An accurate medication history is an essential part of patient assessment and can have vital impact on the person's care. However, manually-acquired histories are prone to inaccuracies. The problem is especially serious in emergency care and in disaster situations due to the lack of time, overloaded staff and special patient conditions (e.g. comatose or confused patients, unaccompanied minors or elderly patients). This study is partially funded by the Bethesda Hospitals Emergency Preparedness Partnership. It focuses on patients attended by the Emergency Department (ED) of a regional hospital (Suburban Hospital, Bethesda, MD) and evaluates the potential added value of electronic prescription history information from Surescripts. We established a secure electronic connection between the hospital and Surescripts so that prescription-filling reports could be retrieved from Surescripts in real-time, based on four pieces of patient identifying information (name, date of birth, gender and zip code) obtained from the registration process. For three months we collected the Surescripts information in parallel with the medication history manually acquired by the ED nurse. We also retrieved demographic, administrative (e.g. class of insurance, mode of arrival) and clinical (e.g. vital signs, Glasgow coma score) information from the hospitals database as predictors. All the information was de-identified before being sent to NLM for analysis. This research only involved de-identified data collected for quality assurance purposes. We conducted the data analysis retrospectively;there was no intervention or patient contact. The ED-collected records of prescriptions were manually typed in by triage nurses, so the drug names were subject to variation and typos. Surescripts data provided standardized names. To make the two sources comparable, we mapped all drugs to their standard names in RxNorm, the U.S. standard reference drug terminology. Mapping was done largely by automatic text matching algorithms followed by manual review of the unmapped items. About two-thirds of all ED patients were registered in the Surescripts database, and for about half of all patients, Surescripts returned some medication history information. We have completed the analysis of this data, and in short, Surescripts provides more complete medication history than the manual history when the Surescripts database has any information about the patient. However, it has no information for about a third of the patients who have medications. The Surescripts information contains drugs that the patient is currently taking but also the full history of prescriptions being filled in the past year. ED providers found the summaries very useful in spotting problematic behaviors (e.g. narcotic drugs abuse, poor drug compliance). We have submitted a paper to the journal Annals of Emergency Medicine, which has been accepted and is pending publication. With the Surescripts prescription data we obtained in the first phase, we performed preliminary analyses of the number of different prescribers per patient as a measure of care continuity, and it seemed quite high. We have also examined the number of interruptions that physicians would experience due to drug interaction messages under different levels of interaction importance. While doing this analysis, we have acquired, a much larger de-identified database of prescriptions covering the entire Washington, D.C. metropolitan area. With this database we will seek to answer many questions about prescribing patterns, including, but not limited to: the degree to which drug interacting pairs written by different providers from different offices would not be seen as an interaction by the provider in either office;the drugs most frequently responsible for drug interactions; the frequency of provider reminder interruptions according to different commercial drug interaction knowledge bases at various levels of threshold settings, and;identification of events (e.g. ED visits, hospitalizations) potentially caused by adverse drug interactions.

Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
National Library of Medicine
Zip Code
Drawz, Paul E; Archdeacon, Patrick; McDonald, Clement J et al. (2015) CKD as a Model for Improving Chronic Disease Care through Electronic Health Records. Clin J Am Soc Nephrol 10:1488-99
Demner-Fushman, Dina; Kohli, Marc D; Rosenman, Marc B et al. (2015) Preparing a collection of radiology examinations for distribution and retrieval. J Am Med Inform Assoc :
Ayvaz, Serkan; Horn, John; Hassanzadeh, Oktie et al. (2015) Toward a complete dataset of drug-drug interaction information from publicly available sources. J Biomed Inform 55:206-17
Abhyankar, Swapna; Demner-Fushman, Dina; Callaghan, Fiona M et al. (2014) Combining structured and unstructured data to identify a cohort of ICU patients who received dialysis. J Am Med Inform Assoc 21:801-7
Pan, Xuequn; Cimino, James J (2014) Locating relevant patient information in electronic health record data using representations of clinical concepts and database structures. AMIA Annu Symp Proc 2014:969-75
Fung, Kin Wah; Kayaalp, Mehmet; Callaghan, Fiona et al. (2013) Comparison of electronic pharmacy prescription records with manually collected medication histories in an emergency department. Ann Emerg Med 62:205-11
Fung, Kin Wah; McDonald, Clement; Srinivasan, Suresh (2010) The UMLS-CORE project: a study of the problem list terminologies used in large healthcare institutions. J Am Med Inform Assoc 17:675-80
Demner-Fushman, Dina; Chapman, Wendy W; McDonald, Clement J (2009) What can natural language processing do for clinical decision support? J Biomed Inform 42:760-72
McDonald, Clement (2009) Protecting patients in health information exchange: a defense of the HIPAA privacy rule. Health Aff (Millwood) 28:447-9
Friedlin, F Jeff; McDonald, Clement J (2008) A software tool for removing patient identifying information from clinical documents. J Am Med Inform Assoc 15:601-10

Showing the most recent 10 out of 14 publications