Aim Acute kidney injury (AKI) is a major complication of coronary artery bypass grafting (CABG). Existing predictive models show limited accuracy and rarely include modifiable intraoperative risk factors. Therefore, a prediction model was developed to identify modifiable variables to guide targeted prevention strategies. Methods We collected data from 1174 patients who underwent CABG at our hospital between January 2020 and May 2025. AKI was defined according to the KDIGO 2012 criteria. Missing data (<2%) were processed using multiple imputation (MICE algorithm). We constructed a preoperative model (baseline variables only) and a complete model (baseline plus intraoperative variables). Model discrimination was performed using the area under the receiver operating characteristic curve (AUC-ROC), and internal validation was conducted using bootstrap resampling (500 iterations). The clinical utility was assessed using decision curve analysis. Results AKI occurred in 297 patients (25.3%). The full model had six predictors: age, body mass index, diabetes, estimated glomerular filtration rate, hemoglobin nadir during cardiopulmonary bypass, and maximum lactate. Hemoglobin nadir was the most robust modifiable predictor (odds ratio 0.73/SD, 95% CI 0.61-0.86, p < 0.001). The bootstrap-corrected AUC of 0.661 with minimal optimism (0.008) suggested solid internal validity. Model discrimination was consistent across subgroups (AUC range 0.614-0.672). Patients with chronic kidney disease demonstrated the highest AKI incidence (39.1%). AKI also increased the hospital length of stay by 26.6% (95% CI 19.6%–34.1%, p < 0.001) after adjusting for confounders. Decision curve analysis yielded a positive net benefit between clinically relevant risk thresholds (15–35%). Conclusion We developed and internally validated a predictive model with acceptable discrimination and reasonable optimism in a contemporary CABG population. The hemoglobin nadir during cardiopulmonary bypass emerged as the most potent modifiable risk factor, with dose-response analysis suggesting a threshold of 8–9 g/dL as an actionable target for AKI prevention. High-risk populations showed markedly elevated AKI incidence, and the model quantified considerable clinical and economic impacts. External validation is required.
Pierri et al. (Sat,) studied this question.