TagMeds is a system that recognizes and marks textual descriptions of a patient's current medications in the unstructured textual content of consultations letters. Medications are found based on their names and on linguistic patterns describing their dose, form of administration, etc. The UMLS is used as the underlying database of terms, and detected medications are encoded into XML tags consistent with and making use of the Health Level 7 (HL7) Clinical Document Architecture. The specific aims of this research are: (1) to review the literature in order to determine the state of the art in tagging free text for search and utilization, (2) to construct a tool that will reliably generate UMLS Concept Unique Identifier tags of current medications within free text. The methods involved are: (1) creating Perl procedures to recognize patterns in free text to retrieve the UMLS Concept Unique Identifiers and to insert these unique identifiers into XML tagging of the text and (2) statistical analysis of the use of TagMeds on a data base of consultation letters from the Endocrinology Clinic of the Children's Hospital of Boston as compared to manual markup by a group of physicians. The performance of an NLP system is found to be at least as sensitive as the performance of physicians in the extraction of current medications and their attributes. The tagged current medication information has the potential to support a personal electronic medical record system, such as PING. Additional development of TagMeds is likely to bring significant improvements, with modest expenditure of time and effort. TagMeds demonstrates that great utility can be achieved with a medical natural language processing system using simple and unsophisticated techniques.
展开▼
