Fig. 3.

Prediction model for evaluating effectiveness of midodrine hydrochloride in children with vasovagal was assessed by receiver operating characteristic (ROC) curve, precision-recall (PR) curve, calibration curve, and decision curve analysis (DCA). a ROC curve. Specificity (the abscissa): 79.3%. Sensitivity (the ordinate): 77.3%. Area under the curve was 0.79 with a 95% confidence interval of 0.66–0.92. b PR curve. Precision is the abscissa and recall is the ordinate. Area under the curve was 0.833. c Calibration curve. The abscissa is the predicted incidence of events (predicted probability), and the ordinate is the observed incidence of actual events (observed probability). The gray line on the diagonal indicates that the predicted and actual probabilities of the present model are identical. The black line represents the actual predictive accuracy of the model. The calibration curve is well-fitted. d DCA. Y axis equals the net benefit, and x axis equals the threshold probability. Serum uric acid-based treatment decision resulted in greater net benefit over that without uric acid especially when threshold probability > 0.35