Determining Dengue Hemorrhagic Fever Severity Levels Using K-Means Clustering: A Data-Driven Approach for Clinical Risk Stratification

Abstract

Background: Dengue Hemorrhagic Fever (DHF) remains a persistent public health burden in Indonesia, with 12,994 cases reported in Central Java Province in 2023. At a regional general hospital in Surakarta, DHF inpatient cases surged from 48 in 2023 to 610 in 2025, underscoring the urgent need for systematic severity classification. Objective: this study aimed to determine DHF severity levels using K-Means Clustering as a data-driven approach for clinical risk stratification. Methods: A quantitative descriptive design with a retrospective approach was applied to all 610 DHF inpatient cases, yielding 592 valid records after data cleaning. Five clinical variables were analyzed: age, platelet count, hematocrit, hemoglobin, and leukocyte count. Results: Data processing followed the CRISP-DM framework in RapidMiner Studio with K=3. The Davies-Bouldin Index (DBI) value of 0.745 confirmed good clustering quality. Three clusters were identified: Cluster 0 (Dengue Without Warning Signs, n=251, 42.40%), Cluster 1 (Dengue With Warning Signs, n=72, 12.16%), and Cluster 2 (Severe Dengue, n=269, 45.44%). Conclusion: These findings demonstrate that K-Means Clustering effectively supports clinical risk stratification of DHF severity levels, offering a reliable data-driven strategy for timely clinical decision-making in tropical endemic settings