact of COVID-19 on the security context in CAR.

AI Digest

Study Justification:

– The Central African Republic (CAR) suffers from a protracted conflict and has a low human development index.
– Previous mortality estimates in CAR vary and differ in methodology.
– This study aimed to obtain reliable mortality data in the Ouaka prefecture.

Study Highlights:

– The study used a population-based two-stage cluster survey to collect data.
– 50 clusters with 591 participating households were included in the study.
– The study found that the crude mortality rate (CMR) and under-five mortality rate (U5MR) in Ouaka exceeded previous estimates.
– The most common specified causes of death were malaria/fever, violence, diarrhoea/vomiting, and respiratory infections.
– The maternal mortality ratio (MMR) was also high in Ouaka.
– Challenges reported by households included health problems, lack of healthcare access, high number of deaths, lack of potable water, insufficient means of subsistence, food insecurity, and violence.

Recommendations:

– Address the violence as a major threat to life and physical and mental wellbeing.
– Improve living conditions and access to healthcare and preventive measures.
– Address the challenges reported by households, such as health problems, lack of healthcare access, and lack of basic necessities.
– Conduct similar studies in other areas of CAR to assess the generalizability of the findings.
– Merge and weight the results of multiple studies to provide country-wide estimates.

Key Role Players:

– Ministry of Health and of the Population
– Central African Institute for Statistics and Economic and Social Studies (ICASEES)
– Local survey team
– Epidemiologist

Cost Items for Planning Recommendations:

– Security measures for addressing violence
– Infrastructure improvements for living conditions and healthcare access
– Healthcare resources and services
– Basic necessities provision
– Research and survey costs

Strength of Evidence

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 utilized a population-based two-stage cluster survey, which is a robust methodology. The sample size and data collection methods are clearly described. The results provide important information on mortality rates and causes of death in Ouaka prefecture. However, there are a few areas that could be improved. First, the abstract does not mention any limitations of the study, such as potential biases or sources of error. Including this information would provide a more complete picture of the strength of the evidence. Additionally, the abstract does not discuss the generalizability of the results beyond Ouaka prefecture. It would be helpful to know if the findings can be applied to other areas of the Central African Republic. Finally, the abstract does not provide any recommendations or implications for action based on the findings. Including actionable steps to address the identified challenges would enhance the practical relevance of the study.

