Xuan-Wen Ru, Rui-Ying Liu, Zi-Hao Yang, Cai-Zhi Huang, Cong Zhang, Tie-Wei Li, Jun-Mei Yang, Yang Liu, Xiao-Yu Cui, Hong-Bing Chen, Xun Zhou, Feng Tang, Xu-Yang Gong, Ping Ling, Xin Lv, Qian Zeng, Zhan Ma, Bao-Yu Yuan, Hong Zhu, Xue-Jun Shao, Jing Wu, Rui-Jie Yu, Xiao-Wei Zhang, Jing-Ran Wang, Feng Cheng, Su-Hong Huang, Kun Chi, Guo-Feng Liu, Zhen-Wen Zhou, Xiao-Chun Liu, Wei-Li Yan, Guo-Qiang Qi, Wen-Xia Shao, Qing Ye
In this study, the PSS demonstrated superior predictive performance in the overall cohort, whereas no statistically significant difference was observed between the two scores in the neonatal sub-cohort. In the hematology sub-cohort, both scores showed limited discriminative ability.
BACKGROUND: This study compared the prognostic accuracy of the Phoenix sepsis score (PSS) and the pediatric sequential organ failure assessment (pSOFA) score for predicting in-hospital mortality in critically ill children with suspected infection, with a focus on neonatal and hematology sub-cohorts.
METHODS: This retrospective, multicenter cohort study analyzed data from 2374 children (< 18 years) with suspected infections admitted to 13 intensive care units (ICUs) between 2023 and 2025. Two predefined sub-cohorts were examined: the neonatal sub-cohort (≤ 28 days; n = 576) and the hematology sub-cohort (n = 494). PSS and pSOFA scores were calculated using the worst available physiological and laboratory data within 24 hours of ICU admission. The primary outcome was in-hospital mortality. Prognostic accuracy was assessed mainly via the adjusted area under the receiver operating characteristic curve (AUROC).
RESULTS: Compared with the pSOFA score, the PSS demonstrated significantly greater discrimination for mortality (adjusted AUROC: 0.840 vs. 0.817; P < 0.001) in the overall cohort (n = 2374) and the hematology sub-cohort (0.718 vs. 0.682; P = 0.006) but not in neonates (0.915 vs. 0.850; P = 0.222). At the ≥ 2-point threshold, both scores maintained high sensitivity (> 80%) but low positive predictive value (PPV) (≤ 50%) across all cohorts. In the high-event-rate hematology sub-cohort, the PSS had an approximately 5% lower sensitivity for a meaningfully higher PPV than the pSOFA score. In contrast, within the low-event-rate neonatal sub-cohort, both tools showed equivalent sensitivity and similarly low PPVs (< 8%).
CONCLUSIONS: In this study, the PSS demonstrated superior predictive performance in the overall cohort, whereas no statistically significant difference was observed between the two scores in the neonatal sub-cohort. In the hematology sub-cohort, both scores showed limited discriminative ability.