Dataset Information

Machine learning guided postnatal gestational age assessment using new-born screening metabolomic data in South Asia and sub-Saharan Africa.

ABSTRACT:

Background

Babies born early and/or small for gestational age in Low and Middle-income countries (LMICs) contribute substantially to global neonatal and infant mortality. Tracking this metric is critical at a population level for informed policy, advocacy, resources allocation and program evaluation and at an individual level for targeted care. Early prenatal ultrasound examination is not available in these settings, gestational age (GA) is estimated using new-born assessment, last menstrual period (LMP) recalls and birth weight, which are unreliable. Algorithms in developed settings, using metabolic screen data, provided GA estimates within 1-2 weeks of ultrasonography-based GA. We sought to leverage machine learning algorithms to improve accuracy and applicability of this approach to LMICs settings.

Methods

This study uses data from AMANHI-ACT, a prospective pregnancy cohorts in Asia and Africa where early pregnancy ultrasonography estimated GA and birth weight are available and metabolite screening data in a subset of 1318 new-borns were also available. We utilized this opportunity to develop machine learning (ML) algorithms. Random Forest Regressor was used where data was randomly split into model-building and model-testing dataset. Mean absolute error (MAE) and root mean square error (RMSE) were used to evaluate performance. Bootstrap procedures were used to estimate confidence intervals (CI) for RMSE and MAE. For pre-term birth identification ROC analysis with bootstrap and exact estimation of CI for area under curve (AUC) were performed.

Results

Overall model estimated GA had MAE of 5.2 days (95% CI 4.6-6.8), which was similar to performance in SGA, MAE 5.3 days (95% CI 4.6-6.2). GA was correctly estimated to within 1 week for 85.21% (95% CI 72.31-94.65). For preterm birth classification, AUC in ROC analysis was 98.1% (95% CI 96.0-99.0; p < 0.001). This model performed better than Iowa regression, AUC Difference 14.4% (95% CI 5-23.7; p = 0.002).

Conclusions

Machine learning algorithms and models applied to metabolomic gestational age dating offer a ladder of opportunity for providing accurate population-level gestational age estimates in LMICs settings. These findings also point to an opportunity for investigation of region-specific models, more focused feasible analyte models, and broad untargeted metabolome investigation.

SUBMITTER: Sazawal S

PROVIDER: S-EPMC8424940 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:BackgroundUnderstanding the age pattern of under-5 mortality is essential for identifying the most vulnerable ages and underlying causes of death, and for assessing why the decline in child mortality is slower in some countries and subnational areas than others. The aim of this study is to detect age patterns of under-5 mortality that are specific to low-income and middle-income countries (LMICs).MethodsIn this modelling study, we used data from 277 Demographic and Health Surveys (DHSs), 58 Health and Demographic Surveillance Systems (HDSSs), two cohort studies, and two sample-registration systems. From these sources, we collected child date of birth and date of death (or age at death) from LMICs between 1966 and 2020. We computed 22 deaths rates from each survey with the following age breakdowns: 0, 7, 14, 21, and 28 days; 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 15, 18, and 21 months; and 2, 3, 4, and 5 years. We assessed how probabilities of dying estimated for the 22 age groups deviated from predictions generated by a vital registration model that reflects the historical mortality of 25 high-income countries.FindingsWe calculated mortality rates of 81 LMICs between 1966 and 2020. In contrast with the other regions of the world, we found that under-5 mortality in south Asia and sub-Saharan Africa was characterised by increased mortality at both ends of the age range (ie, younger than 28 days and older than 6 months) at a given level of mortality. Observed mortality in these regions was up to 2 times higher than predicted by the vital registration model for the younger-than-28 days age bracket, and up to 10 times higher than predicted for the older-than-6 months age bracket. This age pattern of under-5 mortality is significant in 17 countries in south Asia and sub-Saharan Africa. Excess mortality in children older than 6 months without excess mortality in children younger than 28 days was found in 38 countries. In south Asia, results were consistent across data sources. In sub-Saharan Africa, excess mortality in children younger than 28 days was found mostly in DHSs; the majority of HDSSs did not show this excess mortality. We have attributed this difference in data sources mainly to omissions of early deaths in HDSSs.InterpretationIn countries with age patterns of under-5 mortality that diverge from predictions, evidence-based public health interventions should focus on the causes of excess of mortality; notably, the effect of fetal growth restriction and infectious diseases. The age pattern of under-5 mortality will be instrumental in assessing progress towards the decline of under-5 mortality and the Sustainable Development Goals.FundingEunice Kennedy Shriver National Institute of Child Health and Human Development of the National Institutes of Health.

