Introduction. Rapid Emergency Medicine Score (REMS) has been known as a useful instrument to stratify patients in emergency department (ED) based on its ability to predict short term mortality. As the clinical characteristics of elderly patients visiting ED are frequently aypical, validation of REMS in this specific population is very important. To measure predictive validity of REMS in elderly patients visiting ED. Methods. Prospective cohort study in elderly visiting ED of Cipto Mangunkusumo between September to October 2016. Subjects were followed for 30 days after visiting ED. Calibration (Hosmer Lemeshow test), discrimination (Area Under ROC Curve) and REMS predictive value of 30 days mortality were measured. Result. Of 321 elderly subjects visiting ED during study period, 306 subjects were included in the study. Among them, 163 (53,2%) were men. 30 days mortality incidence was 22,8% (95% CI 22,3-23,3). Median age was 68,3 (6,7) year. As many as 83 (27%) subjects were in red triage. There were 154 (50%) patients with REMS 0-7, 140 (46%) with REMS 8-17 and 12(4%) with REMS more than 17. Calibration plot shows r=0,23 with poor calibration (p=0,00) in Hosmer Lemeshow test. AUC of REMS in predicting 30 days mortality in elderly was 0,45 (95% CI 0,23-0,66). Conclusion. REMS shows poor calibration and discrimination in eldery patients visiting ED.