Leveraging Causal Inference and Machine Learning Methods to Advance Evidence-Based Maternal Care and Improve Newborn Health Outcomes

项目摘要/摘要 美国(美国)在新生儿健康方面，几乎落后于所有高收入国家，其比率 不良结局，包括围产期死亡率、早产和低出生体重，已经证明 近些年来进展甚微。巨大的差异是这些结果的基础，在 少数族裔和社会经济地位较低的个人。虽然这样的人口经常 在获得适当护理方面面临障碍，其他研究表明，某些低风险人群经历了 过度提供产妇护理。例如，产前护理和成像方面的上升趋势受到以下因素的影响 有限的严格证据表明，提高利用率已显著改善了结果。这一现实 有必要对产妇保健采取新的办法，包括：1)制定循证政策和 准则，特别是有利于弱势群体的准则；和2)更好地评估干预措施 对改善护理目标的益处不明确。与政策和临床相关的两个具体领域包括 获得产后立即避孕和产前胎儿监测的机会，有潜力的战略 改善新生儿结局，但现有证据有限。因此，这一行动的总体目标是 提案旨在就以下方面提供严格的证据：1)医疗补助政策，以激励立即提供 产后长效可逆避孕(IPP-LARC)；2)产后长效可逆避孕的范围和效果 产前胎儿监护。目标1利用来自三个州的州住院患者出院数据和一个合成的 评估医疗补助提供者为提供IPP-LARC而支付的费用对短产影响的对照设计 间歇期、早产和低出生体重。AIM 2利用来自大型、 综合保健系统和基于机器学习的倾向评分设计，以评估范围和 产前胎儿监护对围产儿死亡率及其他相关新生儿结局的影响。这个 拟议的研究与国家儿童健康和人类发展研究所的 (NICHD)促进生殖健康的研究优先事项，更好地了解 避孕、增强健康孕期能力和解决健康差距。全面的团契培训 该计划将得到一个多学科指导团队和协作培训环境的支持 致力于研究、专业和临床技能的发展。通过机器中的桥接方法 很少在孕产妇保健评价中使用的学习和准实验评价 政策和指南，这项建议将有助于为数据驱动的临床实践提供信息 产后立即避孕和产前胎儿监护。总而言之，这样的证据将有助于 在美国促进公平的新生儿结局。

PROJECT SUMMARY/ABSTRACT The United States (U.S.) trails behind almost all high-income nations with respect to newborn health, with rates of adverse outcomes, including perinatal mortality, preterm birth, and low birthweight, having demonstrated minimal progress in recent years. Significant disparities underlie such outcomes, with higher rates among racial and ethnic minorities and individuals of low socioeconomic status. While such populations frequently face barriers in accessing adequate care, other studies suggest that certain low-risk populations experience over-provision of maternal care. Rising trends in prenatal care and imaging, for example, are tempered by limited rigorous evidence that increases in utilization have meaningfully improved outcomes. This reality necessitates a renewed approach to maternity care, including: 1) generation of evidence-based policies and guidelines, particularly ones that benefit vulnerable populations; and 2) better assessment of interventions with ambiguous benefit to improve targeting of care. Two specific areas of policy and clinical relevance include access to immediate postpartum contraception and antepartum fetal surveillance, strategies with potential to improve newborn outcomes but for which existing evidence is limited. Accordingly, the overall goal of this proposal is to produce rigorous evidence on: 1) a Medicaid policy incentivizing provision of immediate postpartum long-acting reversible contraception (IPP-LARC); and 2) the scope and effect of antepartum fetal surveillance. Aim 1 utilizes state inpatient discharge data from three states and a synthetic control design to estimate the impact of Medicaid provider payments for provision of IPP-LARC on short birth intervals, preterm birth, and low birthweight. Aim 2 leverages electronic health record data from a large, integrated health system and a machine-learning-based propensity score design to assess the scope and effect of antepartum fetal surveillance on perinatal mortality and other related newborn outcomes. The proposed research aligns closely with the National Institute of Child Health and Human Development’s (NICHD) research priorities to promote reproductive health, better understand the health impacts of contraception, empower healthy pregnancies, and address health disparities. The overall fellowship training plan will be supported by a multidisciplinary mentorship team and a collaborative training environment dedicated to research, professional, and clinical skills development. Through bridging methods in machine learning and quasi-experimental evaluation that have rarely been used in the evaluation of maternal care policies and guidelines, this proposal will be instrumental in informing data-driven clinical practice around immediate postpartum contraception and antepartum fetal surveillance. Together, such evidence will help advance equitable newborn outcomes in the U.S.