The SAGE (Standards-Based Sharable Active Guideline Environment) project, a collaboration among research groups at IDX Systems Corporation, the University of Nebraska Medical Center, Intermountain Health Care (IHC), Apelon, Inc., Stanford Medical Informatics, and the Mayo Clinic, seeks to create the technological infrastructure for integrating interoperable computer-interpretable clinical practice guidelines (CPGs) into enterprise clinical information systems (CISs). (3HX)
The main SAGE website is http://www.sageproject.net/. We maintain a Stanford SAGE project website at http://www.smi.stanford.edu/projects/sage/. (3HY)
The SAGE project uses Protégé in two ways: (1) as the development environment for creating the ontologies required for formalizing CPGs in computer-interpretable format, and (2) as the workbench for encoding these computer-interpretable guidelines as Protégé knowledge bases. (3HZ)
For the use of Protégé as an ontology-development environment, we used Protégé to build a suite of models that include (1) a guideline ontology that defines the structure of a computable guideline, (2) an information model that represents the structure and constraints of patient information with which we will generate patient-specific recommendations for guideline-based care, (3) a terminology model that allows us to reference external standard terminologies and to define new concepts, (4) an organizational model that defines the events, clinical settings, and clinical and administrative roles of personnel in patient-care processes. (3I0)
For the use of Protégé as a guideline-encoding workbench, we have built a framework called Kwiz that extends Protégé to especially manage the complexity of the guideline-encoding process, and to improve the efficiency in knowledge acquisition.. We built Kwiz to take advantage of the strengths of Protégé, and incorporated features including what we found relevant from other tools that we had reviewed previously. The Kwiz framework extends Protégé in the following ways: (3I1)
1. Custom views on the knowledge base (3I2)
Kwiz can provide domain-specific views of the knowledge model concepts by creating tabbed entry points to components of the knowledge base. These views can be used to hide the complexities of the underlying computable model from the user. (3I3)
2. Alternative navigation during the encoding process (3I4)
Kwiz provides a navigation paradigm where the user is always working on a single window. A stack of pointers to forms that were previously opened aids the user in knowing where they are in the encoding process, and allow the user to jump back to any form in the stack. Alternatively, a user may navigate to instances that are referenced, directly and indirectly, from a top-level instance, and jumps to the Protégé form for that instance. (3I5)
3. Reusable knowledge repository (3I6)
Kwiz supports a knowledge repository that includes previously encoded guideline knowledge bases. The user can search the contents of the repository, and can reuse appropriate components, such as formal decision criteria, in another guideline knowledge base. (3I7)
4. Explanation, validation, and export (3I8)
Kwiz, together with other Protégé tab plug-ins, include explanation, export, and validation functionalities. At every Kwiz-enabled Protégé form, the user has access to annotations and appropriate knowledge-acquisition instructions. The facet-constraint and PAL-constraint tabs can be used for knowledge-base validation. We also implemented a document-generation tab that exports the content of a knowledge base as one HTML document. (3I9)
The software that implements Kwiz is being packaged for release to the Protégé community. The primary obstacle is the need to provide documentation for Kwiz and to write the instructions for configuring a knowledge base to work with Kwiz. (3IA)