Abstract

We conducted a population-based two-stage cluster survey across Ouaka prefecture between 9 March and 9 April, 2020 in cooperation with the Ministry of Health and of the Population and the Central African Institute for Statistics and Economic and Social Studies (Institut Centrafricaine des Statistique et des Etudes Economique et Sociales (ICASEES)). The recall period commenced on the 26 May, 2019 (Mother’s Day). We chose this date as it is a well-known celebration in CAR, it fit with our requirement that the recall period cover part of the rainy and the dry seasons, and it was close in time to the end of Ramadan in 2019 (4 June) and to the Catholic feast day of the Ascension (30 May), factors which might further aid recall amongst participants. Using ENA software for SMART 2011 [33], we estimated a required sample size of 765 households and 3636 persons based on an estimated CMR of 0.72/10,000/day (double the UN estimate available at the time of protocol development [18]), precision of 0.25/10,000/day, design effect (DEFF) of 2, average household size of 5 [21], a non-response rate of 5%, and an average recall period of 265 days. We aimed to sample 64 clusters of 12 households each. The small cluster size was chosen to ensure it was feasible to complete clusters given the expected travel time to some areas, and to minimise time spent stationary in one location due to security risks. We used two-stage cluster sampling. Cluster sampling is frequently used in settings such as CAR where limited resources, logistical challenges such as poor roads, and security concerns make simple or systematic random sampling methods unfeasible [34–36]. First, we allocated clusters amongst the 16 communes of Ouaka proportional to population size according to population estimates for 2019 from ICASEES [25]. Secondly, we selected buildings as cluster starting points within each commune using systematic random sampling from a dataset of buildings. The sampling frame was created using a dataset of geographical building footprints (based on CAR Ecopia Building Footprint layer,©2019 Digital Globe, Inc) by commune [37, 38]. Buildings in settlements of less than ten buildings were excluded for feasibility reasons. From the cluster starting point, we selected subsequent households in a sequential manner by selecting the next closest building to the right until 12 households were included. We skipped buildings which were not households. In multi-household buildings, we selected one household randomly. We defined a household as a group who slept under the same roof the previous night, or if a group was spread across several huts, who ate together the previous night. In eligible and consenting households, we included all persons who were members of the household during the recall period. A priori, we did not exclude any area. We limited the number of clusters that could be replaced to 25%. If the locality of a starting point was deemed inaccessible in advance because of security concerns, we replaced it with another starting point in the same commune as per the stage two sampling previously described. If, on the day of the survey, we could not reach the starting point, or if there was no settlement at the starting point, we replaced the cluster with the next closest village on the return route. If after a second visit a household was still absent, we replaced it. If we did not achieve the target of 12 households after visiting all households in a settlement, we continued in the next closest settlement. We undertook 4 days of training with the locally recruited survey team on the aim and objectives of the survey, methods, ethics, data protection and smartphone use. During the training, with the assistance of an interpreter, the correct and appropriate phrasing of the survey questions in Sango, the local language, were practiced. We conducted structured interviews (Additional file 1) with the head of household or a designate. The head of the household (or designate) could be any adult member of the household who could provide information on events in the household over the recall period. Households self-identified the head of the household. We started the survey with an open question about challenges the household faces (“What difficulties does your household and community face on a daily basis?”) both to build rapport and to document general difficulties in the community. We noted the responses or a summary on paper. Then, using KoBoToolbox [39] on smartphones, we asked a series of questions on household composition during the recall period. For all identified members of the household during the recall period we asked demographic information, and noted arrivals or departures. For women aged 10–49 years we also asked about pregnancy during the recall period, and the outcome of the pregnancy. For deaths, the reported cause and place of death, and health seeking behaviour in the 2 weeks prior to the death was recorded. For the cause of death, the household was asked an open question and allowed to respond freely. If the response corresponded with one of the pre-defined categories listed in the questionnaire, (see Additional file 1) we marked this. If not, we noted as free text the reported cause of death or any additional information provided. The epidemiologist (ER who is a medical practitioner) reviewed these responses, and if sufficient information was available, categorised the cause of death. The recall period ran from 26 May, 2019 to the interview date. For members who left/died or arrived/were born during the recall period, their person-time contribution was adjusted for the exact date of the event if known. Otherwise, the mid-month was used. Using Poisson regression, we calculated the CMR and U5MR as deaths/10,000 population/day, the MMR as maternal deaths/100,000 live-births, and the neonatal mortality rate as deaths in the first 28 days of life/1000 live-births. We categorised the outcome of pregnancy as live-born, early pregnancy loss ( 3 months or after a woman was visibly pregnant, including stillbirths). While we did not specifically ask if a pregnancy loss was a spontaneous or induced abortion, if the household mentioned it was an induced abortion this was noted. We present descriptive analyses as proportions with 95% CIs for categorical variables and means and standard deviations (SD) or medians including interquartile ranges (IQR) for continuous variables. Where appropriate, we measured differences in proportions using Pearson χ2 test and present a p-value (p). All analyses were conducted accounting for the survey sample weights and the effect of clustering induced by the two-stage sampling method. We undertook quantitative data analysis using Stata version 15.1 [40]. We digitalised the qualitative data collected in response to the introductory questions on difficulties faced. We coded it using a content analysis approach in order to identify themes and patterns from the data [41, 42]. We describe the identified themes and include a selection of illustrative quotations. Of note, during the first 2 days of interviews the responses of each individual household were not documented. Of note, we planned to conduct six other surveys using the same methodology across CAR – four studies in the other prefectures where MSF is present (Ouham, Mambere-Kadei, Mbomou, and Haute Kotto); one in the capital region of Bangui; and one covering all other prefectures. We planned to merge and weight the results of each study to provide country-wide estimates. The studies were due to take place in a staggered fashion in the first half of 2020. Ouaka was the first prefecture to start and was almost complete when the first case of COVID-19/ Sars-CoV-2 was detected in CAR. Unfortunately, the other surveys, which had not yet started, could not proceed for a number of reasons including travel restrictions to and within CAR, redirection of resources towards the COVID-19 response, and uncertainty regarding the impact of COVID-19 on the security context in CAR.

Materials

Innovations

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

1. Mobile Health (mHealth) Solutions: Develop mobile applications or SMS-based systems that provide pregnant women with important health information, reminders for prenatal visits, and access to teleconsultations with healthcare providers.

2. Telemedicine: Implement telemedicine services to enable pregnant women in remote areas to consult with healthcare professionals without the need for travel. This can help address the challenges of limited access to healthcare facilities.

3. Community Health Workers: Train and deploy community health workers who can provide basic prenatal care, education, and support to pregnant women in underserved areas. These workers can also facilitate referrals to healthcare facilities when necessary.

4. Transportation Solutions: Develop innovative transportation solutions, such as mobile clinics or ambulances, to improve access to healthcare facilities for pregnant women in remote areas. This can help overcome the challenges of poor roads and long travel distances.

5. Maternal Health Vouchers: Implement a voucher system that provides pregnant women with financial assistance for accessing maternal health services. This can help reduce financial barriers and improve access to quality care.

6. Maternal Health Education Programs: Develop targeted educational programs that focus on improving maternal health knowledge and practices among pregnant women and their families. This can help empower women to make informed decisions about their health and seek appropriate care.

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

8. Maternal Health Monitoring Systems: Implement robust data collection and monitoring systems to track maternal health indicators and identify areas with high maternal mortality rates. This can help prioritize resources and interventions to areas most in need.

9. Maternal Health Financing Models: Explore innovative financing models, such as social health insurance or community-based health financing, to ensure sustainable funding for maternal health services. This can help improve the availability and affordability of quality care.

