291 Campus Drive
Stanford, CA 94305
USA
Natural language processing approaches to extracting patient functioning from clinical data
Natural language processing (NLP) has become a significant tool in clinical informatics research, leading to advances in electronic phenotyping, adverse drug event detection, and literature retrieval. However, capturing information about patient functioning, recently proposed to join mortality and morbidity as a world indicator of health, is largely unexplored. In this talk, I will present ongoing work in the NIH Clinical Center on developing NLP methods for extracting information on patient functioning from unstructured EHR data. Considering disability adjudication as a motivating example, we analyze the distinguishing characteristics of the language of functioning, and describe development of the first clinical corpus annotated for functioning information. A significant challenge to applying NLP methods to functioning is the lack of robust domain-specific knowledge sources; we present early work exploring unsupervised learning to adapt to this low-resource domain. I will conclude with a roadmap for ongoing research in NLP for functioning.