Protocol for analysing the epidemiology of maternal mortality in Zimbabwe: A civil registration and vital statistics trend study

PLoS ONE, Volume 16, No. 6 June, Year 2021

Interpretation

Background Sub-Saharan Africa (SSA) carries the highest burden of maternal mortality, yet, the accurate maternal mortality ratios (MMR) are uncertain in most SSA countries. Measuring maternal mortality is challenging in this region, where civil registration and vital statistics (CRVS) systems are weak or non-existent. We describe a protocol designed to explore the use of CRVS to monitor maternal mortality in Zimbabwe—an SSA country. Methods In this study, we will collect deliveries and maternal death data from CRVS (government death registration records) and health facilities for 2007–2008 and 2018–2019 to compare MMRs and causes of death. We will code the causes of death using classifications in the maternal mortality version of the 10th revision to the international classification of diseases. We will compare the proportions of maternal deaths attributed to different causes between the two study periods. We will also analyse missingness and misclassification of maternal deaths in CRVS to assess the validity of their use to measure maternal mortality in Zimbabwe. Discussion This study will determine changes in MMR and causes of maternal mortality in Zimbabwe over a decade. It will show whether HIV, which was at its peak in 2007–2008, remains a significant cause of maternal deaths in Zimbabwe. The study will recommend measures to improve the quality of CRVS data for future use to monitor maternal mortality in Zimbabwe and other SSA countries of similar characteristics. To analyse the epidemiology of maternal mortality in Zimbabwe by assessing changes in the MMR and causes of maternal death over a decade and the validity of using CRVS data in future to monitor maternal mortality in Zimbabwe and similar countries. We will conduct an observational, population-based, trend analysis study comparing maternal mortality estimates and cause for two studies that are ten years apart. The study will use the reproductive age mortality survey (RAMOS) method, collecting data for all deaths of women of reproductive age (WRA) to identify maternal deaths [20, 21]. It will collect deaths for the one year from 1 May 2007 to 15 June 2008 and 1 May 2018 to 15 June 2019. The data will be collected from CRVS and health facilities (central, provincial, district, mission and rural hospitals and primary care clinics). Women of reproductive age, 12–49 years in Zimbabwe. The primary study outcomes are change in MMR and changes in proportions of deaths due to the leading causes of maternal mortality in Zimbabwe from 2007–2008 to 2018–2019. Secondary study outcomes are changes in proportions of deaths of WRA and pregnancy-related deaths. The sampling for 2007–2008 and 2018–2019 data is the same. The sample size was estimated using deliveries—the denominator for the MMR. The population was stratified into the ten provinces of the country, and one district was selected from each province using simple random sampling [22–24]. An extra district was sampled from Harare by proportional-to-size sampling, considering the city’s large population and two referral hospitals. In total, the study will be done in 11 districts and 291 health facilities. The sample size of deliveries was estimated using the last verifiable MMR of Zimbabwe and the country’s total fertility rate (TFR), treating MMR as a single proportion [25]. It was calculated using the Wald test, which estimates a one-sample proportion against an expected proportion (the last verifiable MMR as a proportion). Power of 80%, z-value for two-sided 95% CI, normal-approximation continuity correction for the expected proportion, 2.5% error margin (for the alternative hypothesis of MMR outside the 95% CI of the last verified MMR) was applied. A design effect of 2 was also applied, based on the two-step sampling procedure of stratifying the study population into provinces and districts and randomly selecting the districts [23, 24, 26]. Deliveries and deaths occurring in the study districts are consecutively enrolled until the sample size of deliveries is reached. The data collection procedures are summarised in Fig 1. The 2007–2008 dataset will be built from data collected in the first study conducted in 2007–2008. We will validate the 2007–2008 data by re-visiting the vital registration (VR) and hospital records in the 11 districts and collect data for all deaths of WRAs and their causes of death for 2007–2008. We will collect case notes for the deaths from hospitals to code the causes of death. We will use the new data to clean up the data collected in 2007–2008. We will collect the 2018–2019 data