Objectives: Early mortality after starting hemodialysis (HD) is a barrier to improving elderly HD patients' longevity. We aimed to develop a clinical risk prediction model for early mortality in Korean elderly HD patients. Methods: In the retrospective cohort from the Korean Society of Geriatric Nephrology, a total of 2,468 HD patients were enrolled. The mortality risk model for six-month after dialysis initiation was validated using logistic regression for classification. In prediction models, the cohort was randomly divided into a training and test set with a 7:3 ratio, and loge value of Cox proportional hazard ratio (HR) was used for final risk score modeling. Results: We finally analyzed at the records of 1,967 incident HD patients over the age of 70, using 16 different variables. Seven parameters were significantly related to mortality: age, body mass index, malignancy, metastatic malignancy, hypertension, maintenance vascular access and serum albumin level. The crude 6-month mortality rate (n=309) was 15.7%. The average age at dialysis initiation was 77.4 years, and 44% of the patients was male. Patients in the Early-Peak mortality group were older, had a higher proportion of diabetes mellitus, had a lower BMI, and albumin levels than those in the Alive group. After adjusting for confounders, younger age at dialysis initiation, maintenance vascular access such as AVF or AVG and higher serum albumin level were significantly associated with a lower risk of all-cause mortality in multivariate Cox regression analysis. Finally, for HR prediction, survival modeling presented an area under the curve value of 0.771 for predicting the six-month mortality in the final risk score modeling. Conclusions: We developed an available model to predict early mortality after dialysis initiation in Korean elderly patients. Therefore, this would be useful as a clinical decision tool for dialysis in Korean elderly ESRD patients.