Predicting Adverse Events For ICU Patients Using Physiological Measures Taken At Admission
Patients in the critical care setting are closely monitored and constantly intervened upon. While some are deemed higher risks than others using metrics such as SAPS II and other physiological scores patients of all subtypes experience adverse events. We are interested in characterizing common “patterns of care”° in the ICU and how these patterns of care may differ across patient subtypes. We will attempt to better understand patient ICU data by creating individual visualizations of their stays determine subtypes based on physiological states and use machine learning techniques to find patterns of care using data from the MIMIC III database. We hope to learn which patterns of care may be associated with adverse events in order to prevent those.

I am a 6-7 and am interested in the applications of computer science to medicine. I started conducting research in Professor Guttag’s group fall of 2014 applying machine learning to predict medical outcomes. I will continue work on a screening tool for speech and language impairments in children. I hope to learn more about machine learning and am excited to work on something that potentially will have a big impact.