Development and validation of the UCSS score, a novel method to predict septic shock after PCNL

Abstract

Objective: To develop an objective and easily recognizable model to predict septic shock following percutaneous nephrolithotomy (PCNL).

Subjects and methods: First, we identified differences between 431 patients who underwent PCNL with or without septic shock. These data were used to develop existing models and examine their improvement. Multivariate analysis was applied to identify risk factors of septic shock after PCNL based on the scores allocated to the PCNL postoperative test indicators. Finally, we developed a predictive nomogram using the selected factors and compared its performance with that of the existing nomograms SOFA, qSOFA, and SIRS.

Results: Twelve (2.8%) of the patients met the criteria for postoperative septic shock after PCNL. Baseline data analysis revealed differences in sex, preoperative drainage, urinary culture, and urinary leukocyte between groups. After transforming patient data into measurement-level data, we investigated each index score in these conditions, and found that the incidence of septic shock generally increased with the score. Multivariate analysis and early optimization screening revealed that septic shock factors could be predicted using platelets, leukocytes, bilirubin, and procalcitonin levels. We further compared the prediction accuracy of urinary calculi-associated septic shock (UCSS), SOFA, qSOFA, and SIRS scores using the AUC of the ROC curve. As compared to SIRS [AUC 0.938 (95% CI 0.910-0.959)] and qSOFA [AUC 0.930 (95% CI 0.901-0.952)], UCSS [AUC 0.974 (95% Cl 0.954-0.987)] and SOFA [AUC 0.974 (95% CI 0.954-0.987)] scored better at discriminating septic shock after PCNL. We further compared the ROC curves of UCSS with SOFA (95% CI - 0.800 to 0.0808, P = 0.992), qSOFA (95% CI - 0.0611 to 0.0808, P = 0.409), and SIRS (95% CI - 0.0703 to 0.144, P = 0.502), finding that UCSS was non-inferior to these models.

Conclusions: UCSS, a new convenient and cost-effective model, can predict septic shock following PCNL and provide more accurate discriminative and corrective capability than existing models by including only objective data. The predictive value of UCSS for septic shock after PCNL was greater than that of qSOFA or SIRS scores.

Keywords: Percutaneous nephrolithotomy; Risk model; Septic shock; Systemic inflammatory response syndrome; Urinary calculi.