Predictive informatics monitoring in the Neonatal Intensive Care Unit

点击翻译按钮获取中文摘要

PROJECT SUMMARY/ABSTRACT Significance: Premature very low birth weight (VLBW) infants in the Neonatal ICU continue to suffer morbidity and mortality from sepsis and necrotizing enterocolitis (NEC), and early detection and treatment of these illnesses has been shown to improve survival and outcomes. Continuously monitored vital signs contain subtle changes in the early stage of sepsis and NEC, but these physiological markers are invisible with current technology, even for the most sophisticated monitors and experienced clinicians. Our group continues to build on experience and collaboration to develop early warning systems through advanced time series and machine learning data analytics that will improve outcomes for premature infants. Progress: Our first such early warning system, the HeRO monitor, alerts clinicians to abnormal heart rate characteristics and was shown to reduce sepsis-associated mortality by 40% in a randomized trial of 3003 VLBW infants. The past 7 years of NIH support produced several important discoveries, including 1) Adding pulse oximetry oxygenation data (SpO2) to electrocardiogram data improves sepsis and NEC prediction; 2) Center-specific differences in patient demographics and care practices impact algorithm performance; 3) High cross-correlation of heart rate and SpO2 reflects increased apnea and sepsis risk across multiple sites. Innovation: The proposed work represents a paradigm shift in patient care – monitors that report trends of development of health and illness rather than fleeting values, leading to improved outcomes of preterm infants in the NICU. Approach: In the current proposal we are adding a fourth large NICU to accomplish the following aims: Aim 1: Use advanced time series analytics and machine learning to refine and expand predictive monitoring algorithms for sepsis and NEC; Aim 2: Determine the impact of demographics and center on outcomes and algorithm performance; Aim 3: Share multi-center data and analytics globally by building on an existing platform. Investigators: Co-PI's Fairchild and Moorman have a longstanding collaboration and have led successful multicenter clinical and analytical research for the life of this grant and have strengthened the team by adding new collaborators and centers. Environment: The centers involved in this proposal are unique in their ability to collect and analyze large vital sign and clinical data sets, made possible by robust institutional research and computing support.