META-GLARE: A Computer-Interpretable Guideline System Shell

Abstract

Computer-interpretable Guideline (CIG) systems are important tools for ensuring healthcare practice quality and standardization. They usually provide a tool to acquire CIGs, and one to execute them on specific patients. Current CIG systems rely on their own formalism to represent clinical guidelines, so moving to new phenomena/domains may require substantial extensions. We propose an innovative approach, providing a “shell” that facilitates system designers to define new CIG systems (or to update an existing one) through the definition of a new CIG representation formalism, based on the Task-Network model. We based it on our previous work on META-GLARE, and we extend it with a general execution tool, able to operate on any CIG representation formalism acquired through the META-GLARE acquisition tool. Developed with modularity and compositionality principles, the tool exploits an open library of basic execution methods. It offers a general execution mechanism supporting various CIG formalisms. We successfully applied our approach to three practical case studies. We have identified a reference CIG formalism (the one currently supported by the META-GLARE library) and compared its expressiveness to benchmark approaches. META-GLARE constitutes the first shell in the literature to facilitate the (formalism-based) design and development of CIG systems, considering both acquisition and execution.