from the same 11 districts using the same method recording all deaths of WRA at the VR offices and hospitals. Depending on the level of the institution, sources of maternal deaths data in health institutions in Zimbabwe include maternal death line-list in reproductive health offices, delivery unit, casualty unit, theatre, intensive care unit, high dependency unit, female ward, and mortuary registers [27]. We will triangulate and collect additional data from the maternal death surveillance, and response (MDSR) reports. MDSR reports and VR records include community deaths. Deaths of WRA will be identified at the VR offices from birth and death registration forms filed by year in locked records rooms. We will record the data on line-listing forms, including demographic and cause of death information for each woman. A second line-listing form will record more detailed data for each PRD, including personal identifiers such as name, date of birth, national identification number, home address, place of residence, date of death, place of death and cause of death. We will use personal identifiers to track and de-duplicate the women across CRVS and health facility records. PRDs will be identified at health facilities from maternity, delivery, maternal death and mortuary registers, and MDSR reports. The data will be recorded on the same line-listing forms. Case notes for all PRDs will be collected from hospitals. MDSR reports will be collected from the district and provincial health offices. PRDs identified in the RAMOS will be verified with those reported in the district health information system (DHIS2) by reporting month and institution to avoid missing some deaths. The 2007–2008 data were collected on a form adapted from WHO. We will use updated versions of the 2007–2008 questionnaire to collect the 2018–2019 data and use the newly collected 2007–2008 data to update the previous data forms. Variables collected in the form include demographic, antenatal, antepartum and postpartum information and causes of death. Four (4) trained research midwives will collect the data from all districts. A panel of obstetricians will review all PRDs for 2007–2008 and 2018–2019 to confirm maternal deaths, the type of death (direct, indirect or incidental), category and the actual cause of death. They will use the WHO ICD-10 MM guide for the classification and coding. ICD-10 MM groups the causes of maternal death into nine groups, namely: 1: Pregnancies with the abortive outcome, 2: Hypertensive disorders in pregnancy, childbirth and the puerperium, 3: Obstetric haemorrhage, 4: Pregnancy-related infection, 5: Other obstetric complications, 6: Unanticipated complications of management, 7: Non-obstetric complications, 8: Unknown/undetermined causes and 9: Coincidental causes [1]. Groups 1–8 are maternal deaths, while group 9 are pregnancy-related deaths. Obstetrician reviews will be documented on an additional data form page attached to each PRD’s data form. The primary data form and line-listing forms were adapted from the WHO maternal mortality and morbidity systematic review tool [28]. The death registration forms used in Zimbabwe are aligned to ICD-10 MM. For 2007–2008 data, the additional data form page will be used to facilitate changes in the database. For 2018–2019, the line-listings of PRDs will be used to complete the main data form. The data form is completed in the field while data collectors still have access to other patient data from field records. After coding the causes of death, the data will be entered into a password protected MS Access database. The data processing procedures are summarised in Fig 2. The study will include deaths of WRA (12–49 years) from the study districts, including those who died during pregnancy or within 100 days of termination of pregnancy. We will include PRDs due to illnesses, injuries and accidents that were not pregnancy-induced. We will exclude these from MMR estimates. We will analyse these other deaths of WRA as secondary outcomes. Maternal deaths of unknown or non-specific causes coded in ICD-10 MM group 8 will be included in MMR estimates and cause of death analyses. Women who died within but originated from other districts will be excluded from MMR estimates and cause of death analyses but reported in secondary outcomes. The principal investigator (PI) will collect data with the research midwives at the first two districts and at two other districts midway through data collection. During this time, the PI will ensure that data collectors are following the protocol. PRDs identified in the study will be linked across all data sources to minimise missing data and avoid duplicating some deaths. The PI will review the completeness and quality of data on all forms. Obstetricians’ review of every PRD will be another quality control measure on the data. We will compare demographic and clinical characteristics of 2007–2008 and 2018–2019 deaths, including age, parity, gravidity and pregnancy complications, using proportions and mean/medians (with standard deviations) with tests of significance of difference. We will assess risk factors for maternal death using multivariable logistic regression models on 2007–2008 data, which has a comparison group of deliveries where the women did not die. We will include predictor variables with univariate p-values less than 0.25 in the initial model and test model fitness using Hosmer and Lemeshow Chi-square test [29]. Predictor variables with p-values less than 0.2 will be included in the final model [30]. This analysis will be design-adjusted and use survey weights [31, 32]. We will calculate and compare MMRs for 2007–2008 and 2018–2019 using their 95% CIs. We will calculate the MMRs by dividing the total number of maternal deaths with the total live births for that year, multiplied by 100,000 [33, 34]. MMR estimates with overlapping CIs will suggest that the MMR has not changed over the decade. However, we will also consider the clinical significance of differences in MMRs when CIs show no difference. Differences in MMR point estimates greater than 100 will be considered clinically significant [35, 36]. We will calculate proportions (with 95% CIs) of deaths due to each underlying cause of death by dividing the number of deaths for each cause with the total number of deaths in the survey to identify the leading causes of maternal mortality in 2007–2008 and 2018–2019. We will use the 95% CIs to evaluate changes in the contribution of each cause to maternal mortality over the decade [37]. Subject to the suitability of the study design and availability of requisite data, competing risk analysis will be considered in cause-of-death analysis [38]. We will calculate and compare the proportion of HIV-associated maternal deaths in 2007–2008 and 2018–2019. HIV-associated maternal deaths are deaths where an AIDS-defining condition is present [28, 39, 40]. We will calculate the proportions by dividing the number of HIV-associated maternal deaths by the total number of maternal deaths in that year. We will triangulate the number of maternal deaths identified from CRVS, health facilities and MDSR to quantify missingness in the study. We will classify the maternal deaths according to false positives, false negatives, true positives, true negatives, and missing (both true and false maternal deaths) to assess the validity of using CRVS to monitor maternal mortality in Zimbabwe. We will review the quality of records in CRVS records to evaluate its use to monitor maternal mortality. We will discuss the challenges identified in the data systems and recommend measures to strengthen CRVS for future use to monitor maternal mortality in Zimbabwe and similar countries. Some of our analysis will be weighted since this study has multistage sampling [41–43]. The weights will be calculated using non-proportional sample allocation to clusters and possible inter-cluster differences in response rates [22, 44]. The weights are calculated using the product of two probabilities, considering the two clustering levels of province and district in the sampling design. The first probability is for selecting a district in a province, which is one out of the number of districts in the province. The second probability is for identifying a maternal death in the district, estimated at 70% (0.7) by a verification exercise conducted in the 2007–2008 study [19]. The final sample weight is the inverse of the product of the two probabilities [44, 45]. Sample weights will be calculated per district since the probabilities vary by district. The protocol was approved by the UNDP-UNFPA-UNICEF-WHO-World Bank Human Reproduction Program (HRP) (Date: 2019-03-27), World Health Organization Ethics Review Committee (Ref: ERC 0003348; Date: 2020-04-24), Medical Research Council of Zimbabwe (Ref: MRCZ/A/2613; Date: 2020-07-17) and University of Pretoria Faculty Health Research Ethics Committee (Ref: 339/2019; First submission approved– 2020-07-15; Amendment approved: 2021-01-21). The Ministry of Health and Child Care (MoHCC) and the Registrar General’s Department of Zimbabwe granted permissions for access to patient and civil registration records, respectively. Additional clearances for data collection will be received from provincial, district and station heads of the MoHCC and RG’s offices upon presentation of permission letters from their head offices. All approvals waived consenting of study participants since the study will collect data from existing records and not engage living human subjects.

