Background This paper presents a novel free-text information retrieval system designed especifically for medical case finding applications. Methods The proposed system begins by extracting medical information from free-text narrative reports and storing it in a predefined relational clinical data mart. The extraction is performed using a sophisticated medical thesaurus with a fast regular expression based method for pattern match. Following the extraction phase, inclusion/exclusion criteria are provided to the system using a physician-friendly user interface. The system converts the entered criteria into a single SQL command, which can be then executed on the relational data mart. Results Using inclusion and exclusion criteria of a published diabetes clinical trial (Diabetes Care 25:30–34, 2002) on a 500-discharge summaries database of a large hospital in NY, the performance of the system has been compared with the results obtained through a manual search by two physicians. (see Table) Conclusions The initial comparison validates that the proposed system can be successfully used for an ultra rapid case-finding applications, such as identifying appropriate patients to be enrolled in clinical trials with a high sensitivity, while paying only a small price for reduced specificity. Clinical Pharmacology & Therapeutics (2005) 77, P13–P13; doi: 10.1016/j.clpt.2004.11.054 Sensitivity Specificity Time Needed MDs search 71% 100% 50 hours System search 98% 91% 15 seconds
Discussion(0)
No comments yet. Be the first to comment.