Abstract: Objective To determine whether the nomogram for predicting the probability of urosepsis progressed from ureteral calculi was generally applicable to clinical practice. Methods The clinical data of 317 patients with ureteral calculi in our department from January 2016 to December 2016 were collected in this study. A total of 29 patients with urosepsis were selected into the urosepsis group, and the other 288 patients without urosepsis in the same period were selected into the non-urosepsis group. Using the prediction model developed before, the probability of urosepsis was predicted by gender, functional solitary kidney, average CT value of hydronephrosis, urine WBC count and urine nitrite in two groups, and the difference between the predicted and observed probabilities was compared. The discrimination and calibration of the prediction model were validated by using ROC and GiViTI calibration curve belts, respectively. Results The area under ROC curve (AUC) was 0.874 (95%CI: 0.804-0.945), suggesting that the prediction model could distinguish the patients with urosepsis. The 95%CI region of GiViTI calibration belt did not cross the 45-degree diagonal bisector line (P=0.176). Therefore, the prediction probability of the model was consistent with the actual probability, which suggested that prediction model had strong concordance performance. Conclusion The accuracy of the model in predicting the risk of urosepsis in ureteral calculi is improved, and it was helpful to improve the early identification and screening ability of those high-risk patients.

Key words: ureteral calculi, urosepsis, prediction model, external validation