AI Digest

Study Justification:

– Sub-Saharan Africa (SSA) has the highest burden of maternal mortality, but accurate maternal mortality ratios (MMR) are uncertain in most SSA countries.
– Measuring maternal mortality is challenging in this region due to weak or non-existent civil registration and vital statistics (CRVS) systems.
– This study aims to explore the use of CRVS to monitor maternal mortality in Zimbabwe, an SSA country.
– The study will determine changes in MMR and causes of maternal mortality in Zimbabwe over a decade.
– It will assess the validity of using CRVS data to measure maternal mortality in Zimbabwe.
– The findings will contribute to improving the quality of CRVS data for future monitoring of maternal mortality in Zimbabwe and similar countries.

Study Highlights:

– The study will collect deliveries and maternal death data from CRVS and health facilities for 2007-2008 and 2018-2019.
– Causes of death will be coded using classifications in the maternal mortality version of the 10th revision to the international classification of diseases.
– The study will compare the proportions of maternal deaths attributed to different causes between the two study periods.
– Missingness and misclassification of maternal deaths in CRVS will be analyzed to assess the validity of their use.
– Demographic and clinical characteristics of maternal deaths will be compared between the two study periods.
– Risk factors for maternal death will be assessed using multivariable logistic regression models.
– Maternal mortality ratios (MMRs) for 2007-2008 and 2018-2019 will be calculated and compared.
– Proportions of deaths due to each underlying cause of death will be calculated and compared.
– The proportion of HIV-associated maternal deaths will be calculated and compared.
– The study will quantify missingness in the data sources and evaluate the validity of using CRVS to monitor maternal mortality.
– Challenges identified in the data systems will be discussed, and recommendations will be made to strengthen CRVS for future use.

Recommendations for Lay Reader and Policy Maker:

– Strengthening civil registration and vital statistics (CRVS) systems is crucial for accurate monitoring of maternal mortality.
– Improving the quality of CRVS data will help identify changes in maternal mortality ratios (MMR) and causes of maternal death over time.
– Addressing challenges in data systems and ensuring accurate classification and coding of causes of death are important for reliable monitoring.
– HIV remains a significant cause of maternal deaths in Zimbabwe, and efforts to prevent and manage HIV-related complications during pregnancy should be prioritized.
– Enhancing access to antenatal care, skilled birth attendance, and emergency obstetric care can contribute to reducing maternal mortality.
– Policy interventions should focus on addressing the leading causes of maternal mortality identified in the study, such as hypertensive disorders, obstetric hemorrhage, and pregnancy-related infections.
– Collaboration between health facilities, civil registration offices, and other relevant stakeholders is essential for comprehensive and accurate data collection.

Key Role Players:

– Ministry of Health and Child Care (MoHCC)
– Registrar General’s Department of Zimbabwe
– UNDP-UNFPA-UNICEF-WHO-World Bank Human Reproduction Program (HRP)
– World Health Organization Ethics Review Committee
– Medical Research Council of Zimbabwe
– University of Pretoria Faculty Health Research Ethics Committee
– Provincial, district, and station heads of the MoHCC and Registrar General’s offices
– Research midwives
– Panel of obstetricians

Cost Items for Planning Recommendations:

– Training and capacity building for data collectors and research midwives
– Data collection materials and equipment (e.g., forms, computers, data entry software)
– Transportation and logistics for data collection in 11 districts and 291 health facilities
– Review and validation of data by obstetricians
– Data entry and management
– Data analysis and statistical support
– Reporting and dissemination of study findings
– Monitoring and evaluation of the implementation of recommendations
– Collaboration and coordination meetings with key stakeholders
– Overhead costs and administrative support

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 design is well-described, and the methods for data collection and analysis are clear. The abstract also mentions the use of validated tools and the involvement of trained research midwives and obstetricians, which adds to the strength of the evidence. However, there could be more information on the sample size calculation and the statistical methods used for data analysis. Additionally, it would be helpful to include information on potential limitations of the study and how they will be addressed. To improve the evidence, the authors could provide more details on the sample size calculation and statistical methods, as well as discuss potential limitations and mitigation strategies.

