Global, regional, and national estimates and trends in stillbirths from 2000 to 2019: a systematic assessment

Study Justification:
– Stillbirths are a significant public health issue and an indicator of the quality of care during pregnancy and birth.
– The UN Global Strategy for Women’s, Children’s and Adolescents’ Health and the Every Newborn Action Plan call for the prevention of stillbirths.
– Standardized measurement of stillbirth rates across countries is crucial for preventing stillbirths.
Highlights:
– In 2019, an estimated 2.0 million babies were stillborn at 28 weeks or more of gestation globally.
– The global stillbirth rate in 2019 was 13.9 stillbirths per 1000 total births.
– Stillbirth rates varied widely across regions, with the highest rates in west and central Africa.
– Progress in reducing stillbirth rates has been slower compared to decreases in mortality rates for children under 5 years.
Recommendations:
– Accelerated improvements are needed in regions and countries with high stillbirth rates, particularly in sub-Saharan Africa.
– Increased efforts are required to raise public awareness, improve data collection, assess progress, and understand local public health priorities.
– These efforts will require investment and support.
Key Role Players:
– Researchers and scientists specializing in maternal and child health.
– Public health officials and policymakers at national and international levels.
– Non-governmental organizations (NGOs) working in maternal and child health.
– Health professionals, including obstetricians, midwives, and nurses.
– Community leaders and advocates for women’s and children’s health.
Cost Items for Planning Recommendations:
– Funding for research and data collection on stillbirth rates.
– Investment in public awareness campaigns and education programs.
– Resources for improving data collection systems, including vital registration systems and health information systems.
– Training and capacity building for healthcare providers.
– Support for community-based interventions and programs.
– Monitoring and evaluation of progress in reducing stillbirth rates.

The strength of evidence for this abstract is 9 out of 10.
The evidence in the abstract is strong because it is based on a systematic assessment that used a large dataset from multiple sources. The methods used were rigorous and accounted for data quality, definitional adjustments, and source type bias. The estimates were generated using a Bayesian hierarchical temporal sparse regression model. The study compared stillbirth rates to other mortality estimates and provided regional and global trends. To improve the evidence, it would be helpful to provide more details on the specific covariates used in the model and their coefficients.

Background: Stillbirths are a major public health issue and a sensitive marker of the quality of care around pregnancy and birth. The UN Global Strategy for Women’s, Children’s and Adolescents’ Health (2016–30) and the Every Newborn Action Plan (led by UNICEF and WHO) call for an end to preventable stillbirths. A first step to prevent stillbirths is obtaining standardised measurement of stillbirth rates across countries. We estimated stillbirth rates and their trends for 195 countries from 2000 to 2019 and assessed progress over time. Methods: For a systematic assessment, we created a dataset of 2833 country-year datapoints from 171 countries relevant to stillbirth rates, including data from registration and health information systems, household-based surveys, and population-based studies. After data quality assessment and exclusions, we used 1531 datapoints to estimate country-specific stillbirth rates for 195 countries from 2000 to 2019 using a Bayesian hierarchical temporal sparse regression model, according to a definition of stillbirth of at least 28 weeks’ gestational age. Our model combined covariates with a temporal smoothing process such that estimates were informed by data for country-periods with high quality data, while being based on covariates for country-periods with little or no data on stillbirth rates. Bias and additional uncertainty associated with observations based on alternative stillbirth definitions and source types, and observations that were subject to non-sampling errors, were included in the model. We compared the estimated stillbirth rates and trends to previously reported mortality estimates in children younger than 5 years. Findings: Globally in 2019, an estimated 2·0 million babies (90% uncertainty interval [UI] 1·9–2·2) were stillborn at 28 weeks or more of gestation, with a global stillbirth rate of 13·9 stillbirths (90% UI 13·5–15·4) per 1000 total births. Stillbirth rates in 2019 varied widely across regions, from 22·8 stillbirths (19·8–27·7) per 1000 total births in west and central Africa to 2·9 (2·7–3·0) in western Europe. After west and central Africa, eastern and southern Africa and south Asia had the second and third highest stillbirth rates in 2019. The global annual rate of reduction in stillbirth rate was estimated at 2·3% (90% UI 1·7–2·7) from 2000 to 2019, which was lower than the 2·9% (2·5–3·2) annual rate of reduction in neonatal mortality rate (for neonates aged <28 days) and the 4·3% (3·8–4·7) annual rate of reduction in mortality rate among children aged 1–59 months during the same period. Based on the lower bound of the 90% UIs, 114 countries had an estimated decrease in stillbirth rate since 2000, with four countries having a decrease of at least 50·0%, 28 having a decrease of 25·0–49·9%, 50 having a decrease of 10·0–24·9%, and 32 having a decrease of less than 10·0%. For the remaining 81 countries, we found no decrease in stillbirth rate since 2000. Of these countries, 34 were in sub-Saharan Africa, 16 were in east Asia and the Pacific, and 15 were in Latin America and the Caribbean. Interpretation: Progress in reducing the rate of stillbirths has been slow compared with decreases in the mortality rate of children younger than 5 years. Accelerated improvements are most needed in the regions and countries with high stillbirth rates, particularly in sub-Saharan Africa. Future prevention of stillbirths needs increased efforts to raise public awareness, improve data collection, assess progress, and understand public health priorities locally, all of which require investment. Funding: Bill & Melinda Gates Foundation and the UK Foreign, Commonwealth and Development Office.

