Development and evaluation of an authoring environment for a centralized knowledge repository.

A single knowledge authoring environment can accommodate the wide range of needs imposed by the authors and the knowledge engineering processes associated with a centralized knowledge repository. In this dissertation, three separate studies are presented to support this hypothesis.The studies explore various aspects of the knowledge authoring environment, (1) the software environment itself; (2) the reasons for acceptance or rejection of the authoring tools and (3) the importance of automated feedback processes integrated within the system. The first study focuses upon the requirements, design phase, and implementation of the Knowledge Authoring Tool (KAT). This research focuses upon the needs, features, and requirements in building such an application and details how closely the application meets original expectations. It details usage patterns within the first year of KAT's existence in terms of knowledge content produced, the number of clinical applications served, as well as the number of clinical domain experts currently interacting with the tool. The study defends the need for the authoring environment, as well as its distinctions from other authoring tools used in medical informatics. The second study analyzes cognitive measurements among end-users of the authoring environment to identify which aspects of the software are most important in determining overall 'acceptance' of the application. We approached the study by using validated Likert-based survey questions to test the constructs and relationships defined by the Technology Acceptance Model. Our findings show that the model predicts users' intentions to use the software reasonably well, though usefulness is regarded as more important than ease of use in determining acceptance of the authoring tool. The third study details the development and validation of an automated feedback tool built into the knowledge authoring environment. We assessed the validity of the suggestions proposed by the system by presenting proposed changes to order set templates to clinical domain experts. Of the four types of changes proposed by the system, three were acceptable above a 50°lo threshold among the clinical panel. In summary, we have created an authoring environment to support the distributed development of clinical knowledge base content within a centralized repository. The authoring environment is designed to ease some of the burdens associated with the knowledge base creation and maintenance processes.