Abstract

To analyse the epidemiology of maternal mortality in Zimbabwe by assessing changes in the MMR and causes of maternal death over a decade and the validity of using CRVS data in future to monitor maternal mortality in Zimbabwe and similar countries. We will conduct an observational, population-based, trend analysis study comparing maternal mortality estimates and cause for two studies that are ten years apart. The study will use the reproductive age mortality survey (RAMOS) method, collecting data for all deaths of women of reproductive age (WRA) to identify maternal deaths [20, 21]. It will collect deaths for the one year from 1 May 2007 to 15 June 2008 and 1 May 2018 to 15 June 2019. The data will be collected from CRVS and health facilities (central, provincial, district, mission and rural hospitals and primary care clinics). Women of reproductive age, 12–49 years in Zimbabwe. The primary study outcomes are change in MMR and changes in proportions of deaths due to the leading causes of maternal mortality in Zimbabwe from 2007–2008 to 2018–2019. Secondary study outcomes are changes in proportions of deaths of WRA and pregnancy-related deaths. The sampling for 2007–2008 and 2018–2019 data is the same. The sample size was estimated using deliveries—the denominator for the MMR. The population was stratified into the ten provinces of the country, and one district was selected from each province using simple random sampling [22–24]. An extra district was sampled from Harare by proportional-to-size sampling, considering the city’s large population and two referral hospitals. In total, the study will be done in 11 districts and 291 health facilities. The sample size of deliveries was estimated using the last verifiable MMR of Zimbabwe and the country’s total fertility rate (TFR), treating MMR as a single proportion [25]. It was calculated using the Wald test, which estimates a one-sample proportion against an expected proportion (the last verifiable MMR as a proportion). Power of 80%, z-value for two-sided 95% CI, normal-approximation continuity correction for the expected proportion, 2.5% error margin (for the alternative hypothesis of MMR outside the 95% CI of the last verified MMR) was applied. A design effect of 2 was also applied, based on the two-step sampling procedure of stratifying the study population into provinces and districts and randomly selecting the districts [23, 24, 26]. Deliveries and deaths occurring in the study districts are consecutively enrolled until the sample size of deliveries is reached. The data collection procedures are summarised in Fig 1. The 2007–2008 dataset will be built from data collected in the first study conducted in 2007–2008. We will validate the 2007–2008 data by re-visiting the vital registration (VR) and hospital records in the 11 districts and collect data for all deaths of WRAs and their causes of death for 2007–2008. We will collect case notes for the deaths from hospitals to code the causes of death. We will use the new data to clean up the data collected in 2007–2008. We will collect the 2018–2019 data from the same 11 districts using the same method recording all deaths of WRA at the VR offices and hospitals. Depending on the level of the institution, sources of maternal deaths data in health institutions in Zimbabwe include maternal death line-list in reproductive health offices, delivery unit, casualty unit, theatre, intensive care unit, high dependency unit, female ward, and mortuary registers [27]. We will triangulate and collect additional data from the maternal death surveillance, and response (MDSR) reports. MDSR reports and VR records include community deaths. Deaths of WRA will be identified at the VR offices from birth and death registration forms filed by year in locked records rooms. We will record the data on line-listing forms, including demographic and cause of death information for each woman. A second line-listing form will record more detailed data for each PRD, including personal identifiers such as name, date of birth, national identification number, home address, place of residence, date of death, place of death and cause of death. We will use personal identifiers to track and de-duplicate the women across CRVS and health facility records. PRDs will be identified at health facilities from maternity, delivery, maternal death and mortuary registers, and MDSR reports. The data will be recorded on the same line-listing forms. Case notes for all PRDs will be collected from hospitals. MDSR reports will be collected from the district and provincial health offices. PRDs identified in the RAMOS will be verified with those reported in the district health information system (DHIS2) by reporting month and institution to avoid missing some deaths. The 2007–2008 data were collected on a form adapted from WHO. We will use updated versions of the 2007–2008 questionnaire to collect the 2018–2019 data and use the newly collected 2007–2008 data to update the previous data forms. Variables collected in the form include demographic, antenatal, antepartum and postpartum information and causes of death. Four (4) trained research midwives will collect the data from all districts. A panel of obstetricians will review all PRDs for 2007–2008 and 2018–2019 to confirm maternal deaths, the type of death (direct, indirect or incidental), category and the actual cause of death. They will use the WHO ICD-10 MM guide for the classification and coding. ICD-10 MM groups the causes of maternal death into nine groups, namely: 1: Pregnancies with the abortive outcome, 2: Hypertensive disorders in pregnancy, childbirth and the puerperium, 3: Obstetric haemorrhage, 4: Pregnancy-related infection, 5: Other obstetric complications, 6: Unanticipated complications of management, 7: Non-obstetric complications, 8: Unknown/undetermined causes and 9: Coincidental causes [1]. Groups 1–8 are maternal deaths, while group 9 are pregnancy-related deaths. Obstetrician reviews will be documented on an additional data form page attached to each PRD’s data form. The primary data form and line-listing forms were adapted from the WHO maternal mortality and morbidity systematic review tool [28]. The death registration forms used in Zimbabwe are aligned to ICD-10 MM. For 2007–2008 data, the additional data form