Journal Article

Validation and modification of the phoenix sepsis score for predicting in-hospital mortality in children with suspected infection admitted to the intensive care unit

Pan, Wei, Shu, Ting, Qiu, Yehan, Zheng, Liling, Li, Zhe, Wang, Shaobo, Shi, Xianqing, Du, Bin, Huang, Ziyang, Zhou, Xiang, Guan, Yi, Li, Yaguang, Qu, Dong, Su, Longxiang, Jiang, Huizhen, Chen, Jieqing, Chen, Yujie, Wang, Lu, Fu, Yanwen, Ding, Xin, Ning, Botao

2026

Overview

The Society of Critical Care Medicine Pediatric Sepsis Definition Task Force recommends the use of the Phoenix Sepsis Score (PSS) to diagnose pediatric sepsis among children with suspected infection. However, the performance of the PSS in different healthcare settings remains unclear. The present study aimed to validate and modify the ability of the PSS to predict in-hospital death in children with suspected infection admitted to the intensive care unit (ICU) in China. This multicenter retrospective cohort study included children aged ≤18 years with suspected infection admitted to the ICU from five hospitals in China between January 2012 and December 2023. Two children's specialized hospitals and three tertiary general hospitals participated. The primary evaluation of PSS prediction performance was based on in-hospital mortality discrimination through the area under the receiver operating characteristic curve (AUROC). The secondary evaluation was conducted using the area under the precision-recall curve (AUPRC). Given the moderate discriminative performance of the PSS in this cohort, we employed extreme gradient boosting (XGBoost) with SHapley Additive exPlanations (SHAP) to identify key predictors of in-hospital mortality. Predictors selected through this process were incorporated into the original PSS framework while prioritizing clinical feasibility, resulting in a modified score, PSS+, for mortality prediction. Among 9221 ICU admissions for children with suspected infections (13.4% mortality), the PSS showed moderate discrimination, with an AUROC of 0.60 (95% CI: 0.59-0.62). The modified PSS (PSS+), which selectively incorporated comorbidities, vital signs, and demographic variables identified through model development, outperformed the original PSS-4, PSS-8, and pediatric sequential organ failure assessment (pSOFA) in terms of both internal validation (AUROC of 0.75, 95% confidence interval [CI]: 0.70-0.78) and external validation (AUROC of 0.71, 95% CI: 0.69-0.73). The PSS demonstrated only moderate ability to predict in-hospital mortality among pediatric ICU patients in the present cohort, indicating that its application in other healthcare settings should be approached with caution. The modified PSS+ scoring system significantly improved mortality prediction performance and may serve as a more reliable and clinically applicable tool for risk assessment in critically ill children.