A Preoperative Prediction Model for Acute Kidney Injury After Off-Pump Coronary Artery Bypass Grafting.

Abstract

Objectives: Off-pump coronary artery bypass (OPCAB) is considered renal-protective; however, the incidence of acute kidney injury (AKI) remains high. Because AKI is associated with adverse outcomes, accurate prediction and prevention are essential. We developed and evaluated a preoperative prediction model for AKI after OPCAB. Model performance was also evaluated by applying the AKI model to predict renal replacement therapy (RRT).

Design: Diagnostic/prognostic study.

Setting: Two university hospitals and 10 tertiary care centers.

Participants: A total of 1,640 OPCAB patients.

Interventions: None.

Measurements and main results: Model performance was evaluated in terms of discrimination, calibration, and net benefit. AKI occurred in 290 patients (17.7%), and RRT was required in 33 patients (2.0%). Preoperative predictors were male sex, obesity, contrast use within 7 days, diabetes, absence of aspirin therapy, anemia, and chronic kidney disease. Postoperative predictors included unplanned cardiopulmonary bypass, oliguria, and prolonged surgery. The area under the receiver operating characteristic curves (AUCs) were 0.715 for the preoperative model and 0.753 for the postoperative model (p < .001). For RRT prediction, the AUCs of the preoperative and postoperative models were 0.776 and 0.800 (p = .441), respectively. All calibration plots confirmed agreement between predicted and observed probabilities (all Hosmer-Lemeshow goodness-of-fit tests p > .05). The preoperative and postoperative models provided greater net benefit compared with previously reported models using decision curve analysis.

Conclusion: We developed a preoperative AKI prediction model based solely on 7 simple variables, which demonstrated good performance comparable to models incorporating postoperative factors and was also applicable to RRT prediction.