Description of Collaborative Activity: |
The objective of this initiative is for the National Library of Medicine (NLM) to assist the US Food and Drug Administration (FDA) in indexing the content of structured product labels (SPL) for drugs. Natural language processing technology is used to extract drug-drug interaction information from SPLs which will be validated by FDA subject matter experts. Transforming the narrative text to structured information encoded in national standard terminologies is a prerequisite to the effective deployment of drug safety information in drug labels for clinical decision support. Starting in FY16, the Labeling and Health Communication (LHC) and the FDA's Center for Drug Evaluation and Research (CDER) collaborate on advancing automated extraction of Adverse Reactions (ARs) from drug listing information companies have submitted to FDA in the form of Structured Product Labels (SPLs). The goals of the initial agreement were to develop optimal MetaMap settings for extracting ARs from SPLs and provide FDA with the software necessary to replicate the results. To evaluate the extraction quality, LHC and FDA needed to develop a test collection of SPLs annotated with ARs. Working together, LHC and FDA have developed annotation guidelines. LHC have annotated 200 SPLs and coded the results to MedDRA, the official controlled vocabulary used by FDA to normalize the labeled ARs. In FY18, the collaboration continues with the goals to conduct a community-wide evaluation of the available AR extraction tools. The evaluation venue is NIST Text Analysis Conference (TAC). Over 130 new labels have been annotated with DDIs in FY18. |
