Representation and Use of Temporal Information in ONCOCIN
Michael G. Kahn, Jay C. Ferguson, Edward Hance Shortliffe, Lawrence M. Fagan · PubMed Central · 1985
The past medical history of a patient is a complex collection of events yet the understanding of these past events is critical for effective medical diagnostic and therapeutic decisions. Although computers can store vast quantities of patient data, diagnostic and therapeutic computer programs have had difficulty in accessing and analyzing collections of patient information that is clinically pertinent to a specific decision facing a particular patient at a given moment in his disease. Without some model of the patient's past, the computer cannot fully interpret the meaning of the available patient data. We present some of the difficulties that were encountered in ONCOCIN, a cancer chemotherapy planning program. This program must be able to reason about the patient's past treatment history in order to generate a therapy plan that is responsive to the problems he or she may have encountered in the past. A design is presented that supports a more intuitive approach to capture and analyze important temporal relationships in a patient's computer record. In order to represent the time course of a patient, we have implemented a structure called the temporal network and a temporal syntax for data storage and retrieval. Using this system, ONCOCIN is able to quickly obtain data that is patient-specific and context-sensitive. Adding the temporal network to the ONCOCIN system has markedly improved the program's handling of complex temporal issues.