<![CDATA[Pending claims remain a challenge in the implementation of the INA-CBGs-based payment system within Indonesia’s National Health Insurance (JKN) program, resulting in administrative and financial burdens for hospitals. This study aimed to analyze the patterns of pending BPJS claims in cardiac arrest cases using K-Modes clustering and Random Forest classification. A quantitative analytical study with a cross-sectional design was conducted using 177 cardiac arrest medical records submitted for BPJS claims in 2024. The study population consisted of all cardiac arrest claim records, and samples were selected using simple random sampling. Inclusion criteria included complete medical records and finalized claim verification outcomes. Data were collected using a structured data extraction form from medical records and claim documents. K-Modes clustering was applied to identify claim patterns, while Random Forest was used to classify claim status and determine the relative importance of predictor variables. Model performance was evaluated using classification accuracy and feature importance measures. The analysis identified distinct claim clusters characterized by differences in diagnosis, medical interventions, and claim outcomes. Random Forest achieved an accuracy of 89.93% in classifying claim status. Cluster membership, final diagnosis, and medical interventions were identified as the most influential variables associated with pending claims. These findings demonstrate that combining K-Modes clustering and Random Forest classification can effectively identify patterns of pending BPJS claims in cardiac arrest cases and support hospital management by improving claim submission strategies and service quality. Future studies are recommended to use larger multicenter datasets and incorporate additional clinical and administrative variables to enhance model performance and generalizability.]]>
