The repeated measurements of heart rate variability (HRV) is more relevant than a single HRV measurement in predicting patient prognosis but is less addressed previously. This prospective study aimed to investigate the association between repeated measurements of HRV and long-term mortality in chronic hemodialysis patients. The 164 patients (65.0 ± 13.1 years; woman, 57.3%) were enrolled from June 1, 2010, to August 31, 2010, and received four HRV measurements (before and during the index hemodialysis session) after the enrollment. The baseline characteristic and clinical variables, including mortality, were documented. The joint modeling method and Cox regression were used for statistical analyses. After an 8-year follow-up, 79 patients expired, and 85 patients survived. We found that higher normalized high-frequency (nHF) (hazard ratio [HR] 1.033) as well as lower very-low-frequency (HR 0.990), Variance (HR 0.991), normalized low-frequency (HR 0.999, P = 0.006), and low-frequency/high-frequency ratio (HR 0.796) were independent predictors for cardiovascular mortality. Whereas the independent predictors for infection-associated mortality included higher nHF (HR 1.033) as well as higher age (HR 19.29) and lower serum albumin (HR 0.01, P = 0.001). (all P < 0.001 unless otherwise stated) In conclusion, HRV measurement predicts long-term mortality among hemodialysis patients.