DDC-Outlier: Preventing medication errors using unsupervised learning
Author: Henrique D. P. dos Santos, Ana Helena D. P. S. Ulbrich, Vinicius Woloszyn, and Renata Vieira
Abstract: Electronic Health Records (EHR) have brought valuable improvements to hospital practices by integrating patient information. In fact, the understanding of this data can prevent mistakes that may put patients' lives at risk. Nonetheless, to the best of our knowledge, there are no previous studies addressing the automatic detection of outlier prescriptions, regarding dosage and frequency. In this paper, we propose an unsupervised method, called Density-Distance-Centrality (DDC), to detect potential outlier prescriptions. A dataset with 563 thousand prescribed medications was used to assess our proposed approach against different state-of-the-art techniques for outlier detection. In the experiments, our approach achieves better results in the task of overdose and underdose detection in medical prescriptions, compared to other methods applied to this problem. Additionally, most of the false positive instances detected by our algorithm were potential prescriptions errors.
Keywords: Electronic Health Records, Prescription Errors, Unsupervised Learning
Complete Reference: Henrique D. P. dos Santos, Ana Helena D. P. S. Ulbrich, Vinicius Woloszyn, and Renata Vieira. 2018. DDC-Outlier: Preventing medication errors using unsupervised learning. IEEE Journal of Biomedical and Health Informatics, 2018, 8 pages. DOI: 10.1109/JBHI.2018.2828028
This project belongs to GIAS at PUCRS, Brazil