Predicting acute renal failure after coronary bypass surgery: cross-validation of two risk-stratification algorithms

Kidney Int. 2000 Jun;57(6):2594-602. doi: 10.1046/j.1523-1755.2000.00119.x.


Background: Acute renal failure (ARF) requiring dialysis after coronary artery bypass grafting (CABG) occurs in 1 to 5% of patients and is independently associated with postoperative mortality, even after case-mix adjustment. A risk-stratification algorithm that could reliably identify patients at increased risk of ARF could help improve outcomes.

Methods: To assess the validity and generalizability of a previously published preoperative renal risk-stratification algorithm, we analyzed data from the Quality Measurement and Management Initiative (QMMI)1 patient cohort. The QMMI includes all adult patients (N = 9498) who underwent CABG at 1 of 12 academic tertiary care hospitals from August 1993 to October 1995. ARF requiring dialysis was the outcome of interest. Cross-validation of a recursive partitioning algorithm developed from the VA Continuous Improvement in Cardiac Surgery Program (CICSP) was performed on the QMMI. An additive severity score derived from logistic regression was also cross-validated on the QMMI.

Results: The CICSP recursive partitioning algorithm discriminated well (ARF vs. no ARF) in QMMI patients, even though the QMMI cohort was more diverse. Rates of ARF were similar among risk subgroups in the CICSP tree, as was the overall ranking of subgroups by risk. Using logistic regression, independent predictors of ARF in the QMMI cohort were similar to those found in the CICSP. The CICSP additive severity score performed well in the QMMI cohort, successfully stratifying patients into low-, medium-, high-, and very high-risk groups.

Conclusions: The CICSP preoperative renal-risk algorithms are valid and generalizable across diverse populations.

MeSH terms

  • Acute Kidney Injury / etiology*
  • Aged
  • Aged, 80 and over
  • Algorithms
  • Cohort Studies
  • Coronary Artery Bypass*
  • Evaluation Studies as Topic
  • Female
  • Forecasting
  • Humans
  • Male
  • Postoperative Complications*
  • ROC Curve
  • Regression Analysis
  • Risk Assessment