User-Customizable Knowledge Coupling for Medical Decision-Support.

摘要

The clinical problem of data / alarm overload is well documented, but debate continues as to how it is addressed best. This thesis reviews clinical factors contributing to success and failure of medical decision-support systems (MDSSs) and it presents a prototype system, demonstrating several novel concepts to address clinical overload. Five basic clinical factors were derived: (a) graphical user interface; (b) user-customizable sentinel query design; (c) intelligent computing inference engine cueing from user-defined factors and focusable from general to specific patient viewpoints; (d) user-defined alert annunciation messaging, facilitating clinical knowledge transfer; and (e) quasi-flexible data sources for query building, including physiological parameters, various allied and focussed MDSS tool results, plus other data as clinical users require. Feasibility of the prototype "Watchdog" query system is demonstrated using the clinical problem of shock for data analysis. Practical suggestions for clinical use of these design concepts are provided for future work.