For this systematic assessment, we present a summary of the source data and methods; a more detailed description of the methods is available in Wang et al.9 We created a database with 2833 country-year datapoints from 171 countries starting in the year 2000 up to the year 2019, updating and further developing the database used in 2016 by WHO.7 We extracted stillbirth rates from nationwide administrative registration systems such as vital registration systems, medical birth or death registries and HMIS, nationally representative household surveys with pregnancy histories or reproductive calendars, and population-based studies. Subnational population-based study data were sought for all countries without high coverage of routine administrative data from registration systems (appendix pp 3–5). Definitional adjustment of stillbirth data was required, given that stillbirths were reported inconsistently in countries according to different combinations of definitional criteria, including gestational age, birthweight, or, occasionally, length at birth, and with varying thresholds (panel 1). In some instances, no clear criteria or thresholds were provided. These differences make it difficult to compare stillbirth rates and trends across countries and to calculate the global burden, as highlighted previously.7, 13, 14, 15 We estimated stillbirth rate using a 28 weeks’ gestation or more definition of stillbirth (panel 1). If information for the 28 weeks’ gestation definition was not available, adjustments and additional uncertainty associated with alternative definitions were accounted for in the model fitting.9 For each definitional conversion, we estimated the mean and variance associated with the ratio of the expected stillbirth rate based on an alternative definition, to the expected stillbirth rate based on the 28 weeks’ gestation or more definition. For low-income and middle-income countries (LMICs), high quality data from studies16, 17 were used to calculate adjustments and variance; for high-income countries (HICs), national administrative data were used. The World Bank Group income classification of countries from the year 2020 was used. The term stillbirth generally applies to a baby born with no signs of life after a given viability threshold, with viability typically assessed on the basis of gestational age, birthweight, or length at birth. A stillbirth is defined as the birth of a baby following fetal death before labour (antepartum stillbirth) or during labour or birth (intrapartum stillbirth). Although most stillbirths occur within hours or days of fetal death, occasionally, such as in the case of twins, this can be delayed by months. For international comparisons of stillbirths, the International Classification of Diseases (ICD) definition (ICD 10th and 11th revisions)10, 11 of late fetal deaths is used. ICD defines late fetal death as the in-utero death of a baby (ie, born with no signs of life at birth) with a birthweight of 1000 g or more; or if birthweight is not available, at a gestational age of 28 weeks or more, or (if gestational age is not available), a body length of 35 cm or more at birth. Early fetal death is defined as the in-utero death of a baby (ie, born with no signs of life at birth) with a birthweight of 500–999 g; or, if birthweight is not available, at a gestational age at birth of 22–27 weeks, or a body length of 25–34 cm at birth. Since gestational age and birthweight thresholds do not perfectly correspond, the UN Inter-agency Group for Child Mortality Estimation (UN IGME) and the Core Stillbirth Estimation Group (CSEG) recommend the use of gestational age rather than birthweight to define a stillbirth. Gestational age is a better predictor of maturity and hence viability; of note, gestational age is the most commonly available criterion across data sources globally. Gestational age is typically measured from the first day of the last normal menstrual period,10 although in circumstances in which early ultrasound dating scans are available, gestational age should be based on the best obstetric estimate to avoid recall errors and differences in the length of menstrual cycles.12 Recommendations from the UN IGME and the CSEG also include omitting the birth length criterion and making a clearer distinction between stillbirth and fetal death. These recommendations are under review for inclusion in an updated edition of ICD-11. Consistent with these recommendations, in this paper we defined a stillbirth as the birth of a baby with no signs of life at or after 28 weeks of gestation. When possible, data with a 28 weeks or more gestation definition were extracted. When data were collected according to a different definition (eg, based on birthweight or an alternative gestational age definition), stillbirth rates were adjusted in the modelling to allow for consistent international comparisons. For the estimates presented in this paper, stillbirth rate was defined as the number of stillbirths at 28 weeks’ gestation or more per 1000 total births (ie, livebirths plus stillbirths). We assessed the quality of the various types and sources of stillbirth data by evaluating completeness and consistency. Data were excluded if the definition of stillbirth used or the method of data collection was not specified, more than 50% of reported stillbirths had unknown gestational age or birthweight, or coverage of livebirths in administrative registration data systems was estimated to be lower than 80% (or 75% for HMIS). Registration data with incomplete coverage of child deaths (<95%) were also excluded on the basis of WHO completeness assessments that used the same threshold.18 Additionally, data were excluded on the basis of external information that suggested some stillbirth rate observations were unreliable, for example due to poor quality of the data source, known data quality issues, undercapture of stillbirths, or inconsistency in reported numbers. As part of the assessment of data quality, the plausibility of the ratio of stillbirth rate (measured according to the 28 weeks’ gestation or more definition) to neonatal mortality rate (for babies aged <28 days) from the same data source was determined. In the case of HMIS data, for which data on neonatal mortality rate might be less reliable than data on stillbirths as neonatal deaths are more likely to occur outside the health facility, or if the HMIS or other data source did not contain neonatal mortality rate, the UN IGME neonatal mortality rate estimates19 were used to calculate the ratio of stillbirth rate to neonatal mortality rate for assessment purposes. The UN IGME neonatal mortality rates were estimated within a Bayesian hierarchal framework at the country level and aggregated to region and global levels. The observed stillbirth rate to neonatal mortality rate ratios were compared to the distribution of ratios obtained from high quality LMIC study data.16, 17 We excluded observations with extremely low ratios using methods detailed in Wang et al.9 In summary, if stillbirths were under-reported relative to neonatal deaths for a country-year datapoint, the associated observed ratio of stillbirth rate to neonatal mortality rate would be lower than the true ratio. To quantify whether an observed ratio from our global dataset was extremely low, we calculated