page will be used to facilitate changes in the database. For 2018–2019, the line-listings of PRDs will be used to complete the main data form. The data form is completed in the field while data collectors still have access to other patient data from field records. After coding the causes of death, the data will be entered into a password protected MS Access database. The data processing procedures are summarised in Fig 2. The study will include deaths of WRA (12–49 years) from the study districts, including those who died during pregnancy or within 100 days of termination of pregnancy. We will include PRDs due to illnesses, injuries and accidents that were not pregnancy-induced. We will exclude these from MMR estimates. We will analyse these other deaths of WRA as secondary outcomes. Maternal deaths of unknown or non-specific causes coded in ICD-10 MM group 8 will be included in MMR estimates and cause of death analyses. Women who died within but originated from other districts will be excluded from MMR estimates and cause of death analyses but reported in secondary outcomes. The principal investigator (PI) will collect data with the research midwives at the first two districts and at two other districts midway through data collection. During this time, the PI will ensure that data collectors are following the protocol. PRDs identified in the study will be linked across all data sources to minimise missing data and avoid duplicating some deaths. The PI will review the completeness and quality of data on all forms. Obstetricians’ review of every PRD will be another quality control measure on the data. We will compare demographic and clinical characteristics of 2007–2008 and 2018–2019 deaths, including age, parity, gravidity and pregnancy complications, using proportions and mean/medians (with standard deviations) with tests of significance of difference. We will assess risk factors for maternal death using multivariable logistic regression models on 2007–2008 data, which has a comparison group of deliveries where the women did not die. We will include predictor variables with univariate p-values less than 0.25 in the initial model and test model fitness using Hosmer and Lemeshow Chi-square test [29]. Predictor variables with p-values less than 0.2 will be included in the final model [30]. This analysis will be design-adjusted and use survey weights [31, 32]. We will calculate and compare MMRs for 2007–2008 and 2018–2019 using their 95% CIs. We will calculate the MMRs by dividing the total number of maternal deaths with the total live births for that year, multiplied by 100,000 [33, 34]. MMR estimates with overlapping CIs will suggest that the MMR has not changed over the decade. However, we will also consider the clinical significance of differences in MMRs when CIs show no difference. Differences in MMR point estimates greater than 100 will be considered clinically significant [35, 36]. We will calculate proportions (with 95% CIs) of deaths due to each underlying cause of death by dividing the number of deaths for each cause with the total number of deaths in the survey to identify the leading causes of maternal mortality in 2007–2008 and 2018–2019. We will use the 95% CIs to evaluate changes in the contribution of each cause to maternal mortality over the decade [37]. Subject to the suitability of the study design and availability of requisite data, competing risk analysis will be considered in cause-of-death analysis [38]. We will calculate and compare the proportion of HIV-associated maternal deaths in 2007–2008 and 2018–2019. HIV-associated maternal deaths are deaths where an AIDS-defining condition is present [28, 39, 40]. We will calculate the proportions by dividing the number of HIV-associated maternal deaths by the total number of maternal deaths in that year. We will triangulate the number of maternal deaths identified from CRVS, health facilities and MDSR to quantify missingness in the study. We will classify the maternal deaths according to false positives, false negatives, true positives, true negatives, and missing (both true and false maternal deaths) to assess the validity of using CRVS to monitor maternal mortality in Zimbabwe. We will review the quality of records in CRVS records to evaluate its use to monitor maternal mortality. We will discuss the challenges identified in the data systems and recommend measures to strengthen CRVS for future use to monitor maternal mortality in Zimbabwe and similar countries. Some of our analysis will be weighted since this study has multistage sampling [41–43]. The weights will be calculated using non-proportional sample allocation to clusters and possible inter-cluster differences in response rates [22, 44]. The weights are calculated using the product of two probabilities, considering the two clustering levels of province and district in the sampling design. The first probability is for selecting a district in a province, which is one out of the number of districts in the province. The second probability is for identifying a maternal death in the district, estimated at 70% (0.7) by a verification exercise conducted in the 2007–2008 study [19]. The final sample weight is the inverse of the product of the two probabilities [44, 45]. Sample weights will be calculated per district since the probabilities vary by district. The protocol was approved by the UNDP-UNFPA-UNICEF-WHO-World Bank Human Reproduction Program (HRP) (Date: 2019-03-27), World Health Organization Ethics Review Committee (Ref: ERC 0003348; Date: 2020-04-24), Medical Research Council of Zimbabwe (Ref: MRCZ/A/2613; Date: 2020-07-17) and University of Pretoria Faculty Health Research Ethics Committee (Ref: 339/2019; First submission approved– 2020-07-15; Amendment approved: 2021-01-21). The Ministry of Health and Child Care (MoHCC) and the Registrar General’s Department of Zimbabwe granted permissions for access to patient and civil registration records, respectively. Additional clearances for data collection will be received from provincial, district and station heads of the MoHCC and RG’s offices upon presentation of permission letters from their head offices. All approvals waived consenting of study participants since the study will collect data from existing records and not engage living human subjects.

