Measuring Neonatal Regionalization

点击翻译按钮获取中文摘要

Care and outcomes for the 60,000 very low-birth-weight (VLBW; <1500g) infants born annually in the United States varies widely. National guidelines recommend that care be organized along hierarchical regionalized care delivery networks, but too often these vulnerable infants are born in hospitals whose capabilities don't match patient need. This necessitates postnatal transfer which has been associated with excess morbidity and mortality. To date, research on regional care networks has been thwarted by a lack of appropriate linked data sets and mathematical tools to understand care network characteristics and their effect on neonatal outcomes. We propose to bridge this gap and advance health outcomes science by gaining a deep understanding of network characteristics and their links to clinical care and outcomes. We will accomplish this by using linked data sets, not available elsewhere, that allow for analysis of the individual and joint contributions of multi-level factors, including network factors on clinical outcomes. In addition, we will apply network analysis, a branch of graphical mathematics to visually display and quantify regionalized care network characteristics. We propose a large, near population-scale, observational study to analyze routinely collected data from 2010 to 2020 from >290,000 VLBW infants (>50% of all VLBW infants in the United States) in ~520 NICUs using linked vital records and patient discharge data from 17 states. This study is designed to achieve 3 specific aims: 1) Quantify regionalization and structure of transfer networks for VLBW infants across the United States; 2) Test the association of network structure with clinical quality of care and outcomes; and 3) Model optimized structure of perinatal transfers networks. Our analyses will employ network analysis as an innovative tool to measure care regionalization focusing on a high impact primary outcome (survival without major morbidity), as a substantive departure from prior work. Machine learning will be used to provide information on optimal network structures in terms of effectiveness, equity and efficiency. These models will reveal how networks would need to be modified to satisfy optimization goals and reveal potential trade-offs. We have a long track record of impactful research funded by the National Institute of Health using this data. We also have an opportunity to investigate more granular questions in California (140 NICUs), which has unique existing linkages to maternal and infant clinical and transport data. We expect our research to have an immediate positive impact because it is designed to result in actionable information for policy makers, administrators and clinicians to improve perinatal care delivery and equity.