the probability of obtaining a ratio that is smaller than the observed ratio (taking account of the uncertainty associated with the observed ratio) using the distribution of ratios obtained from the high quality data. If this probability was less than 0·05, the observation was excluded from the database. This approach was applied to all observations in the database with 28 weeks’ gestation or more definitions and adjusted definitions (appendix pp 5–6). Due to data quality concerns, 1302 (46·0%) of 2833 datapoints on stillbirths were excluded from the model (regional distribution shown in the appendix [p 4]). Among 195 countries for which we generated stillbirth estimates, 24 countries had no stillbirth data at all and 38 countries had no good quality stillbirth data, after excluding data according to our criteria (appendix pp 37–40). We estimated stillbirth rates using a Bayesian hierarchical temporal sparse regression model for all country-years (appendix p 6). In the model, stillbirth rate was estimated assuming that the logarithm of the observed stillbirth rate plus adjustments and random measurement error equals the logarithm of the true stillbirth rate. Adjustments included those related to application of definition conversions and source type bias. Source type bias was equal to zero for all observations except for those from surveys, which were assumed to have a negative bias associated with them as surveys have been shown to underestimate stillbirths.20 Random measurement error referred to the sum of the stochastic or sampling error, the random definitional adjustment, and a random error related to source type. Each error was expected to be zero on average but included a variance term that reflected how much uncertainty was associated with the error. The stochastic or sampling error was due to not observing the complete population or survey sampling design. The random definitional adjustment error was non-zero for alternative definitions of stillbirth (ie, not the ≥28 weeks’ gestation definition) and followed from the analysis of the definitional adjustment ratios. The source type error referred to variances specific to source type, which accounted for random errors that might occur in the data collection process, and potential non-representativeness of observations. The distinct data source types considered in the model were administrative registration data (including vital registration systems and birth and death registries), HMIS, household surveys, and population-based studies. The estimated stillbirth rate (on the logarithmic scale) for each country for the years 2000–19 was given by the sum of a regression function with a country-specific intercept and a country-specific temporal smoothing process. Resulting estimates were a weighted combination of information from adjusted country data and covariates associated with stillbirth rate, and accounted for the varying uncertainty associated with the adjusted observations. If data were precise when accounting for biases, uncertainty, and non-sampling errors, the stillbirth rate point estimates followed the adjusted country data. In cases of no data or imprecise data, the estimates were based on covariates. The uncertainty associated with the stillbirth rate estimates depended on data availability and precision for the respective country-period; uncertainty decreased as data availability and precision increased. Uncertainty in stillbirth rate estimates increased when extrapolating to periods without data. The candidate covariates were based on a conceptual framework published in 2016 by Blencowe and colleagues.7 The framework included distal determinants such as socioeconomic factors, inter-related and overlapping demographic and biomedical factors (eg, adolescent fertility rate, maternal age, and malaria prevalence), perinatal outcome markers associated with stillbirth, and access to health care. The covariate data from household surveys, such as coverage of antenatal care visits and proportion of caesarean deliveries, were smoothed with a time-series trend to reduce small fluctuations in measured covariates. In the model fitting, regression coefficients for covariates with low predictive power were shrunk towards zero with sparsity-inducing priors9 as part of the Bayesian hierarchical temporal sparse regression model. The final covariates used in the model were neonatal mortality rate, low birthweight rate (on a logarithmic scale), coverage of four or more antenatal care visits, caesarean section rate, mean years of schooling of females, and gross national income per capita (on a logarithmic scale). A more detailed description of the model is available in Wang et al9 and covariate coefficients are available in the appendix (page 10). We used a Hamiltonian Monte Carlo algorithm implemented with the use of Stan21 and R package RStan22 to generate samples from the posterior distributions of stillbirth rate. Given the inherent uncertainty in stillbirth rate estimates, 90% uncertainty intervals (UIs) are used by the UN IGME instead of the more conventional 95% intervals. Although reporting intervals that are based on higher uncertainty (ie, 95% instead of 90%) would reduce the chance of not including the true value in the interval, the disadvantage of choosing higher uncertainty is that intervals lose their utility in presenting meaningful summaries of a range of likely outcomes when the indicator of interest is highly uncertain. The resulting UIs are not necessarily symmetrical around the point estimates, as stillbirth rates were estimated on the logarithmic scale, but reflect the uncertainty range associated with the stillbirth rates. The UIs for the number of stillbirths generated by the UN IGME do not account for uncertainty associated with other inputs required for calculation, such as the number of livebirths, because uncertainty assessments of these inputs are not yet available. The number of stillbirths was calculated from the number of livebirths estimated by the UN Population Division,23 according to the formula: number of stillbirths=livebirths × [stillbirth rate/(1 – stillbirth rate)]. The codes used in the model are available on request. We generated stillbirth rate estimates for 195 countries from 2000 to 2019. We produced regional aggregates from the country estimates and countries within these regions were defined according to UNICEF's regional classifications (appendix pp 7–8) and the World Bank 2020 income classification. We computed the percentage change and the annual rate of reduction with 90% UIs in the stillbirth rate and number of stillbirths for selected periods (2000–19, 2000–09, and 2010–19). The annual rate of reduction was defined as log(rate in t2/rate in t1)/(t1–t2), where t1 and t2 refer to different years (t1<t2). We compare the estimated progress in stillbirth rate with the UN IGME estimates for mortality among children younger than 5 years (neonatal mortality and child mortality at 1–59 months)19 and the WHO, UNICEF, UN Population Fund, World Bank Group, and UN Population Division estimates of maternal mortality ratio (number of maternal deaths per 100 000 livebirths).24 We also present the ratio of stillbirth rate to UN IGME neonatal mortality rate estimates (with 90% UIs). The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.

