Adverse medical event reduction technologies: Identification of primary risk factors of adverse medical events using artificial neural networks

This thesis presents research into the area of adverse medical events and their resultant medical errors. Key terminology associated with adverse medical events is clarified and five specific categories of adverse medical events are identified. Information technologies that can be very beneficial in the reduction of adverse medical events are discussed and their importance in the healthcare delivery field is emphasised.; Utilising the United States Food and Drug Association's Adverse Event Reporting System database from the first two quarters of 2004 and artificial neural networks, a neural network with optimal performance was determined and the resultant weights were used to calculate the relative importance of the inputs into the ANN. The optimal ANN was a three-layer feed-forward back-propagation structure using the hyperbolic tangent transfer function.; An expanded model was prepared to capture a larger set of input variables that may be more highly linked to the occurrence of adverse medication events. The recommended model contains seventeen additional input parameters. A recommended model was also prepared for similar databases in the other areas of adverse medical events in the healthcare field based on the adverse medical event risk factors identified in this work. The FDA AERS is a mandatory reporting system. Recommendation from this work is to utilise this expanded model to establish a mandatory adverse medication event reporting system in Canada.