TITLE:  Predictive Health Analytics

ABSTRACT:

In 2014, the United States spent $3 trillion in health care, equivalent to 17.2% of the US GDP. About one third of this amount ($971.8 billion) is attributed to hospital care. Evidently, even modest efforts for preventing and/or streamlining care in a hospital setting matter. 

In this talk, I will outline our recent work leveraging Electronic Health Records at the Boston Medical Center to predict hospitalizations and re-hospitalizations. Our current work makes hospitalization predictions within the next year for heart-disease patients and patients with diabetes, as well as, re-hospitalization predictions within 30 days of general surgery. There are interesting lessons to be learned by working with such data. Specifically, very sparse classifiers that focus on a small feature subset for each patient perform remarkably well. Yet, there are no features that can be eliminated for all patients. In fact, our methods substantially outperform classifiers based on a small set of medically recommended risk factors. It is also true that there are patients with very similar features. This has led us to develop a new joint clustering and sparse classification method which produces strong predictions while discovering hidden patient clusters. Cluster membership of positive predictions can be used to justify them, which is particularly useful to clinicians.  

I will also briefly outline work with the Brigham and Women's Hospital on predicting the effect of hard to titrate ICU medications and developing an automated controller, using ideas from adaptive control methods, to optimally infuse such medications. 

BIO:

Yannis Paschalidis is a Professor and Distinguished Faculty Fellow of Electrical and Computer Engineering, Systems Engineering, and Biomedical Engineering at Boston University. He is the Director of the Center for Information and Systems Engineering (CISE). He obtained a Diploma (1991) from the National Technical University of Athens, Greece, and an M.S. (1993) and a Ph.D. (1996) from the Massachusetts Institute of Technology (MIT), all in Electrical Engineering and Computer Science. He has been at Boston University since 1996. His current research interests lie in the fields of systems and control, networks, applied probability, optimization, operations research, computational biology, medical informatics, and bioinformatics. 

Prof. Paschalidis' work on communication and sensor networks has been recognized with a CAREER award (2000) from the National Science Foundation, the second prize in the 1997 George E. Nicholson paper competition by INFORMS, and the best student paper award at the 9th Intl. Symposium of Modeling and Optimization in Mobile, Ad Hoc, and Wireless Networks (WiOpt 2011) won by one of his Ph.D. students for a joint paper. His work on protein docking (with his collaborators) has been recognized for best performance in modeling selected protein-protein complexes against 64 other predictor groups (2009 Protein Interaction Evaluation Meeting). He was an invited participant at the 2002 Frontiers of Engineering Symposium organized by the National Academy of Engineering, and at the 2014 National Academies Keck Futures Initiative (NAFKI) Conference.  Prof. Paschalidis is a Fellow of the IEEE and the Editor-in-Chief of the IEEE Transactions on Control of Network Systems.