Materials

https://efashare.b-cdn.net/materials/52d0c7650f8824716cf06ecd5989061e16129c67987056dad00eae2ec9e87d8f/pone.0252106.s001.docx

Innovations

The innovation described in the provided text is a protocol for analyzing the epidemiology of maternal mortality in Zimbabwe using civil registration and vital statistics (CRVS) data. This protocol aims to improve access to maternal health by monitoring maternal mortality rates (MMR) and causes of death over a decade. It involves collecting data from CRVS systems and health facilities for two study periods (2007-2008 and 2018-2019) and comparing MMRs and causes of death. The study also includes analyzing missingness and misclassification of maternal deaths in CRVS to assess the validity of using this data to measure maternal mortality. The findings of this study will inform recommendations to improve the quality of CRVS data for monitoring maternal mortality in Zimbabwe and similar countries.

AI Innovations Description

The recommendation to improve access to maternal health based on the described protocol is to strengthen the civil registration and vital statistics (CRVS) systems in Zimbabwe and similar countries. This can be achieved through the following measures:

1. Enhance data collection: Improve the collection of deliveries and maternal death data from both CRVS and health facilities. This includes ensuring accurate and complete recording of demographic and cause of death information for each woman.

2. Standardize classification and coding: Use internationally recognized classifications, such as the maternal mortality version of the 10th revision to the international classification of diseases (ICD-10 MM), to code the causes of death. This will enable consistent and comparable analysis of maternal mortality data.

3. Validate and clean data: Validate the data collected by revisiting vital registration and hospital records to ensure accuracy and completeness. Collect case notes from hospitals to code the causes of death and use the new data to clean up previously collected data.

4. Strengthen data sources: Improve the quality and availability of data sources, such as maternal death line-lists, delivery units, and reproductive health offices in health institutions. Triangulate data from multiple sources, including maternal death surveillance and response (MDSR) reports, to ensure comprehensive coverage.

5. Enhance data analysis: Conduct trend analysis studies over a decade to assess changes in maternal mortality ratios (MMRs) and causes of maternal death. Use multivariable logistic regression models to identify risk factors for maternal death.