Project description:BackgroundWidespread increases in facility delivery have not substantially reduced neonatal mortality in sub-Saharan Africa and South Asia over the past 2 decades. This may be due to poor quality care available in widely used primary care clinics. In this study, we examine the association between hospital delivery and neonatal mortality.Methods and findingsWe used an ecological study design to assess cross-sectional associations between the share of hospital delivery and neonatal mortality across country regions. Data were from the Demographic and Health Surveys from 2009 to 2018, covering 682,239 births across all regions. We assess the association between the share of facility births in a region that occurred in hospitals (versus lower-level clinics) and early (0 to 7 days) neonatal mortality per 1,000 births, controlling for potential confounders including the share of facility births, small at birth, maternal age, maternal education, urbanicity, antenatal care visits, income, region, and survey year. We examined changes in this association in different contexts of country income, global region, and urbanicity using interaction models. Across the 1,143 regions from 37 countries in sub-Saharan Africa and South Asia, 42%, 29%, and 28% of births took place in a hospital, clinic, and at home, respectively. A 10-percentage point higher share of facility deliveries occurring in hospitals was associated with 1.2 per 1,000 fewer deaths (p-value < 0.01; 95% CI: 0.82 to 1.60), relative to mean mortality of 22. Associations were strongest in South Asian countries, middle-income countries, and urban regions. The study's limitations include the inability to control for all confounding factors given the ecological and cross-sectional design and potential misclassification of facility levels in our data.ConclusionsRegions with more hospital deliveries than clinic deliveries have reduced neonatal mortality. Increasing delivery in hospitals while improving quality across the health system may help to reduce high neonatal mortality.

Project description:There is great untapped potential for farm mechanization to support rural development initiatives in low- and middle-income countries. As technology transfer of large machinery from high-income countries was ineffective during the 1980s and 90s, mechanization options were developed appropriate to resource poor farmers cultivating small and scattered plots. More recently, projects that aim to increase the adoption of farm machinery have tended to target service providers rather than individual farmers. This paper uses the Scaling Scan tool to assess three project case studies designed to scale different Mechanization Service Provider Models (MSPMs) in Mexico, Zimbabwe, and Bangladesh. It provides a useful framework to assess the gap between international lessons learned on scaling captured in forty tactical questions over ten "scaling ingredients" as perceived by stakeholders involved in the projects, as well as private sector actors and government employees. Although at first sight the case studies seem to successfully reach high numbers of end users, the assessment exposes issues around the sustainable and transformative nature of the interventions. These are highly influenced by the design of the projects and by the environment and context of the intervention areas. Across the three case studies, large-scale adoption of the models was found to be hampered by lack of finance to set up MSPMs and insufficient collaboration among the value chain actors to strengthen and foster Mechanization Service Provider (MSP) entrepreneurs. Applying a scaling perspective on each case study project exposed important lessons on minimizing project dependencies. Positive examples include integration of capacity development materials in vocational training centers in Zimbabwe, promotion of MSPMs by other donors in East Africa and levering of nearly USD six million of private sector investment in appropriate machinery in Bangladesh. On the other hand, there is still a high dependency on the projects in terms of coaching of service providers, facilitating collaboration along the value chain, and provision of leadership and advocacy to address issues at governance level. These results have important implications for similar development interventions aimed at increasing smallholder access to mechanization services at scale and is to our knowledge the first cross-continental assessment of these issues to date.

Project description:ObjectiveTo assess the comparative costs and effects of interventions to combat five neuropsychiatric conditions (schizophrenia, bipolar disorder, depression, epilepsy, and heavy alcohol use).DesignCost effectiveness analysis based on an epidemiological model.SettingTwo epidemiologically defined World Health Organization sub-regions of the world: countries in sub-Saharan Africa with very high adult and high child mortality (AfrE); and countries in South East Asia with high adult and high child mortality (SearD).Data sourcesPublished studies, costing databases.Main outcome measuresCost per capita and cost per disability adjusted life year (DALY) averted, expressed in international dollars ($Int) for the year 2005.ResultsAcross 44 assessed intervention strategies for the five neuropsychiatric conditions, cost effectiveness values differed by as much as two orders of magnitude (from $Int100-250 to $Int10,000-25,000 for a year of healthy life gained). In both sub-regions, inpatient based treatment of schizophrenia with newer antipsychotic drugs was the most costly and least cost effective strategy. The most cost effective strategies in the African sub-region related to population based alcohol control, while in the South East Asian sub-region the most cost effective intervention was drug treatment of epilepsy in primary care. The cumulative cost per capita of the most cost effective set of interventions covering all five conditions was estimated at $Int4.90-5.70. This package comprises interventions for epilepsy (older first line antiepileptic drugs); depression (generically produced newer antidepressants and psychosocial treatment); bipolar disorder (mood stabiliser drug lithium); schizophrenia (neuroleptic antipsychotic drugs and psychosocial treatment); and heavy alcohol use (increased taxation and its enforcement, reduced access, and, in the African sub-region, advertising bans and brief advice to heavy drinkers in primary care).ConclusionsReallocation of resources to cost effective intervention strategies would increase health gain, save money and help implement much needed expansion of services for neuropsychiatric conditions in low resource settings.