Background: A recently published score predicts the occurrence of acute kidney injury (AKI) after liver resection based on preoperative parameters (chronic renal failure, cardiovascular disease, diabetes, and alanine-aminotransferase levels). By inclusion of additional intraoperative parameters we aimed to develop a new prediction model.
Methods: A series of 549 consecutive patients were enrolled. The preoperative score and intraoperative parameters (blood transfusion, hepaticojejunostomy, oliguria, cirrhosis, diuretics, colloids, and catecholamine) were included in a multivariate logistic regression model. We added the strongest predictors that improved prediction of AKI compared to the existing score. An internal validation by fivefold cross validation was performed, followed by a decision curve analysis to evaluate unnecessary special care unit admissions.
Results: Blood transfusions, hepaticojejunostomy, and oliguria were the strongest intraoperative predictors of AKI after liver resection. The new score ranges from 0 to 64 points predicting postoperative AKI with a probability of 3.5–95 %. Calibration was good in both models (15 % predicted risk vs. 15 % observed risk). The fivefold cross-validation indicated good accuracy of the new model (AUC 0.79 (95 % CI 0.73–0.84)). Discrimination was substantially higher in the new model (AUCnew 0.81 (95 % CI 0.76–0.86) versus AUCpreoperative 0.60 (95 % CI 0.52–0.69), p < 0.001). The new score could reduce up to 84 unnecessary special care unit admissions per 100 patients depending on the decision threshold.
Conclusions: By combining three intraoperative parameters with the existing preoperative risk score, a new prediction model was developed that more accurately predicts postoperative AKI. It may reduce unnecessary admissions to the special care unit and support management of patients at higher risk.