Data-Driven Identification of the Acute Respiratory Distress Syndrome

点击翻译按钮获取中文摘要

PROJECT SUMMARY/ABSTRACT This K01 proposal will complete Michael Sjoding, MD, MSc's training towards his long-term career goal of improving care of patients with acute respiratory disease. Dr. Sjoding is a Pulmonary and Critical Physician at the University of Michigan with master's level training in clinical study design and biostatistics. This proposal builds on Dr. Sjoding's prior expertise, providing protected time for additional training in data science, the technical methods for deriving new knowledge about human disease from “Big Biomedical Data” in the rich training environment at the University of Michigan. The project's research goal is to develop real-time systems to improve accuracy and timeliness of Acute Respiratory Distress Syndrome (ARDS) diagnosis using electronic health record data. ARDS is a critical illness syndrome affecting 200,000 people each year with high mortality. Under-recognition of this syndrome is the key barrier to providing evidence-based care to patients with ARDS. The research will be completed under the guidance of primary mentor Theodore J. Iwashyna, MD, PhD and co-mentors Timothy P. Hofer, MD, MSc, and Kayvan Najarian, PhD, and a scientific advisory board with additional expertise in data science and applied clinical informatics. The 5-year plan includes didactic coursework, mentored research, and professional development activities, with defined milestones to ensure successful transition to independence. The mentored research has 2 specific Aims: Aim 1. Develop a novel system for identifying ARDS digital signatures in electronic health data to accurately identify patients meeting ARDS criteria. Aim 2. Define the early natural history of developing ARDS, to more accurately predict patients' future ARDS risk. Both Aims will utilize rigorous 2-part designs, with the ARDS diagnostic and prediction models developed in the same retrospective cohort and validated in temporally distinct cohorts. In completing these high-level aims, the research will leverage high-resolution electronic health record and beside-monitoring device data to study ARDS with unprecedented detail, providing new insights into ARDS epidemiology and early natural history. This work will build to at least two R01 proposals: (1) testing the impact of a real-time electronic health record- based ARDS diagnostic system to improve evidence-based care practice, (2) defining ARDS subtypes using deep clinical phenotypic data. The work will build toward a programmatic line of research using high-resolution electronic health data to improve understanding of critical illness and respiratory disease. In completing this proposal, Dr. Sjoding will acquire unique computational expertise in data science methods, complementing his previous training, which he can then readily apply to address other research challenges in respiratory health. The ambitious but feasible training and mentored research proposed during this K01 award will allow him to achieve his goal of becoming an independent investigator.