Improving risk prediction of adverse outcomes in hemodialysis patients by incorporating non-traditional risk factors

点击翻译按钮获取中文摘要

PROJECT SUMMARY: Candidate: The primary objective of this application is to support Dr. Lili Chan's career development into an independently funded clinical investigator leveraging electronic health records (EHR) and improve risk prediction of adverse outcomes in patients on hemodialysis (HD) by incorporating social determinants of health. To accomplish this goal, Dr. Chan has assembled a multidisciplinary mentoring and advisory team lead by Dr. Steven Coca, Associate Professor of Medicine and Director of Clinical Research in Nephrology at the Icahn School of Medicine at Mount Sinai, and co-mentor Dr. Peter Kotanko, Adjunct Professor of Medicine at Mount Sinai and Research Director of the Renal Research Institute. Her advisory team consists of Dr. Weng, an expert and in machine learning and natural language processing (NLP), Dr. Alex Federman, who has contributed significantly to the literature on the effects of psychosocial factors on patient care, and Dr. Mazumdar, an expert in biostatistics and risk prediction modeling. Dr. Chan's proposed training plan focuses on four areas, (1) advanced statistical methodology; (2) bioinformatics; (3) patient centered outcomes; and (4) career development. Environment: The Icahn school of Medicine at Mount Sinai is a national leader in research. Specifically the Division of Nephrology has over 30 funded investigators and has successfully mentored five faculty members from K awards to R01 awards. Research: Given the high morbidity and mortality of HD patients, there is a critical need for better risk stratification and identification of high risk groups in order for targeted interventions to be tested. This project utilizes prospectively collected surveys and retrospective chart review of a cohort of diverse patients on chronic HD who receive care from four Renal Research Institute and six Mount Sinai Health System hemodialysis units located throughout New York City. The Specific Aims of the research are: (1) to determine the association between domains of social determinants of health and hospitalizations using survey research methods; (2) to identify social determinants of health in an accurate manner using natural processing language; and (3) to create risk prediction models for hospitalization among patients on HD utilizing both standard measures and social determinants of health using standard statistical methods and machine learning. This research leverages novel computational methods to examine the association of social determinants of health and hospitalizations in HD patients and incorporates SDOH into risk prediction models which will allow for identification of high risk HD patients for inclusion in future intervention trials. The results of this proposal sets the foundation for future R01 studies validating these findings in external data sets and testing the utility of EHR integrated clinical decision tools on reducing hospitalizations, readmissions, and mortality.