This study was aimed to develop a novel algorithm which incorporated medoid partitioning into multiple correspondence analysis (MCA) to determine the risk factors of rare disease events, which were easily ignored by patients or physicians and may result in more medical resources utilizations. This study used the Taiwan National Health Insurance Research Database (NHIRD) to investigate the onset of transient ischemic attack (TIA) among low risk atrial fibrillation (AF) patients younger than 60 years old, which is a rare disease event among AF patients and is easily ignored by AF patients and cardiologists.