Background: A predictive model for hospital mortality in postoperative acute renal failure (ARF) patients requiring renal replacement therapy (RRT) may aid clinicians' therapeutic decision-making and research design.
Methods: A prospective observational study of 398 postoperative ARF patients requiring RRT was conducted in four hospitals. The derivation cohort consisted of 334 patients recruited between January 2002 and December 2005. The validation cohort consisted of 64 patients recruited between January 2006 and December 2006.
Results: The hospital mortality rates for the derivation and validation cohorts were 65.6 and 62.5%, respectively. A modified Sequential Organ Failure Assessment (SOFA) score was constructed at the commencement of RRT by a formula of serum lactate level (mM) + 2 x (generic SOFA score) + 3 x (age per decade) + 8 (if mechanical circulatory support required) + 10 (if total parenteral nutrition required) + 11 (if status postcardiopulmonary resuscitation) + 13 (if positive sepsis sign). The area under the receiver operating characteristic curve of the model for the derivation and validation cohorts was 0.804 and 0.839, respectively.
Conclusion: This validated score at dialysis commencement might assist clinicians in estimating hospital mortality, planning future clinical trials, and providing quantitative guidance for decision making in postoperative ARF patients requiring RRT.
Copyright 2008 S. Karger AG, Basel.