A comparison of pediatric sepsis definitions based on systemic inflammatory response syndrome and Phoenix criteria: a single-center PICU retrospective study

BACKGROUND: The 2024 international pediatric sepsis consensus definition has undergone a paradigm shift from a systemic inflammatory response syndrome (SIRS)-based framework to the organ dysfunction-centered Phoenix Sepsis Criteria (PSC). We aimed to evaluate the diagnostic concordance, predictive performance for 28-day pediatric intensive care unit (PICU) mortality, and phenotypic overlap between these two pediatric sepsis definitions. METHODS: This single-center retrospective cohort study included 1034 children aged > 1 month to < 18 years with confirmed or suspected infection who were directly admitted to the PICU of Children's Hospital of Chongqing Medical University between January 1, 2020, and November 21, 2023. All patients were independently evaluated for sepsis using both the 2005 International Pediatric Sepsis Consensus Conference (IPSCC) SIRS criteria and the 2024 PSC. Diagnostic agreement was assessed using the Kappa coefficient. Binary logistic regression was employed to establish association models between factors and phenotypes, as well as between factors and PICU 28-day mortality. Predictive performance was compared using the C-statistic. RESULTS: Among 1034 patients, 613 (59.3%) met the Sepsis-SIRS criteria with a 28-day PICU mortality of 15.2% (93/613), 744 (72.0%) met the Sepsis-Phoenix criteria with a mortality of 16.3% (121/744), 489 (47.3%) met both criteria with a mortality of 18.6% (91/489), and 166 (16.1%) met neither criterion with a mortality of 2.4% (4/166). Agreement between the two criteria was poor (kappa = 0.202, 95% CI: 0.143-0.261). After adjusting for clinically relevant confounders, the PSC remained a strong independent predictor of 28-day mortality (adjusted OR = 5.123, 95% CI: 2.128-12.333, p < 0.001), whereas the SIRS criteria showed no independent predictive value (adjusted OR = 0.937, 95% CI: 0.523-1.678, p = 0.827). The PSC demonstrated significantly superior discriminatory ability compared with the SIRS criteria (C-statistic = 0.809 vs. 0.589, p < 0.001). Notably, 20.2% of SIRS-positive patients were not classified as sepsis by the Phoenix Criteria, and this subgroup had an extremely low mortality of 1.6%, reflecting higher specificity of the PSC. CONCLUSIONS: The SIRS and PSC identify partially overlapping populations with distinct risk stratification, showing poor diagnostic concordance. The PSC has superior independent predictive performance for PICU 28-day mortality and may be considered for prognostic assessment of infected children in the PICU setting. Importantly, historical study results based on the SIRS criteria should be extrapolated cautiously to PSC-defined populations.