Fig. 5

Receiver operator characteristic (ROC); Calibration curve analysis (CCA); Decision Curve Analysis (DCA). In predicting a “Un-worsened” outcome, the ROC curve (A) shows a good performance (AUC = 0.824). The CCA (B) shows a passable consistency between the actual predicting performance (Solid blue line) and the ideal predicting performance (Dotted gray line). The bar-chart (C) shows the best cut-off point of this model, which allows for a perfect prediction of a negative outcome (“Worsened”, Blue bar). Yet, the predicting performance for a “Un-worsened” outcome is unsatisfactory because of the presence of multiple false-positive cases, which results in a comprehensive predicting accuracy of 0.711. The DCA (D) shows the advantages of this model (Solid red line) compared with the “treat-none strategy” (Dotted black line), while its advantages over the “treat-all strategy” (Dotted black line) are not as significant