Developing innovative analytics to estimate age-and cause-specific child mortality for low- and middle-income countries

开发创新分析来估计低收入和中等收入国家的年龄和特定原因儿童死亡率

• 批准号: 9766323
• 负责人: Li Liu
• 金额: $ 19.02万
• 依托单位: JOHNS HOPKINS UNIVERSITY
• 依托单位国家: 美国
• 财政年份: 2018
• 资助国家: 美国
• 起止时间: 2018-08-17 至 2021-07-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/9766323
• 关键词: 5 year old; Age; Area; Autopsy; Bangladesh; Bayesian Modeling; Categories; Cause of Death; Cessation of life; Child; Child Mortality; Childhood; China; Communities; Country; Custom; Data; Data Quality; Databases; Decision Making; Dependence; Development; Disease; Effectiveness; Epidemiology; Evaluation; Foundations; Future; Generations; Goals; Government; Grouping; Heterogeneity; Impact evaluation; Income; International; Intervention; Investments; Knowledge; Life; Life Table Models; Logic; Longevity; Measurement; Methodology; Methods; Modeling; Neonatal Mortality; Pattern; Performance; Policies; Policy Developments; Policy Making; Program Evaluation; Publishing; Reproducibility; Research; Resolution; Resource Allocation; Sampling; Savings; Specificity; Sustainable Development; System; Uncertainty; Ursidae Family; Variant; Work; age group; base; human mortality; innovation; insight; low and middle-income countries; low income country; mortality; novel; prevent; programs; prototype; scale up; theories

Project Summary Globally, an estimated 5.9 million children died before reaching their fifth birthday in 2015. The majority died in low- and middle-income countries (LMICs), where quality information on age- and cause-specific child mortality (ACSCM) is rarely available. Recently, the US government and the international community have renewed their commitment to end preventable child deaths in a generation. We have been publishing modeled national COD distributions for LMICs since 2010, where we estimated COD distribution for 0-1 and 1-59-month olds. However, demographic and epidemiological evidence amounts to the conclusion that child COD is not uniform in the 1-59-month period. National empirical data at levels of specificity below 1-59 months are often not available in LMICs due to weak civil registration systems. Such data and estimates bear considerable scientific value to inform the development and impact evaluation of age-specific childhood interventions and their scale-up. Therefore, understanding the COD distribution among finer age groups in the 1-59-month period is warranted. Previous research has suffered from four main drawbacks: (i) using custom-collected data to understand age dynamics in a single cause; (ii) estimating ACSCM only in broad age groups; (iii) producing estimates in each age group separately and independently; and (iv) developing ACSCM in two separate estimation frameworks. The goals of this study are to systematically describe and make publicly available empirical age patterns of child COD in LMICs with accurate uncertainty intervals, and to develop the innovative theory-driven, parsimonious Bayesian hierarchical modeling framework to derive estimates of national COD distributions in LMICs with partial data among finer age groups than previous research. We will achieve the goals through three aims: 1) To extend and evaluate all-age demographic models to estimate age patterns in child deaths with VR data. 2) To conceptualize, develop and evaluate novel simultaneous ACSCM estimators using VR data in high-income countries; and 3) To extrapolate the unified ACSCM estimation framework to LMICs. The proposed study has two important innovations. First, it proposes the first unified framework for simultaneously estimating all-age, all-cause, age- and cause-specific child mortality. If successful, the study will offer systematically estimated ACSCM with valid uncertainty for selected LMICs, and lay the foundation for developing methods to systematically estimate ACSCM for all LMICs, including those low quality, limited, or even no data. Second, this framework produces estimates at an age granularity not yet seen in published research. This additional information is crucial to enable under-five child survival policy development and program evaluation at granular levels of ages and causes that would further contribute to the Sustainable Development Goals of equitably in reducing under-5 and neonatal mortality rates across countries. This study will lay the groundwork for future research, such as extending the Human Mortality Database to children under five to produce quality assessed, bias adjusted, and systematically organized ACSCM estimates.

项目摘要 2015年，全球估计有590万儿童在五岁生日前死亡。大多数人死于 在低收入和中等收入国家，按年龄和原因分列的儿童死亡率的高质量信息 （ACSCM）很少可用。近日，美国政府和国际社会再度对 他们承诺在一代人的时间内结束可预防的儿童死亡。我们一直在出版全国性的 自2010年以来，LMIC的COD分布，我们估计了0-1和1-59个月奥尔兹的COD分布。 然而，人口统计学和流行病学证据得出结论，儿童死因是不统一的 在1-59个月期间。具体程度低于1-59个月的国家经验数据往往不是 由于民事登记系统薄弱，中低收入国家的人口不多。这些数据和估计具有相当大的科学性 为制定和评价针对特定年龄儿童的干预措施及其 按比例放大。因此，了解1-59个月期间较细年龄组之间的COD分布， 有正当理由以前的研究有四个主要缺点：（一）使用定制收集的数据， 了解单一原因的年龄动态;（ii）仅在广泛的年龄组中估计ACSCM;（iii）产生 （iv）以两个不同的年龄组别，分别发展儿童及青少年自我评估及自我评估策略 估计框架。本研究的目标是系统地描述并公开提供 经验的年龄模式的儿童COD在低收入国家与准确的不确定性区间，并制定创新的 理论驱动的，简约的贝叶斯分层建模框架，以获得国家COD的估计 LMIC中的分布，部分数据在比以前的研究更精细的年龄组中。我们将实现 （1）扩大和评价所有年龄人口模型，以估计 儿童死亡VR数据2)概念化，开发和评估新的同步ACSCM估计器 使用高收入国家的VR数据;以及3）推断统一的ACSCM估计框架， 中低收入国家。这项研究有两个重要的创新。首先，它提出了第一个统一的框架， 同时估计所有年龄、所有原因、按年龄和原因分列的儿童死亡率。如果成功，研究 将为选定的低收入国家提供具有有效不确定性的系统估计的ACSCM，并为 制定方法，系统地估计所有低收入国家的ACSCM，包括那些低质量、有限或 甚至没有数据。第二，这个框架产生了一个年龄粒度的估计，在出版的 research.这一补充资料对于制定五岁以下儿童生存政策至关重要， 在年龄和原因的粒度级别进行计划评估，这将进一步有助于可持续发展 公平地降低各国5岁以下儿童和新生儿死亡率。本研究 这将为未来的研究奠定基础，例如将人类死亡率数据库扩展到儿童， 第五，产生质量评估，偏差调整，并系统地组织ACSCM估计。

项目成果

Methods for correcting inference based on outcomes predicted by machine learning.

• DOI: 10.1073/pnas.2001238117
• 发表时间: 2020-12-01
• 期刊: Proceedings of the National Academy of Sciences of the United States of America
• 影响因子: 11.1
• 作者: Wang S;McCormick TH;Leek JT
• 通讯作者: Leek JT