MIMIC-II: a database of 30,000+ patients for ICU research

Thursday, October 11, 2012 — Poster Session III
10:00 a.m. – Noon	Natcher Conference Center, Building 45	NLM	EPID-1

Authors

• S. Abhyankar
• D. Demner-Fushman
• F. Callaghan
• K. Leishear
• C.J. McDonald

Abstract

Secondary use of electronic health record data for clinical research is becoming increasingly important. The Multiparameter Intelligent Monitoring in Intensive Care II (MIMIC-II) database (maintained by MIT, sponsored by NIBIB) contains de-identified data on over 32,000 adult and neonatal ICU patients. Clinical data include vital signs, ventilator settings, lab results, radiology reports, medications, and documents such as progress notes and discharge summaries. Administrative data include demographic and billing information. We are collaborating with MIT to standardize and improve the quality of the data; we are also using the database to answer current questions in critical care and to develop natural language processing (NLP) techniques. To date, we have confirmed the obesity paradox as related to ICU mortality and found that vitamin B12 is not a predictor of ICU mortality once we account for liver function. We are currently studying the relationship between blood transfusions and necrotizing enterocolitis in neonates and are using NLP to extract data from the free-text notes that traditionally have been difficult to access efficiently, such as smoking status and patient discharge information. MIMIC-II is an important critical care database, and we welcome suggestions from the NIH community on new study ideas and potential collaborations.

back to top