Machine Learning Risk Prediction of Kidney Disease After Extremely Preterm Birth

点击翻译按钮获取中文摘要

PROJECT SUMMARY The overall goals of this 3 year mentored patient-oriented research career development award are to develop tools to improve recognition of kidney disease risks among extremely preterm born children and to support the development of Dr. Keia Sanderson into an independent investigator. Dr. Sanderson is a motivated clinical researcher at the University of North Carolina at Chapel Hill (UNC), with a specific interest in early recognition of kidney disease in children. Over the next 3 years, Dr. Sanderson will work with her mentorship committee toward her goal of independence building upon her prior 3 years of KL2-supported training. Her mentorship committee includes experts in long-term neonatal outcomes research (O’Shea, Laughon), neonatal nephrology (Askenazi), machine learning (Kosorok), qualitative research (Flythe), and clinical decision support science (Kistler). Each member has an established track record of mentoring junior faculty, consistent peer-reviewed support, and high research productivity. Dr. Sanderson’s career development objectives are to: 1) gain multi- center research leadership experience; 2) develop skills in machine learning methods for large data risk stratification; 3) develop qualitative research skills to support practical application of machine learning models; 4) acquire skills in research mentorship; and 5) improve publication and grant writing in preparation for NIH R01 applications. The research and training environment at UNC is well established to support these objectives. To achieve her career development objectives, Dr. Sanderson will participate in structured coursework, conduct mentored research, and workshops through North Carolina Translational and Clinical Sciences Institute R-Writing Group. The specific aims of this research project are to 1) utilize machine learning approaches within large databases of prenatal, neonatal, and early life exposure variables to predict kidney disease in adolescents born extremely premature; 2) develop and evaluate the usability and acceptability of a prototype web-based risk stratification tool for pediatric kidney disease after extremely preterm birth using variables hypothesized to predict kidney disease. Expanding on two existing prospective cohorts (Extremely Low Gestational Age Newborn-Environmental influences on Child Health Outcomes (ELGAN-ECHO) and the Preterm Erythropoietin Neuroprotection Trial (PENUT) cohorts), Dr. Sanderson will utilize clinical variables to identify the combination of “at-highest risk,” and “at-risk” predictors for pediatric chronic kidney disease in children after extremely preterm birth. This research will be the basis for R-level NIH applications to develop a finalized web-based risk prediction tool informed by this proposal for public dissemination, to expand the use of machine learning derived risk stratification tools to other medically complex pediatric populations, and to test whether use of risk prediction tools increases the use of critical interventions (e.g. angiotensin converting enzyme inhibitors) to prevent kidney disease progression in high risk children.