J Crit Care. 2021 Sep 7;66:96-101. doi: 10.1016/j.jcrc.2021.08.011. Online ahead of print.
PURPOSE: Reliable prediction of early mortality after initiation of renal replacement therapy (RRT) in critically ill patients may inform decision-making regarding this treatment. Our primary objective was to identify predictors of mortality within 2 days of starting RRT.
MATERIALS AND METHODS: Patients with acute kidney injury (AKI), receiving RRT, and admitted to intensive care units of one hospital were included. Associations between baseline risk factors and mortality at 2 days and at hospital discharge were analyzed using logistic regression. Discrimination of both models was assessed.
RESULTS: We included 626 patients, treated initially with intermittent RRT (n = 300, 47.9%), continuous RRT (n = 211, 33.7%), or sustained low-efficiency dialysis (n = 115, 18.4%). Two-day mortality after starting RRT was 12.9% (n = 81), and hospital mortality was 50.5% (n = 316). Independent predictors of 2-day mortality included primary diagnostic category (p = 0.004) and sepsis-related organ failure assessment (SOFA) score (odds ratio [OR] 1.36 per point, 95% confidence interval [CI] 1.24-1.50). Independent predictors of hospital mortality included SOFA (1.29, 95%CI 1.21-1.37), Charlson score (1.20, 95%CI 1.18-1.43), and interhospital transfer (OR 0.55, 0.38-0.81). C-statistics were 0.81 (2-day mortality) and 0.80 (hospital mortality).
CONCLUSIONS: Higher SOFA was associated with 2-day mortality after RRT initiation and with hospital mortality. Discrimination in both models was modest.