I am a recent doctoral graduate from the Design Science Program at the University of Michigan College of Engineering and a member of the Machine Learning for Learning Health Systems lab of Dr. Karandeep Singh. Design Science is a program spanning multiple colleges and schools across the University of Michigan focused on interdisciplinary problem-solving. I applied this approach to data science, combining methods from computer science, engineering, and medicine. My research focuses on using electronic health record data to inform decision-making with healthcare analytics and clinical predictive modeling.
I earned my Ph.D. under the advisement of Dr. Karandeep Singh, Assistant Professor of Learning Health Sciences, and Dr. Thomas Klumpner, Assistant Professor of Anesthesiology. During my doctoral research, I co-developed and applied a framework to prepare electronic health record data for clinical predictive modeling. I developed the framework to reduce time spent manually preparing data and facilitate multi-disciplinary research through a common language, which we call a grammar of prediction models. I applied this framework to predicting postpartum hemorrhage using data from Michigan Medicine on a research team with several clinicians funded by the U-M Precision Health initiative. Using this framework, we developed an early warning system model that outperformed existing systems used in maternal care at Michigan Medicine. Though applied to the clinical area of obstetrics and gynecology, my doctoral work is readily transferable to other domains facing similar challenges in data preparation for predictive modeling. In addition to Precision Health funding, I also applied for and received funding support from the Rackham Merit Fellowship, Institute for Healthcare Policy and Innovation, and the Michigan Integrated Center for Health Analytics & Medical Prediction throughout my doctoral work. Improving clinical care requires health systems to identify high-risk patients and then shift resources to ensure that care is delivered equitably. Predictive modeling offers efficiencies that can complement provider education and patient care. With a human-centered approach, harmonizing the balance between domain expertise and risk estimates generated in real-time using machine learning, we can improve clinical and non-clinical outcomes.