6. Assess data validity: Evaluate the validity of using CRVS data to monitor maternal mortality by comparing the number of maternal deaths identified from CRVS, health facilities, and MDSR. Classify maternal deaths according to false positives, false negatives, true positives, true negatives, and missing cases to assess the accuracy of CRVS data.

7. Strengthen CRVS systems: Based on the challenges identified in the data systems, recommend measures to strengthen CRVS for future use in monitoring maternal mortality. This may include improving data collection procedures, enhancing training and capacity-building for data collectors, and implementing quality control measures.

By implementing these recommendations, the accuracy and reliability of maternal mortality data can be improved, leading to better monitoring and evaluation of maternal health programs. This, in turn, can contribute to the development of targeted interventions and policies to improve access to maternal health services and reduce maternal mortality rates.

AI Innovations Methodology

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

1. Strengthen Civil Registration and Vital Statistics (CRVS) Systems: Enhance the capacity and infrastructure of CRVS systems in Zimbabwe and similar countries to accurately capture and record maternal deaths. This can include improving data collection methods, training personnel, and implementing standardized reporting procedures.

2. Improve Data Quality and Validity: Implement measures to ensure the accuracy and validity of data collected through CRVS systems. This can involve conducting regular data audits, verifying and validating data sources, and addressing issues of missing or misclassified maternal deaths.

3. Enhance Maternal Death Surveillance and Response (MDSR): Strengthen the MDSR system to capture comprehensive data on maternal deaths, including deaths that occur outside of health facilities. This can involve improving reporting mechanisms, enhancing data collection tools, and promoting collaboration between health facilities and community-based organizations.

4. Increase Awareness and Education: Implement awareness campaigns to educate communities, healthcare providers, and policymakers about the importance of maternal health and the need for accurate data collection. This can help reduce stigma, encourage early detection and reporting of maternal deaths, and promote accountability in the healthcare system.

5. Strengthen Healthcare Infrastructure: Invest in improving healthcare infrastructure, particularly in rural areas, to ensure access to quality maternal healthcare services. This can include building and upgrading healthcare facilities, providing essential medical equipment and supplies, and training healthcare personnel.

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

1. Define Key Indicators: Identify key indicators that reflect access to maternal health, such as maternal mortality ratio (MMR), proportion of deaths due to specific causes, and coverage of essential maternal health services.

2. Establish Baseline Data: Gather existing data on the selected indicators to establish a baseline for comparison. This can include data from CRVS systems, health facilities, and other relevant sources.

3. Define Scenarios: Develop different scenarios that represent the potential impact of the recommendations. For example, one scenario could assume full implementation of all recommendations, while another scenario could assume partial implementation or no implementation.

4. Simulate Impact: Use statistical modeling techniques to simulate the impact of each scenario on the selected indicators. This can involve analyzing the changes in MMR, cause-specific mortality rates, and service coverage rates under different scenarios.

5. Evaluate Results: Compare the simulated results of each scenario to the baseline data to assess the potential impact of the recommendations on improving access to maternal health. This can involve analyzing the magnitude of change in the selected indicators and identifying any significant improvements or challenges.

6. Refine and Iterate: Based on the evaluation results, refine the recommendations and simulation methodology as needed. Iterate the process to further explore different scenarios and assess their potential impact on improving access to maternal health.

It is important to note that the methodology for simulating the impact of recommendations may vary depending on the available data, resources, and specific objectives of the study.

Author

Dima Health

Statistics:

Citations: 4

Identifiers:

DOI: 10.1371/journal.pone.0252106

Research Areas:

Community Interventions, Health System and Policy, Infectious Diseases, Maternal and Child Health, Noncommunicable Diseases, Sexual and Reproductive Health, Social Determinants, Violence and Injury, Workforce

Study Countries:

Zimbabwe

Study Design:

Cross Sectional Study, Grounded Theory, Quasi Experimental Study

Study Approach:

Mixed-methods, Quantitative, Systematic Review

Participants Gender:

Female