K23 Abstract This application is for a K23 Mentored Clinical Scientist Research Career Development Award entitled ?Developing Machine Learning-Driven Prediction Models and Therapeutic Strategies for Circulatory Shock in Critically-ill Patients?. I am a pulmonary and critical care physician at the University of Pittsburgh. This award will facilitate my acquisition of advanced training in clinical research methods, clinical informatics, and computer science to develop my career as a physician-scientist focused on data-driven studies of dynamic physiology in critically-ill patients. The main objective of this proposal is to develop individualized prediction models and treatment strategies for shock among critically-ill patients.
The aims of this study are: 1) To build machine learning-based prediction models of tachycardia and hypotension following blood donation using non-invasive waveform data in healthy blood donor volunteers, and create baseline features to compare with circulatory shock 2) To provide an operational definition, prediction models, differentiation of physiologic evolution towards shock, and personalized treatment of circulatory shock in ICU patients. Through this proposal, I will develop advanced skills in machine-learning, clinical bioinformatics, and clinical research. I will complete a Master of Science in Biomedical Informatics to learn advanced data- driven research methodologies to strengthen my technical training. This award will be a critical step towards my long-term goal, being an independent physician scientist, with expertise in prediction analytics in critical care medicine through clinical trials. I have committed mentors Dr. Michael Pinsky (physiology, functional hemodynamics) and Dr. Gilles Clermont (critical care, algorithms, data science) who will ensure successful completion of my proposed aims. My mentoring committee also includes an advisor, Dr. Milos Hauskrecht - a renowned computer scientist in the Computer Science and Information Sciences at the University of Pittsburgh. My work will be completed within the Division of Pulmonary, Allergy, and Critical Care Medicine at the University of Pittsburgh, which has an extensive track record of committing to the development of physician scientists.
Circulatory shock is a major cause of morbidity and mortality among critically ill patients. Early identification or forecasting of shock would be clinically actionable and improve prognosis, but no personalized prediction model for shock has been presented yet. We propose to develop physiologically and clinically relevant machine learning-based prediction models and therapeutic strategies that could guide personalized diagnosis and treatment of circulatory shock.
