Date

August 2009

Document Type

Capstone

Degree Name

M.B.I.

Department

Dept. of Medical Informatics and Clinical Epidemiology

Institution

Oregon Health & Science University

Abstract

Electronic health record (EHR) systems provide a means of tracking a broad range of patient health information including demographics, progress notes, problems, medications, vital signs, past medical history, immunizations, laboratory data and radiology reports[superscript 1]. To the author's best knowledge, there exists no universal standard in medication list formatting. As a result, one can imagine the ability for a clinician to quickly extract critical details regarding a patient's medication profile may be hindered. Indeed, a recent Healthcare Information and Management Systems Society (HIMMS) publication[superscript 2] reported that approximately 25 percent of medication errors in the 2006 Pharmacopeia MEDMARX involved computer technology as a contributing cause and cited several studies documenting instances of 'terminology confusion' as a significant source of issues. We therefore believe the availability of a means of categorizing clinical drugs should therefore serve to promote greater expediency as well as realization of best treatments in the delivery of patient care. Here, we assess the feasibility of utilizing several complimentary machine learning techniques to extract categorical information for eventual use in creating a comprehensive pharmaceutical drug/category ontology. We obtain a list of generic and proprietary drug names from the RxNorm database while using the web-based encyclopedia, Wikipedia, as our primary data set from which to extract semantic knowledge of the drugs. Support vector machine (SVM) algorithms are utilized on a pared-down, manually-curated test set in attempts to develop a robust classifier to distinguish drug class from non-drug class entries with the intent of identifying valid medication categories and subsequently using them to group drugs. We evaluate classifier performance and suggest additional approaches that may prove more effective.

Identifier

doi:10.6083/M4PK0D4N

School

School of Medicine

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.