Risk prediction of postoperative acute kidney injury in patients undergoing off-pump coronary artery bypass grafting based on machine learning.

ObjectiveTo identify risk factors for postoperative acute kidney injury in patients undergoing off-pump coronary artery bypass grafting and to develop a machine learning model for effective risk prediction.MethodsWe extracted data from the Medical Information Mart for Intensive Care-IV database for patients who underwent off-pump coronary artery bypass grafting. Eight machine learning algorithms were trained, including logistic regression, support vector machine, gradient boosting machine, neural network, eXtreme gradient boosting, adaptive boosting, light gradient boosting machine, and categorical boosting. Model performance was assessed using the area under the receiver operating characteristic curve and other relevant metrics. The best-performing model was further interpreted using Shapley additive explanations.ResultsPatients were randomly divided into a training set and a testing set at a 7:3 ratio. Models were trained on the training set, internally validated on the testing set, and externally validated using the electronic intensive care unit Collaborative Research Database. Independent predictors of postoperative acute kidney injury were age, body weight, bicarbonate, international normalized ratio, respiratory rate, and diastolic blood pressure. Among all models, categorical boosting achieved the highest performance, with an area under the receiver operating characteristic curve of 0.826 in the training set and 0.725 and 0.705 in the testing and external validation sets, respectively.ConclusionThe categorical boosting model, incorporating age, body weight, bicarbonate, international normalized ratio, respiratory rate, and diastolic blood pressure, provides a potentially valuable tool for predicting the risk of postoperative acute kidney injury in patients undergoing off-pump coronary artery bypass grafting.