DISEASE PATTERN DISCOVERY USING K-MEANS CLUSTERING FOR DATA-DRIVEN HEALTHCARE DECISION SUPPORT

Abstract

The increasing availability of electronic healthcare records creates opportunities to discover disease patterns and support evidence-based clinical management. This study aims to identify disease patterns among patients enrolled in Indonesia's National Health Insurance (BPJS) program using K-Means clustering and examine their potential contribution to data-driven healthcare decision support at Bandung Mitra Insani Clinic. A quantitative data mining approach was employed following the Knowledge Discovery in Databases (KDD) framework, encompassing data selection, preprocessing, transformation, clustering, and evaluation. Patient age, gender, and disease diagnosis were analyzed using the K-Means algorithm implemented in Orange Data Mining. The clustering process identified two distinct patient groups with different demographic and clinical characteristics. The first cluster predominantly comprised younger patients diagnosed with common acute conditions, including dyspepsia, gastritis, respiratory infections, and migraine. Conversely, the second cluster primarily represented older patients with chronic conditions, particularly hypertension and diabetes mellitus, indicating potentially greater needs for continuous monitoring and long-term care. Cluster quality assessment using the Silhouette Coefficient yielded values ranging from 0.55 to 0.71, suggesting moderate to relatively strong cluster separation. Scatter plots and box plots further facilitated the interpretation of patient distributions and disease characteristics. These findings demonstrate the potential of unsupervised learning to transform routine healthcare records into actionable insights for identifying disease patterns, prioritizing healthcare services, anticipating resource requirements, and informing preventive interventions. The proposed approach provides an analytical foundation for data-driven healthcare decision support in primary care settings, although further validation is required to establish its generalizability and clinical utility