Automatically building a repository to support evidence-based practice. .
Demner-Fushman D, Karpinski J, Thoma GR
Proceedings of the 2nd Workshop on Building and evaluating resources for biomedical text mining (BioTxtM 2010), 7th Language Resources and Evaluation Conference (LREC 2010). Valetta, Malta. May 2010.
Abstract:
Our long-term goal is to find, store, update, and provide access to key facts needed to support clinical decision making. Presently, the facts are extracted automatically from clinical narrative and biomedical literature sources, primarily MEDLINE, and stored in the Repository for Informed Decision Making. We envision expert community validation of the extracted facts and peer-reviewed direct deposit of key facts in the future. The key facts reported in publications do not change and can be extracted in advance and retrieved as needed. We chose an alternative approach to building the repository: extraction of key facts for specific clinical tasks and clinical scenarios. Combined with providing extracted facts (linked to the original publications) as service at the point of care, this approach allows capturing clinicians’ relevance judgments and leads to gradual, weakly-supervised construction of a collection of facts and documents pertaining to specific clinical situations and expert judgments of relevance and quality of the documents. In this paper, we demonstrate our approach to corpus construction using the clinical task of patient care plan development. We provide an overview of the process and focus on the automatic construction of PubMed queries – an essential step in finding documents containing key facts.
Demner-Fushman D, Karpinski J, Thoma GR. Automatically building a repository to support evidence-based practice. .
Proceedings of the 2nd Workshop on Building and evaluating resources for biomedical text mining (BioTxtM 2010), 7th Language Resources and Evaluation Conference (LREC 2010). Valetta, Malta. May 2010.