Our MELD-Plus study was covered by Hospital Review, Health Data Management, MD&DI, and Massachusetts General Hospital. Our findings in another liver-related study were externally validated by researchers from Cleveland Clinic (see both studies published in The American Journal of Gastroenterology).
Planning to apply complex algorithms known as "best" or "used by millions" to extract features from narrative notes? Before you do that, consider trying Text Nailing (read more in Communications of the ACM and in Wikipedia). You may find that both your micro-average and macro-average F-scores get closer to 1.0.
Interested in working with EMRs but don't have access to any? Consider practicing your algorithms using EMRBots. These artificially generated medical records have helped other researchers to develop a new type of neural network (published in KDD) as well as to develop a new cloud-based platform (published in IEEE HealthCom). EMRBots were criticized in a JAMIA paper, but were mentioned more positively within the context of an extension to Turing Test applied in medical informatics (see: "A leap from artificial to intelligence" / Communications of the ACM, Jan. 2018).
I joined IBM Research in Cambridge, Massachusetts, in 2016. Before that I was a research fellow at Harvard Medical School/Massachusetts General Hospital (read how). I obtained my PhD focused on human-robot collaboration from Ben-Gurion University of the Negev, Israel. I hold dual citizenship (American and Israeli).