10. Maternal Health Quality Improvement Initiatives: Implement quality improvement programs that focus on enhancing the quality of maternal health services, including training healthcare providers, improving infrastructure, and strengthening referral systems. This can help ensure that pregnant women receive safe and effective care.

It’s important to note that the specific context and needs of the Central African Republic should be taken into consideration when implementing these innovations.

AI Innovations Description

Based on the provided description, the following recommendation can be developed into an innovation to improve access to maternal health:

1. Strengthening Healthcare Infrastructure: Develop and implement strategies to improve healthcare infrastructure in the Central African Republic (CAR), particularly in areas like Ouaka prefecture where access to healthcare is limited. This can include building and upgrading healthcare facilities, ensuring the availability of essential medical equipment and supplies, and training healthcare professionals to provide quality maternal health services.

2. Increasing Awareness and Education: Launch awareness campaigns to educate the population, especially women and their families, about the importance of maternal health and the available healthcare services. This can involve community outreach programs, workshops, and the use of media platforms to disseminate information on prenatal care, safe delivery practices, and postnatal care.

3. Mobile Health (mHealth) Solutions: Utilize mobile technology to improve access to maternal health services. Develop mobile applications or SMS-based systems that provide information and reminders about antenatal and postnatal care, as well as emergency services. This can help overcome geographical barriers and reach women in remote areas.

4. Strengthening Emergency Obstetric Care: Enhance emergency obstetric care services in CAR, particularly in areas with high maternal mortality rates. This can involve training healthcare providers in emergency obstetric procedures, ensuring the availability of emergency medical supplies and equipment, and establishing referral systems to transfer high-risk cases to higher-level healthcare facilities.

5. Community-Based Maternal Health Programs: Implement community-based programs that involve trained community health workers who can provide basic maternal health services, conduct health education sessions, and facilitate referrals to healthcare facilities when needed. This can help improve access to care, especially for women in underserved areas.

6. Collaboration and Partnerships: Foster collaboration between the government, non-governmental organizations, and international agencies to support and implement innovative solutions to improve access to maternal health. This can involve sharing resources, expertise, and best practices to address the complex challenges faced in CAR.

By implementing these recommendations, it is expected that access to maternal health services in CAR, particularly in Ouaka prefecture, can be improved, leading to a reduction in maternal mortality rates and better overall maternal health outcomes.

AI Innovations Methodology

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

1. Strengthening healthcare infrastructure: Invest in building and upgrading healthcare facilities, particularly in areas with limited access to maternal health services. This includes ensuring the availability of essential equipment, supplies, and skilled healthcare providers.

2. Mobile health clinics: Implement mobile health clinics that can reach remote and underserved areas, providing maternal health services such as prenatal care, delivery assistance, and postnatal care. These clinics can travel to different locations on a regular schedule, ensuring access to healthcare for pregnant women.

3. Community health workers: Train and deploy community health workers who can provide basic maternal health services, education, and referrals in their communities. These workers can play a crucial role in bridging the gap between healthcare facilities and pregnant women, particularly in areas with limited access.

4. Telemedicine and telehealth services: Utilize technology to provide remote consultations, advice, and support to pregnant women. Telemedicine platforms can connect pregnant women with healthcare providers, allowing them to receive guidance and care without the need for physical travel.

To simulate the impact of these recommendations on improving access to maternal health, a methodology could include the following steps:

1. Define indicators: Identify key indicators to measure the impact of the recommendations, such as the number of pregnant women accessing prenatal care, the number of deliveries attended by skilled healthcare providers, and the reduction in maternal mortality rates.

2. Baseline data collection: Gather data on the current state of maternal health access, including the number of healthcare facilities, the availability of skilled healthcare providers, and the utilization of maternal health services.

3. Model implementation scenarios: Develop different scenarios based on the recommendations, considering factors such as the number of healthcare facilities to be built or upgraded, the number of mobile health clinics to be deployed, and the number of community health workers to be trained and deployed.

4. Simulate impact: Use modeling techniques to simulate the impact of each scenario on the defined indicators. This can involve estimating the increase in the number of pregnant women accessing care, the reduction in maternal mortality rates, and other relevant outcomes.

5. Evaluate cost-effectiveness: Assess the cost-effectiveness of each scenario by comparing the estimated impact with the resources required for implementation. This can help prioritize the most effective and feasible recommendations.

6. Refine and iterate: Based on the simulation results, refine the scenarios and iterate the process to optimize the recommendations and their potential impact on improving access to maternal health.

It’s important to note that the methodology may vary depending on the specific context and available data.

Authors & Co-Authors

Statistics:

Citations: 3

Authors: 18

Identifiers:

DOI: 10.1186/s13031-021-00385-2

Research Areas:

Community Interventions, COVID, Disparities, Environmental, Food Security, Health System and Policy, Infectious Diseases, Maternal and Child Health, Sexual and Reproductive Health, Social Determinants

Study Countries:

Central African Republic

Study Design:

Cohort Study, Cross Sectional Study

Study Approach:

Mixed-methods, Qualitative, Quantitative

Participants Gender:

Female