The major concern of the RX Project is with the development of tools for the collection, analysis, and interpretation of health care data. Specifically, the objectives of the project are 1) to increase the validity and ease of data analysis and interpretation and 2) to facilitate the process of hypothesis generation and preliminary confirmation. The principal focus of our research will continue to be on the RX Study Module. Utilizing an on-line knowledge base of clinical medicine and of statistical procedures, the Study Module is designed to assist a researcher with the task of creating a comprehensive, mathematically correct study design using appropriate data from a large clinical database. The Study Module determines an overall study design choosing between cross-sectional and longitudinal methods, it determines significant confounding variables and methods for controlling them, it determines patient eligibility criteria, and it formulates a mathematical model for testing the hypothesis. Finally, the Study Module automatically runs the study design on a statistical package and assists with the interpretation of results. A second focus of our work is in methods for aiding hypothesis generation. We seek to facilitate this process by developing improved means for labeling and accessing patient records. We are also interested in developing methods for automating the dicovery of hypotheses by using methods from the field of artificial intelligence. Our method will use a medical knowledge base to focus the search for promising hypotheses. Our planned research for all these endeavors will build on our present algorithms and the current RX prototype. We will test the resulting system by the performance of specific rheumatologic studies using a 1,700 patient database provided to us by the American Rheumatism Association Medical Information System research group.
| de Zegher-Geets, I M; Freeman, A G; Walker, M G et al. (1988) Summarization and display of on-line medical records. MD Comput 5:38-45, 54 |
| Blum, R L (1986) Computer-assisted design of studies using routine clinical data. Analyzing the association of prednisone and cholesterol. Ann Intern Med 104:858-68 |
