Application of K-Medoids Clustering Method on Disease Clustering Based on Patient Medical Records

Fatika, Dian
Bustami, Bustami
Afrillia, Yesy

International Journal of Engineering, Science and Information Technology
Vol 5, No 1 (2025)

Dr. Fauziah Bireuen Regional General Hospital (RSUD) faces daily challenges in managing the ever-increasing medical record data. Currently, the medical record data only consists of reports containing information on the number of patients and their diseases, which are then archived without further processing to generate valuable information. This research aims to cluster diseases based on patient medical records using the K-Medoids Clustering method, thereby providing information on the patterns of disease spread in various regions of the Bireuen Regency. The data used are patient medical records from RSUD Dr. Fauziah Bireuen from 2021–2023, focusing on five common diseases: stroke, hypertension, schizophrenia, dyspepsia, and pneumonia. We conducted Clustering in 17 sub-districts in Bireuen Regency using the K-Medoids method and determined the optimal number of clusters using the Elbow method. The research results show that the K-Medoids method successfully grouped each disease into 3 clusters: high, medium, and low. The results showed that the K-Medoids method successfully grouped each disease into 3 clusters: high, medium, and low. The cluster distribution for stroke disease consists of 7 sub-districts in the high cluster, 7 in the medium, and 3 in the low. Hypertension disease consists of 6 sub-districts in the high cluster, 3 in the medium, and 8 in the low. Schizophrenia disease comprises seven sub-districts in the high cluster, 8 in the medium, and 2 in the low. Dyspepsia disease includes six sub-districts in the high cluster, 2 in the medium, and 9 in the low. Meanwhile, pneumonia disease consists of 8 sub-districts in the high cluster, 5 in the medium, and 4 in the low.