Skip to main content

Table 2 Performance of the risk models in the derivation dataset

From: Prediction of acute kidney injury risk after cardiac surgery: using a hybrid machine learning algorithm

  Discrimination (AUC) Calibration (Hosmer–Lemeshow) Performance
Machine Learning model 0.75 χ2 = 7.35, p = 0.393 Predicted risk of 3%*:
Sensitivity = 67.1%
Specificity = 94.1%
PPV = 50.2%
NPV = 87.6%
Traditional logistic regression model 0.72 χ2 = 50.69, p < 0.001 Predicted risk of 2%*:
Sensitivity = 62.2%
Specificity = 65.8%
PPV = 40.9%
NPV = 82.1%
Enhanced logistic regression model 0.74 χ2 = 9.65, p = 0.290 Predicted risk of 2%*:
Sensitivity = 66.3%;
Specificity = 79.1%;
PPV = 47.5%;
NPV = 84.4%
  1. *Predicted probability threshold with the optimal operating characteristics
  2. AUC, area under the Receiver-operating characteristics curve; PPV, positive prediction value; NPV, negative predictive value