WHAT IS KNOWN AND OBJECTIVE: Drug-related problems (DRPs) are of serious concern worldwide, particularly for the elderly who often take many medications simultaneously. Medication reviews have been demonstrated to improve medication usage, leading to reductions in DRPs and potential savings in healthcare costs. However, medication reviews are not always of a consistently high standard, and there is often room for improvement in the quality of their findings. Our aim was to produce computerized intelligent decision support software that can improve the consistency and quality of medication review reports, by helping to ensure that DRPs relevant to a patient are overlooked less frequently. A system that largely achieved this goal was previously published, but refinements have been made. This paper examines the results of both the earlier and newer systems. METHODS: Two prototype multiple-classification ripple-down rules medication review systems were built, the second being a refinement of the first. Each of the systems was trained incrementally using a human medication review expert. The resultant knowledge bases were analysed and compared, showing factors such as accuracy, time taken to train, and potential errors avoided. RESULTS AND DISCUSSION: The two systems performed well, achieving accuracies of approximately 80% and 90%, after being trained on only a small number of cases (126 and 244 cases, respectively). Through analysis of the available data, it was estimated that without the system intervening, the expert training the first prototype would have missed approximately 36% of potentially relevant DRPs, and the second 43%. However, the system appeared to prevent the majority of these potential expert errors by correctly identifying the DRPs for them, leaving only an estimated 8% error rate for the first expert and 4% for the second. WHAT IS NEW AND CONCLUSION: These intelligent decision support systems have shown a clear potential to substantially improve the quality and consistency of medication reviews, which should in turn translate into improved medication usage if they were implemented into routine use.
WHAT IS KNOWN AND OBJECTIVE: Drug-related problems (DRPs) are of serious concern worldwide, particularly for the elderly who often take many medications simultaneously. Medication reviews have been demonstrated to improve medication usage, leading to reductions in DRPs and potential savings in healthcare costs. However, medication reviews are not always of a consistently high standard, and there is often room for improvement in the quality of their findings. Our aim was to produce computerized intelligent decision support software that can improve the consistency and quality of medication review reports, by helping to ensure that DRPs relevant to a patient are overlooked less frequently. A system that largely achieved this goal was previously published, but refinements have been made. This paper examines the results of both the earlier and newer systems. METHODS: Two prototype multiple-classification ripple-down rules medication review systems were built, the second being a refinement of the first. Each of the systems was trained incrementally using a human medication review expert. The resultant knowledge bases were analysed and compared, showing factors such as accuracy, time taken to train, and potential errors avoided. RESULTS AND DISCUSSION: The two systems performed well, achieving accuracies of approximately 80% and 90%, after being trained on only a small number of cases (126 and 244 cases, respectively). Through analysis of the available data, it was estimated that without the system intervening, the expert training the first prototype would have missed approximately 36% of potentially relevant DRPs, and the second 43%. However, the system appeared to prevent the majority of these potential expert errors by correctly identifying the DRPs for them, leaving only an estimated 8% error rate for the first expert and 4% for the second. WHAT IS NEW AND CONCLUSION: These intelligent decision support systems have shown a clear potential to substantially improve the quality and consistency of medication reviews, which should in turn translate into improved medication usage if they were implemented into routine use.
Authors: Margie E Snyder; Omolola A Adeoye-Olatunde; Stephanie A Gernant; Julie DiIulio; Heather A Jaynes; William R Doucette; Alissa L Russ-Jara Journal: Res Social Adm Pharm Date: 2020-11-04
Authors: Lipika Samal; Helen N Fu; Djibril S Camara; Jing Wang; Arlene S Bierman; David A Dorr Journal: Health Serv Res Date: 2021-10-05 Impact factor: 3.734
Authors: Sally H Preissner; Paolo Marchetti; Maurizio Simmaco; Björn O Gohlke; Andreas Eckert; Saskia Preissner; Robert Preissner Journal: Eur J Clin Pharmacol Date: 2021-12-27 Impact factor: 2.953