Integrating Medical Knowledge and Clinical Data Banks: Discovery and Representation of Causal Relationships from a Large Time-Oriented Clinical Database: The RX Project

摘要

In the next few years computerized databases will become ubiquitous in hospitals and clinics. Medical data is needed not only by practitioners but by hospital administrators, health insurers, regulatory agencies, and the pharmaceutical industry. If appropriately analyzed, the data will be a resource of tremendous potential value for epidemiologic research, for post-marketing drug surveillance, for exploratory analysis of costs and benefits of new tests and therapy programs, and for assessment of long-term outcome in chronic diseases. The objective of the RX Project is to develop methods for (1) increasing the validity of medical knowledge derived from large time-oriented databases containing routine, non-randomized clinical data, (2) providing knowledgeable assistance to a research investigator in studying medical hypotheses on large databases, (3) fully automating the process of hypothesis generation and exploratory analysis. The RX computer program consists of a knowledge base (KB), a Discovery Module, a Study Module, and a clinical database. Utilizing techniques from the field of artificial intelligence, the KB contains medical and statistical knowledge hierarchically organized, and is used to assist in the discovery and study of new hypotheses. Confirmed results from the database are automatically encoded into the KB. The Discovery Module uses lagged, non-parametric correlations to generate causal hypotheses which are then studied in detail by the Study Module which automatically determines confounding variables and creates and tests an appropriate statistical model.