The systematic assessment mentioned in the description focuses on estimating stillbirth rates and trends globally from 2000 to 2019. The study uses a Bayesian hierarchical temporal sparse regression model to estimate stillbirth rates for 195 countries. The model incorporates data from various sources, including registration and health information systems, household surveys, and population-based studies. The study also compares the estimated stillbirth rates with neonatal mortality rates and child mortality rates to assess progress. The findings highlight the need for accelerated improvements in regions and countries with high stillbirth rates, particularly in sub-Saharan Africa. The study emphasizes the importance of raising public awareness, improving data collection, assessing progress, and understanding local public health priorities to prevent stillbirths. The study was funded by the Bill & Melinda Gates Foundation and the UK Foreign, Commonwealth and Development Office.
AI Innovations Description
The systematic assessment described in the source aims to estimate stillbirth rates and trends globally from 2000 to 2019. The study collected data from various sources, including registration systems, household surveys, and population-based studies. The researchers used a Bayesian hierarchical temporal sparse regression model to estimate stillbirth rates for 195 countries.

The findings of the assessment revealed that globally, an estimated 2.0 million babies were stillborn in 2019, with a stillbirth rate of 13.9 stillbirths per 1000 total births. Stillbirth rates varied across regions, with the highest rates observed in west and central Africa, eastern and southern Africa, and south Asia. The study also found that progress in reducing stillbirth rates has been slower compared to decreases in mortality rates among children under 5 years.

