Predictive Factors Upon Discontinuation of Renal Replacement Therapy for Long‐Term Chronic Dialysis and Death in Acute Kidney Injury Patients

The specific timing for discontinuing renal replacement therapy (RRT) in acute kidney injury (AKI) patients is debatable. The predictive abilities of variables at the time of discontinuation of RRT for the long‐term prognoses of patients have not been explored. This study aimed to explore the prognostic factors upon discontinuation of RRT for long‐term chronic dialysis and death of patients with acute RRT‐requiring AKI, thus improving decision making regarding the discontinuation of RRT and the follow‐up of patients thereafter. A cohort of 302 AKI patients who required acute RRT and remained alive and free of dialysis for at least 30 days after discharge from January 2009 to December 2012 were followed up. The predictive abilities of general characteristics, RRT details, and variables upon discontinuation of RRT for long‐term chronic dialysis and all‐cause death were evaluated using Cox proportional hazards models. Kaplan–Meier analysis with a log‐rank test was used to compare the survival curves between the strata of levels of good predictors upon discontinuation of RRT. After a median follow‐up time of 4.1 years, 20 (6.6%) patients initiated chronic dialysis and 56 (18.5%) patients died. A higher CysC level upon discontinuation of RRT (HR 1.520, 95% CI 1.082–2.135; P = 0.016), comorbid chronic kidney disease, and a higher non‐renal Charlson comorbidity index (CCI) were independently predictive for chronic dialysis. The hemoglobin level upon discontinuation of RRT was inversely predictive of death (HR 0.986, 95% CI 0.973–0.999; P = 0.035), and comorbid malignancy, the presence of multiple organ dysfunction syndrome, and a higher non‐renal CCI also predicted death. Urine output upon discontinuation of RRT was marginally inversely predictive of death (HR 0.997, 95% CI 0.994–1.000; P = 0.056). Patients who discontinued RRT with CysC levels 85 g/L, and urine output >1130 mL/24 h showed significantly higher non‐chronic dialysis and survival rates according to a log‐rank test. Our study suggested that upon discontinuation of RRT, higher serum CysC levels had the most promising predictive value for long‐term chronic dialysis, and lower hemoglobin levels predicted long‐term death; lower urine output also marginally predicted long‐term death. Based on the remission of the comprehensive condition, lower CysC levels and higher hemoglobin levels and urine output should be considered in the decision to stop RRT. Patients showing worse le