Efforts to apply computer methods to assess and improve the quality of care in the hospital have been stymied by limited access to clinical data. Free-text data have detailed clinical descriptions of patients that would be useful in computer altering systems and computer reminder systems. However, free-text data cannot be interpreted by most clinical computer systems. In this proposal, we describe research specifically aimed at making free-text data accessible to computer-based applications for assessing and improving the quality of care. In particular the research plan focuses on the development of technologies that would allow free-text data to be used in clinical alert systems for critical test results; in reminder systems to encourage adherence to practice guidelines; and in data collection systems for severity of illness models applied in the assessment of risk adjusted outcomes. The approach described in the research plan emphasizes the development of statistical and probabilistic methods for interpretation of data derived from medical language processing systems. We will test the methods developed for language processing and interpretation developed under this proposal in 3 area: 1) the identification of concepts related to severity from the MedisGroups and the Computerized Severity Index models of patient severity of illness; 2) the identification of chest x ray reports and mammography reports with potentially malignant findings that require radiological follow-up; 3) and the automatic assessment of appropriateness of coronary artery bypass grafting (CABG) surgery from free-text descriptions of patients based on the application of a clinical practice guideline for CABG surgery.
Showing the most recent 10 out of 15 publications