Based on the assessment, the following recommendations can be made to improve access to maternal health and reduce stillbirth rates:

1. Raise public awareness: Increase efforts to educate communities about the importance of maternal health and the prevention of stillbirths. This can be done through targeted campaigns, community engagement, and health education programs.

2. Improve data collection: Enhance the quality and coverage of data on stillbirths by strengthening registration systems, improving data collection methods, and ensuring consistent reporting of stillbirths according to standardized definitions.

3. Assess progress: Regularly monitor and evaluate the progress in reducing stillbirth rates at the national and global levels. This can help identify areas of improvement and guide the allocation of resources and interventions.

4. Understand local public health priorities: Conduct research and analysis to identify the specific factors contributing to high stillbirth rates in different regions and countries. This understanding can inform the development of targeted interventions and policies to address the underlying causes.

5. Increase investment: Allocate sufficient resources and funding to support maternal health programs, including access to antenatal care, skilled birth attendance, emergency obstetric care, and postnatal care. Investment in healthcare infrastructure, training of healthcare providers, and community-based interventions is crucial to improving access to maternal health services.

By implementing these recommendations, it is possible to develop innovative approaches and interventions that can effectively improve access to maternal health and reduce stillbirth rates globally.
AI Innovations Methodology
The systematic assessment described in the provided text focuses on estimating stillbirth rates and trends globally from 2000 to 2019. The study aims to provide standardized measurements of stillbirth rates across countries and assess progress over time. The methodology used in this assessment involves creating a database of country-year data points from various sources, including registration and health information systems, household surveys, and population-based studies. The data quality is assessed, and adjustments are made for inconsistencies in stillbirth definitions used by different countries. A Bayesian hierarchical temporal sparse regression model is then used to estimate country-specific stillbirth rates, taking into account covariates and temporal smoothing processes. The estimated stillbirth rates and trends are compared to mortality estimates for children under 5 years of age. The study highlights the need for accelerated improvements in reducing stillbirth rates, particularly in regions with high rates such as sub-Saharan Africa. The methodology used in this assessment provides a framework for estimating stillbirth rates and can be used to monitor progress and inform public health interventions.

Partagez ceci :
Facebook
Twitter
LinkedIn
WhatsApp
Email