SOFTWARE ARCHITECTURE FOR GUIDELINE DIRECTED THERAPY

我们建议开发一种新的计算机体系结构来支持

We propose to develop a novel computer architecture that will support automated reasoning about protocol-based care. We will build on our previous experience in developing the ONCOCIN and T-HELPER systems to create a highly integrated set of computer programs that will significantly ease the construction of new automated systems that can use knowledge of clinical protocols and guidelines to interpret electronically stored patient data and to monitor patient management. Our approach will allow a wide range of clinical information systems to determine when clinical practice guidelines and protocols may apply to given patients, to recommend therapy in accordance with those guidelines, and to critique patient care that deviates from recommended practice patterns. The goal of our research is to address fundamental deficiencies in existing automated approaches to guideline-directed therapy by developing both new data models that can capture the necessary temporal distinctions in patient data and new problem-solving methods for reasoning about guideline- directed therapy. Our research plan consists of a four-step process that defines our specific aims. (1) We will create a domain model of medical concepts that is sufficient to represent the elements of clinical medicine required for a subset of protocol-based care in three areas of medicine: oncology, HIV infection, and hypertension. (2) We will create a temporal data management system that will derive automatically time-related abstractions from primary clinical data. This system will support a new database query language that will allow software clients to query for the presence of time-related abstractions in clinical data, in addition to the primary data themselves. (3) We will develop several extensible problem-solving methods for protocol-based care that will be reusable within the three clinical domains that we will explore. These methods will automate the tasks of therapy planning, of determining patient eligibility for protocols and guidelines, and of critiquing provider's management of patients when practice guidelines apply. (4) We will evaluate the applicability of our problem-solving architecture to different classes of clinical protocols. We will reintegrate our architecture into the T-HELPER medical workstation, and will test the architecture's ability to support reasoning about guidelines and protocols in the three clinical application areas that we will explore. At the conclusion of our work, we will have developed a validated software architecture that can support temporal-data management and automated reasoning about complex clinical protocols and guidelines. Because of our commitment to industry standards such as UNIX, C, TCP/IP, and SQL, our system will be usable as an open "server" for temporal-data management within a variety of application programs.