Harvard Catalyst Profiles

MIT, Cambridge, MA	BS	2001	Computer Science and Engineering
MIT, Cambridge, MA	MEng	2004	Electrical Engineering and Computer Science
Indiana University, Indianapolis, IN	PhD	2012	Health Informatics
Regenstrief Institute, Indianapolis, IN	fellowship	2012	Medical Informatics

I believe that technology can and should have the same positive impact in the medical world it has in the consumer world, and that the nation's massive investment in Health Information Technology should translate into accelerated innovation and discovery and improved quality, efficiency, and safety of healthcare. I feel that computers are a paradigm shift in medicine, rather than an evolution of paper-based processes. To that end, I am interested in fundamentally new ways of using medical data that were not possible in paper records.

Knowledge discovery for clinical decision support: How can we learn from the mass of EMR data, and what can we learn from it to help clinicians perform their work more accurately and efficiently? In the world-wide-web, we believe the crowd. We trust them to help us select movies (Netflix), to shop (Amazon), and to filter our spam (Gmail). However, clinical decision support systems and health standards are generally developed by experts carefully studying small cohorts of patients. In healthcare, how much is the crowd (of highly trained physicians) to be trusted? How can we leverage data mining techniques developed over the last decade to extract crowd wisdom from the masses of data? I have developed and continue to refine methods to transform medical data into drafts of decision support content.

Sharing medical data to improve population health: I am currently engaged in two projects in this area. One, I am the lead researcher in enabling i2b2 for distributed cross-platform measurement of population health (the Query Health initiative). i2b2 is a clinical data analytics platform in use at over 80 sites nationwide. My contributions are currently showcased at two national pilots, with more coming. Two, I lead a project to improve patient safety by making IHI Trigger Tool methodologies for adverse event detection more accessible across institutions.

Revolutionizing user interfaces: Services like Google use a single interface to perform search, calculations, and many other functions. EMRs, on the other hand, are often rigid point-and-click interfaces prevalent before the web. Many clinicans find the amount of clicking and the awkward drop-down-menus to be very aggravating, which is unacceptable in a time-constrained environment. Additionally, EMRs need to display very complex historical and anatomical data in easy-to-digest ways. How can we summarize complex data in a rapidly digestable format? I have published work in: summarizing medical records through a problem-oriented view, visualizing the patient record as a timeline, filtering irrelevant problems from the problem list, searching the patient record, and automatically listening to and interpreting doctor-patient conversations through microphones in the office. Several of these projects are ongoing, such as producing a predictive problem-oriented view for dermatology and continued development of a substitutable technology platform to enable innovate medical apps to run on multiple systems.

Making PHRs viable: Personal Health Records (PHRs) have appeared over the last several years with the promise of helping people manage their health information more easily. Yet most existing PHRs are only slightly more useful than an unstructured collaborative document. Few have adopted PHRs, except for those with massive information-cataloging needs (e.g., caregivers of elderly adults). However, there may be a use-case for the PHR. Patients are aggravated they have no access to their health information. Physicians, also, express frustration at having limited access to patient-gathered information (e.g., blood glucose logs). Patient portals have been successful in providing secure communication and some medical information exchange between patients and single hospital systems. Services such as PatientsLikeMe and CureTogether have been successful in using collective patient information to build community and collaborative disease profiles. A useful PHR must integrate these features. Essentially, the PHR should be part of the social web, allowing us to share medical information (in a controlled way) in the same vein that we share our activities with Facebook friends. It is not a disconnected application, but a tool in our increasingly information-hungry society. Of course, this raises many issues of security, privacy, liability, and data ownership.