Tool-support is a crucial issue for guideline development and use. We are planning to build tools to support browsing, authoring, validation, downloading, and execution of GLIF-encoded guidelines. Following is a short description of the tools we are building already and/or planning to build.

1.1 Guideline authoring/viewing tools
At Stanford and Columbia, we have been using Protege as an authoring and viewing tool for GLIF3. The Stanford team completed two Protege “widgets” to custom-tailor Protege’s RDF support for GLIF. One widget is used to fine-tune the way guidelines, encoded in GLIF3 and created using Protege, are saved into RDF format. The other widget is used when RDF files containing GLIF-encoded guidelines are loaded into Protege. It automatically lays out the flowcharts diagrammatically on the screen. We configured the Protege-2000 GLIF3 authoring tool in two ways: Domain experts can use the first configuration for creating abstract flowcharts, while informaticians can use the other configuration to create detailed computable specifications. The first configuration allows a guideline author to specify clinical algorithms, codes for clinical terms, rules for ranking alternative treatment options written in natural language, and documentation attributes. The latter elements do not specify guideline logic but contain information that is important for assessing guideline validity. Examples include links to information describing supporting data, the target audience of the guideline, and the strength of evidence associated with each guideline step.

The research team at Harvard created an alternative guideline-authoring/viewing tool for graphically creating and editing guidelines in GLIF. The tool provides flowchart and tree views of guidelines. Steps specified in the flowchart can be encoded in more detail using editing forms. The encoded guidelines can be saved to disk as an RDF file or can be submitted to a guideline server for storage. The authoring tool has been designed so that it can be easily custom tailored for different types of guidelines and can be extended with more features such as specialized views of guidelines or links to external applications (e.g., vocabulary databases). A prototype guideline viewing application was developed. The application provides a list of guidelines available in a server. The user can select a guideline that is then automatically laid out as a flowchart. This viewer can be modified for use as a web-based applet.

1.2 Vocabulary tool
The Harvard team developed a tool for editing vocabularies. The tool provides graphical views of the vocabularies (as networks and hierarchies). Users can add, delete, and modify concepts, terms, and relationships. The tool also provides a read-only view of the UMLS meta-thesaurus. This tool will be linked with the guideline authoring-tool once the patient data model specifications are finalized.

1.3 Guideline storage and retrieval tool
We have implemented a simple guideline retrieval tool that enables guidelines to be retrieved by selection from a list, by their name. We also developed a set of primitives for server functionality. Additional work was done to develop a classification scheme for guidelines that would aid in retrieval.  Nonetheless, server development remains incomplete. Functionality intended but not yet implemented includes the ability to manage version control, including updates of guidelines as well as derivative guidelines, and also the ability to retrieve guidelines based on patient-specific eligibility criteria associated with the entire guideline or with particular entry points (Patient state steps).

1.4 Validation tool
The Stanford team used Protege-2000 to develop a validation tool for GLIF3. For each attribute in a class, Protege allows developers to define allowed data types, cardinality constraints, and lower and upper limits on numerical values. We used Protege’s constraint language to define structural integrity constraints (e.g., a branch step should not immediately be followed by a synchronization step). The constraints are included in Protege's GLIF ontology version 3.5. Using this ontology, Protege can be used as an authoring and vlidation tool. We used this tool to author guidelines and to check them for errors. We found errors in two of the four guidelines that the first author of this paper encoded in GLIF3. These errors include: (1) decision steps that were linked to fewer than 2 decision options, (2) synchronization steps that immediately followed branch steps, and (3) guideline steps that were not part of any algorithm. No errors were found in the headache guidelines. This may be due to experience gained by the guideline encoder.

1.5 Execution engine
The Columbia site developed a GLIF3 GuideLine Execution Engine (GLEE), which can be integrated into the clinical information system of a local institution. GLEE is able (1) to recommend clinical tasks that should be performed and assist in clinical decision making based on the scheduling of primitive execution tasks and the evaluation of decision criteria; (2) to support event-driven clinical decision making by integration of GLEE with a local clinical event monitor; (3) to provide a standard application interface to the EMR of the host institution that can be used for communications between GLEE and the local EMR (such as patient data retrieval, automatic enactment of clinical tasks, and clinical event monitoring; (4) to provide a standard application interface to the clinical applications of the host institution that can be used to present the user interface of GLEE for clinical decision support; (5) to provide a stand-alone graphical user interface for GLEE that can be used as a tool for the purpose of demonstration, medical education, and guideline development and testing; and (6) to provide a standard application interface to a guideline representation model (in particular, the GLIF3 model) that facilitates the maintenance of GLEE during the evolution of the guideline representation model.

Contact person: Mor Peleg, Stanford Medical Informatics, Stanford University