• Surgical Care and Outcomes Assessment Program (SCOAP) - Data Element Extraction with Natural Language Processing: SCOAP is a voluntary performance surveillance, sharing and feedback platform derived from clinical records. SCOAP improves quality by increasing adoption of evidence-based process of care measures and performing real world Comparative Effectiveness Research (CER). SCOAP includes data collection over 150 CER and expert opinion informed processes of care and risk adjusted outcome domains from more than a dozen procedures. Currently, Quality Improvement (QI) personnel from participating SCOAP hospitals manually review electronic medical records to extract clinical data for SCOAP. The high cost of manual abstraction (nearly 50 minutes per case) and the need for advanced clinical knowledge to extract data presents a real challenge for hospitals with busy clinicians and limited staff time. The purpose of the project is to automate this process by using Natural Language Processing approaches.


• Critical Illness Phenotype ExtRaction (deCIPHER):Clinical and translational research involving critical illness phenotypes relies heavily on the identification of clinical syndromes defined by consensus definitions (e.g. pneumonia, sepsis, acute lung injury). The overall goal of this project is to apply natural language processing, machine learning, and network analysis to develop an automated screening tool that accurately identifies critical illness phenotypes and their interactions among ICU patients.


• Using Natural Language Processing to Automatically Detect Critical Recommendations in Radiology Reports: Communication of recommendations and abnormal test results is prone to error. If important imaging findings and recommendations are not systematically identified and promptly communicated to referrers, poor patient outcomes can result. In this project, we investigate text processing approaches to identify critical recommendations in radiology reports.

Past Projects

• LitLinker: The explosive growth in biomedical literature has made it difficult for researchers to keep up with the advancements, even in their own narrow specializations and to explore connections to their own work from other parts of the literature. LitLinker is a text mining system that incorporates knowledge based technologies, natural language processing techniques and data mining algorithms to mine the biomedical literature for new, potential causal links between biomedical terms.