Predicting acute renal failure after cardiac surgery: validation and re-definition of a risk-stratification algorithm.

BACKGROUND: Acute renal failure (ARF) after cardiac surgery is associated with significant morbidity and mortality, irrespective of the need for dialysis. Previous studies have attempted to identify predictors of ARF and develop risk stratification algorithms. This study aims to validate the algorithm in an independent cohort of patients that includes a significant proportion of female and black patients and compares two different definitions of renal outcome.
METHODS: A large single center cardiac surgery database was examined (n, 24,660; 1993-2000) which included 29.9% females and 3.7% black patients. Post-operative ARF was defined as: a) ARF requiring dialysis, b) > 50% reduction in creatinine clearance relative to baseline or requiring dialysis. Clinical variables related to baseline renal function and cardiovascular disease were used in recursive partitioning analysis for both outcome definitions. Chi-square goodness of fit analysis was performed to validate the algorithm.
RESULTS: The frequency of post-operative ARF requiring dialysis ranged between 0.5 and 15.5% based on the risk categories with the area under the receiver operating characteristic (ROC) curve of 0.78. Using the more inclusive definition of ARF, the frequency was significantly higher ranging from 2.6 to 25%(P < 0.001) with an area under ROC curve of 0.65.
CONCLUSIONS: The renal risk stratification algorithm is valid in predicting post-operative ARF in an independent cohort of patients, well represented by differences in gender and race. Since the need for dialysis remains subjective, a more objective and inclusive definition of ARF may help in identifying a larger number of patients 'at-risk'.