Dynamic Patient Categorization Based on Medical Records Using Fuzzy C-Means Clustering Technique

Abstract

Health agencies have actively implemented electronic medical records (EMR) on paper-based medical records. Currently, EMR is not very informative for extracting useful information or for tracking the patient disease. Hidden patterns that can be removed through data mining can help practitioners understand discreet relationships, such as inpatient categorization. Categorizing patients in determining groups in hospitals is very helpful for medical personnel in understanding work and actions to provide the right decisions and fast in taking action. However, there are several challenges in categorizing patients using data mining, one of which is selecting methods for the right cluster results according to the data used. Although there have been many studies that discuss the categorization of patients, no one has addressed the categorization of dynamic patients, especially in medical records. We consider this an important issue because, in medical records, there are similarities between one data and another, which causes one patient data to fall into two categories and affects health practitioner decision making. These challenges can be overcome using dynamic patient categorization. After implementation, we do an accuracy test. The test is done using Silhouette, Sum Squared Error, and Duns Fuzziness Coefficients. The result is that the accuracy is close to 85.2%. Identifying the types of diseases that become cluster labels is a good future work to do.