Automated Extraction and Encoding of Information in Armed Forces Health Records

This proposal employs an existing medical language processing and encoding system called MedLEE (Medical Language Extraction and Encoding System), and will adapt it to be fully compatible with the Armed Forced Health Longitudinal Technology Application. MedLEE is ideally suited to perform this task because it has comprehensive coverage of the clinical domain, has a proven ten year record of effective performance in diverse clinical areas and applications, and is already integrated and used operationally in a complex clinical environment in a major medical center. MedLEE can readily be adapted to different coding systems, is easy to extend, and has successfully undergone numerous independent evaluations. Notably, MedLEE has been used to provide clinical data for important tasks such as i) data mining to find correlations between medical conditions, ii) medical error detection to automatically find errors that should be reported, iii) surveillance and alerting to detect specified clinical conditions, and iv) clinical research to automatically find variables for research databases. The research and development effort of this proposal will be led by a team of well-known and highly respected international leaders in the fields of Medical Informatics, Natural Language Processing, Data Mining, and Evaluation.