Predicting the risk of acute kidney injury after cardiopulmonary bypass: development and assessment of a new predictive nomogram

Abstract

Background: Acute kidney injury (AKI) is a common and severe complication of cardiac surgery with cardiopulmonary bypass (CPB). This study aimed to establish a model to predict the probability of postoperative AKI in patients undergoing cardiac surgery with CPB.

Methods: We conducted a retrospective, multicenter study to analyze 1082 patients undergoing cardiac surgery under CPB. The least absolute shrinkage and selection operator regression model was used to optimize feature selection for the AKI model. Multivariable logistic regression analysis was applied to build a prediction model incorporating the feature selected in the previously mentioned model. Finally, we used multiple methods to evaluate the accuracy and clinical applicability of the model.

Results: Age, gender, hypertension, CPB duration, intraoperative 5% bicarbonate solution and red blood cell transfusion, urine volume were identified as important factors. Then, these risk factors were created into nomogram to predict the incidence of AKI after cardiac surgery under CPB.

Conclusion: We developed a nomogram to predict the incidence of AKI after cardiac surgery. This model can be used as a reference tool for evaluating early medical intervention to prevent postoperative AKI.

Keywords: Acute kidney injury; Cardiac surgery; Cardiopulmonary bypass; Nomogram; Prevention.

Fig. 1

The LASSO binary logistic regression model was used to select the risk factors of AKI after CPB. A The choice of the best parameter (lambda) in the LASSO model passed the lowest standard for five times of cross-validation. The partial likelihood deviation (binomial deviation) curve is plotted against log (λ). Draw a dashed vertical line with the best value by using the minimum standard and the minimum standard 1 SE (1-SE standard). B LASSO coefficient curve of 17 features. A coefficient distribution map for the logλ sequence is generated. Vertical lines are drawn at the values selected using five-fold cross-validation, where the optimal λ produces sixteen features with non-zero coefficients

Fig. 2

The nomogram of AKI after CPB surgery. The Nomogram of AKI after CPB surgery estimates the probability of AKI after the operation. Mark the patient value on each axis, draw a straight line perpendicular to the point axis, and add the points of each variable. Next, mark the sum on the total point axis and draw a straight line perpendicular to the probability axis. RBC- intraoperative transfusion of red blood cells; SBS-intraoperative 5% sodium bicarbonate solution infusion; Urine-intraoperative urine output

Fig. 3

Calibration curve for nomogram prediction of AKI after CPB in the cohort. The x-axis represents the predicted risk of AKI after CPB. The y-axis represents the actual occurrence of AKI after CPB. The dotted line on the diagonal represents the perfect prediction of the ideal model. The solid line represents the performance of the nomogram. The closer the fit to the diagonal dashed line is, the better the prediction effect

Fig. 4

Decision curve analysis for the AKI nomogram. The y-axis is net income. The blue line represents the distribution of AKI after CPB. The thin solid line indicates that it is assumed that all patients undergoing CPB surgery do not develop AKI after surgery. The thick solid line represents the hypothesis that all patients have AKI. The decision curve shows that if the threshold probability is greater than 4%, the use of the nomogram of AKI after CPB in the current study to predict the risk of AKI will increase the benefit more than the full-patient intervention program or the non-intervention program

Fig. 5

The importance of normalizing. The independent variable importance map can be used to analyze which factors have a greater impact on the predicted value, and the more obvious the importance, the greater the impact on the predicted value. This figure suggests that Age has the greatest impact on postoperative AKI.

Fig. 6

ROC curve of neural network model. The ROC curve is a curve reflecting the relationship between sensitivity and specificity. The abscissa (X-axis) is 1 – specificity, also known as the false positive rate (false positive rate), the closer the X-axis is to zero, the higher the accuracy; the ordinate (Y-axis) is called sensitivity, also known as true positives rate (sensitivity), the larger the Y-axis, the better the accuracy. The area under the ROC curve (AUC) can evaluate the quality of the model. If the area under the ROC curve is greater than 0.5, it proves that the model has certain value. The closer the AUC is to 1, the better the authenticity of the model is proved. The area under the ROC curve of this neural network model was 0.749, which confirmed the good value of this model