PROTOKOL: a system-building aid for developing Protocol-type knowledge base.

We have identified the difficulties encountered in developing and maintaining the knowledge base component of a decision-support system as the primary factors impeding system development and contributing to the degradation of subsequent system performance. Our analysis of these difficulties suggests that the direct involvement of medical experts in knowledge base development and maintenance is critical to system success. We have accordingly developed a system-building aid (PROTOKOL) designed to facilitate the direct involvement of nonprogrammer medical experts in the acquisition, verification, and maintenance of domain-specific knowledge bases. The major problem addressed in this work is the creation of a system-building aid to facilitate the construction of knowledge-based decision-support systems within HELP. The goal of this work is to provide significant computer-based solutions to the problems referenced above. PROTOKOL is designed not only to involve medical experts directly in the knowledge engineering process but to actively assist those experts in successfully achieving their knowledge engineering responsibilities. The nature of this research is best characterized as software engineering research. After expending significant amounts of effort to analyze the task of creating a high-level system-building aid, the proposed software system was programmed and tested. PROTOKOL has been evaluated in regards to performance objectives established during the analysis stage of our work. Our method of evaluation involved using PROTOKOL to develop a beta-test application based on screening protocols associated with the Cardiac Arrhythmia Suppression Trial (CAST), in collaboration with researchers from the Cardiology Division at LDS hospital. This application was tested clinically for two 4-week periods. The results from these performance evaluations support the conclusion that PROTOKOL may be used to adequately and accurately encode the knowledge requirements of treatment protocols. Further evaluations have demonstrated the generalizability of these performance capabilities to at least one other protocol-based application, i.e., the screening stage of th eLovastatin multicenter trial. PROTOKOL as a system-building aid holds great promise for application to other medical problems that may be formalized as clinical algorithms or protocols.