An Intelligent Decision Support System for Automated Medication Review

Knowledge Acquisition techniques are not historically designed for domains where the expert knowledge being modelled is inconsistent and poorly defined. This study explores the applicability of the MCRDR technique of knowledge acquisition towards a domain of this nature, medication reviews. Through this experimentation it is also sought to improve the quality of service that the medication reviewers can provide to their patients, by reducing the incidence of missed classifications. To facilitate this study a Knowledge-Based System was developed that employed the MCRDR method and which was capable of both being instructed in the domain of medication review through its routine use by the expert, and acting similarly to the expert when producing its classifications based on genuine medication review cases. It was found that there was a high incidence of missed classifications by the expert which were automatically repaired by the system, and it was also shown that the incidence of missed classifications reduced dramatically as the systems knowledge of the domain grew more complete. It was also found that the inconsistent nature of the expert's knowledge of the domain did not appear to significantly affect the functioning of the system, with none of the tests performed indicating any deviation from what would be expected in a normal MCRDR system.