Derivation and validation of a model to predict acute kidney injury following cardiac surgery in patients with normal renal function

Background The study aimed to construct a clinical model based on preoperative data for predicting acute kidney injury (AKI) following cardiac surgery in patients with normal renal function. Methods A total of 22,348 consecutive patients with normal renal function undergoing cardiac surgery were enrolled. Among them, 15,701 were randomly selected for the training group and the remaining for the validation group. To develop a model visualized as a nomogram for predicting AKI, logistic regression was performed with variables selected using least absolute shrinkage and selection operator regression. The discrimination, calibration, and clinical value of the model were evaluated. Results The incidence of AKI was 25.2% in the training group. The new model consisted of nine preoperative variables, including age, male gender, left ventricular ejection fraction, hypertension, hemoglobin, uric acid, hypomagnesemia, and oral renin-angiotensin system inhibitor and non-steroidal anti-inflammatory drug within 1 week before surgery. The model had a good performance in the validation group. The discrimination was good with an area under the receiver operating characteristic curve of 0.740 (95% confidence interval, 0.726–0.753). The calibration plot indicated excellent agreement between the model prediction and actual observations. Decision curve analysis also showed that the model was clinically useful. Conclusions The new model was constructed based on nine easily available preoperative clinical data characteristics for predicting AKI following cardiac surgery in patients with normal kidney function, which may help treatment decision-making, and rational utilization of medical resources.