Assessing the Prognostic Accuracy of the PSS, pSOFA, and SIRS for In-Hospital Mortality in Children with Suspected Infection in Non-ICU Settings: A Multicenter Retrospective Study

To systematically compare the predictive accuracy of the Phoenix Sepsis Score (PSS), the Pediatric Sequential Organ Failure Assessment (pSOFA), and systemic inflammatory response syndrome (SIRS) in assessing in-hospital mortality risk among pediatric sepsis patients in non-ICU wards, thereby providing evidence-based support for clinical risk stratification. This study employed a multicenter retrospective cohort design, enrolling non-ICU pediatric patients with suspected infections (excluding preterm infants and neonates hospitalized immediately after birth), to construct an overall cohort and a neonatal subgroup cohort. Clinical parameters were collected through a data acquisition system, with parallel calculations of PSS, pSOFA, and SIRS scores. The primary endpoint was in-hospital mortality. Receiver operating characteristic (ROC) curves were constructed via R language version 4.3.1, and the discriminatory performance of each scoring system was evaluated via the area under the ROC curve (AUROC). From January 2023 to September 2024, 965 non-ICU pediatric patients with infections were enrolled from 13 medical centers (overall cohort mortality: 1.1%, 11/965; neonatal subgroup: 193 cases, mortality: 4.1%, 8/193). In the overall cohort, all 11 patients experienced deterioration on the ward and were subsequently transferred to the PICU in accordance with each hospital's protocol; eight of them died from sepsis, and three from septic shock. After multivariate adjustment, the AUROC value of the PSS score in the overall cohort was 0.756 (99% CI: 0.731-0.780), representing a 3.6% improvement over the SIRS criteria (0.730, 99% CI: 0.543-0.873) but lower than the pSOFA score (0.845, 99% CI: 0.568-0.966). In the neonatal subgroup, the AUROC value of the PSS score was 0.764 (99% CI: 0.478-0.927), comparable to that of the SIRS criteria (0.757, 99% CI: 0.475-0.905) but still lower than that of the pSOFA score (0.819, 99% CI: 0.533-0.983). No statistically significant differences were observed among the three scoring systems (P > 0.05). These findings suggest that the PSS score does not demonstrate superior predictive value compared with traditional assessment tools in non-ICU clinical scenarios.