Detection and Visualization of Variants in Typical Medical Treatment Sequences

Abstract

Electronic Medical Records (EMRs) are widely used in many large hospitals. EMRs can reduce the cost of managing medical histories, and can also improve medical processes by the secondary use of these records. Medical workers including doctors, nurses, and technicians generally use clinical pathways as their guidelines for typical sequences of medical treatments. The medical workers traditionally generate the clinical pathways themselves based on their experiences. It is helpful for the medical workers to verify the correctness of existing clinical pathways or modify them by comparing the frequent sequential patterns in medical orders computationally extracted from EMR logs. Thinking that the EMR is a database and a typical clinical pathway is a frequent sequential pattern in the database in our previous work, we proposed a method to extract typical clinical pathways as frequent sequential patterns with treatment time information from EMR logs. These patterns tend to contain variants that are influential in verification and modification. In this paper, we propose an approach for detecting the variants in frequent sequential patterns of medical orders while considering time information. Since it is important to provide visual views of these variants so the results can be used effectively by the medical workers, we also develop an interactive graphical interface system for visualizing the results of variants in clinical pathways. The results of applying the approach to actual EMR logs in an university hospital are reported.