Abstract: Objective Results describing the relationship between hemoglobin variability (Hb-Var) of patient-level and mortality were ambiguous in hemodialysis (HD) patients. However, clinical anemia management pattern has important effects on patients’hemoglobin condition as well as on Hb-Var of patient- level
(PL Hb-Var) and Hb-Var of facility-level (FL Hb-Var). This study aimed to analyze the effect of PL Hb-Var and FL Hb-Var on mortality in HD patients. Methods This retrospective cohort analysis used the data in Beijing Hemodialysis Quality Control and Improvement Center (BJHDQCIC) database. Patients on stable HD
(＞3 month) before Jan. 1st, 2010 in this database were included. Patients should survive the baseline phase (Jan.-Jun., 2010) and had at least 3 Hb measurements during the baseline phase. Study subjects were followed up until Dec. 31st, 2011. Dialysis facilities with 10 or more eligible patients were included in this study. For every patient, demographics, primary cause of end stage renal disease, date of first HD, date of censoring and reason for censoring, date of death, cause of death, and 3 Hb measurements were extracted. The primary outcome was all-cause mortality. Patients were divided into 3 groups according to the FL Hb-Var. Two Cox regression models were established to analyze the relationship between mortality and PL Hb-Var and FL Hb-Var. The possible effective factors were also analyzed. Results This study included 2,013 subjects from 91 facilities. There were 276 death and 66 patients censored during follow-up period. Compared with the lowest PL Hb-Var group (＜4.4 g/L), mortality in the highest PL Hb-Var group (＞8.8 g/L) increased by 47.5% (HR=
1.475, 95% CI 1.105～1.968, P=0.008). Compared with the patients ＜40 years old, the crude hazard ratio of death for patients＞65 years old was 6.764 (95% CI 2.983～15.339, P＜0.001). The hazard ratio of death for males was 1.280 (95% CI 1.007～1.627, P=0.044). Outcomes for patients with diabetes (HR=1.000) and hypertension (HR=0.700, 95% CI 0.458～1.070, P=0.100) as the primary ESRD causes were relatively worse. While outcome for patients with glomerulonephritis was better (HR=0.521, 0.458～1.070, P=0.004). After adjustments, the mortality risk increased by 42.2% in the highest FL Hb-Var group compared with the lowest group (HR=1.422, 95% CI 1.056～1.914, P=0.020). Conclusions Moderate fluctuation of PL Hb-Var did not affect mortality. However, greater fluctuation (＞8.8g/L) increased death risk. Patients with age ＞65 years old, male gender, and diabetes and hypertension as the primary ESRD causes had higher mortality risk. Patients in facilities in the highest FL Hb-Var group had higher mortality risk. Effort aiming to improve anemia management pattern in facilities will reduce the mortality.

Key words: hemoglobin variability, hemodialysis, mortality