Abstract
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.
Supplementary Information
The online version contains supplementary material available at https://doi.org/10.1186/s13052-026-02298-3.
Keywords: SIRS, Phoenix, Sepsis, Septic shock, PICU 28-day mortality, Prediction models
Introduction
Infectious diseases account for 27.7% of the global disease burden and are the leading cause of child mortality worldwide [1, 2]. However, the clinical manifestations of these infected patients vary greatly, ranging from mild symptoms to severe infection or multiple organ failure and even death. Over the past three decades, experts in critical care medicine have worked to develop a standardized definition of “sepsis” to help healthcare providers identify, treat and study infected patients that are at a higher risk of morbidity and mortality [3–5].
The 2005 First International Pediatric Sepsis Consensus Conference (IPSCC) [6] defined sepsis as the presence of systemic inflammatory response syndrome (SIRS) in patients with suspected or confirmed infection. At least two SIRS criteria must be met, one of which must be an abnormal body temperature or white blood cell count. Although this definition is convenient, its specificity is limited, does not allow for risk stratification in both lower- and higher-resource settings, and may be discordant with clinician-based diagnosis [7]. In 2024, the Society of Critical Care Medicine (SCCM) [8] released the latest International Consensus Criteria for Pediatric Sepsis and Septic Shock, defining sepsis as suspected or confirmed infection accompanied by life-threatening organ dysfunction, specifically a Phoenix Sepsis Score ≥ 2 points.
Although the Sepsis-Phoenix criteria demonstrate superior discriminatory ability in predicting mortality compared to the Sepsis-SIRS criteria [7, 9], few studies have examined the extent of overlap between populations identified by these two pediatric sepsis definitions [10]. If the populations are nearly identical, historical studies of Sepsis-SIRS patients are likely generalizable to Sepsis-Phoenix patients. Additionally, in resource-poor settings where laboratory indicators are difficult to obtain, hospitals may continue using the Sepsis-SIRS criteria. In such cases, determining the generalizability of data from Sepsis-SIRS patients to Sepsis-Phoenix patients is crucial.
We hypothesized that patients identified by the two criteria would have different mortality profiles, particularly among three groups: (1) those meeting both criteria, (2) those meeting only SIRS criteria, and (3) those meeting only Phoenix criteria. Our specific objectives were to: (1) evaluate the diagnostic concordance between SIRS and Phoenix sepsis criteria; (2) compare their predictive performance for PICU 28-day mortality; and (3) analyze the phenotypic overlap characteristics and clinical applicability of the identified populations.
Methods
Study design and ethics approval
This single-center retrospective cohort study was conducted at the PICU of Children’s Hospital of Chongqing Medical University, a tertiary care children’s hospital in Southwest China. The study protocol was approved by the Institutional Review Board (IRB) of Children’s Hospital of Chongqing Medical University (IRB No. 2025367) on September 12, 2025, with waiver of informed consent due to the retrospective nature of the study. All procedures were performed in accordance with the ethical standards of the institutional committee and the 1975 Helsinki Declaration.
Study population
Inclusion criteria: Patients (> 1 month and < 18 years) were directly admitted to the PICU of Chongqing Medical University Children’s Hospital from January 1, 2020 to November 21, 2023 who were infected or suspected infection. To date, no universally standardized operational definition of confirmed or suspected infection exists in pediatric sepsis research. We adopted the time-window definition established in classic adult sepsis studies [11, 12]: (1) initiation of systemic antibiotics within 72 h of blood culture collection, or (2) blood culture collection within 24 h of the first dose of systemic antibiotics. The onset of presumed infection was defined as the earlier of the culture or the antibiotic administration [11].
Exclusion criteria
For patients with multiple PICU admissions during the study period, only data from the first admission were included.
PICU characteristics
The Children’s Hospital of Chongqing Medical University is a national tertiary comprehensive children’s hospital and a national regional medical center for children in Southwest China. It operates two campuses (Yuzhong and Liangjiang) with a total of 81 PICU beds (29 in Yuzhong, 52 in Liangjiang). Both PICUs have integrated medical and surgical teams, making this the largest comprehensive pediatric intensive care center in the region.
Data collection
Clinical data were extracted from electronic health records (EHRs) and included: (1) demographic characteristics: age, gender, weight; (2) clinical variables: primary infection site, comorbidities (neoplastic disease, hematological disease, etc.), and presence of septic shock; (3) sepsis definition parameters: all variables required for calculating SIRS and Phoenix scores; and (4) outcome data: 28-day survival status.
We reviewed the documented medical history to assess the primary site of infection in patients with sepsis. The SIRS score was calculated according to the 2005 IPSCC criteria, ranging from 0 to 4 points [6]. The Phoenix score was calculated based on the 2024 SCCM recommended Phoenix sepsis criteria (PSC), ranging from 0 to 13 points [8]. We used the worst values within 24 h of PICU admission to calculate SIRS and Phoenix scores [7, 9, 13]. Neoplastic disease refers to solid tumor diseases. Immune suppression refers to rheumatologic or other pediatric diseases that were treated with immunosuppressive agents [14, 15].
Outcomes
The primary outcome was 28-day PICU mortality, defined as death occurring within 28 days of PICU admission, regardless of location (in-hospital or post-discharge). In-hospital survival status was tracked via the EHR system. For patients discharged alive before 28 days, survival status was confirmed via the hospital’s electronic follow-up system, supplemented by telephone follow-up when necessary. Patients lost to follow-up and those surviving to 28 days were censored.
This study used 28-day PICU mortality rather than the in-hospital mortality endpoint used in the original PSC validation studies, due to differences in PICU admission and discharge practices that exist between China and high-income countries [16]. This difference may reduce comparability with previous literature.
Statistical analysis
Statistical analysis was performed using SPSS 26.0 software (IBM), with data visualization conducted using Origin 2024 and MedCalc software. Quantitative data not meeting normal distribution requirements were expressed as median (interquartile range, IQR), with intergroup comparisons analyzed using the Mann-Whitney U test. Categorical data were presented as frequency and percentage (n%), with intergroup comparisons evaluated using the chi-square test. The overlap between patient populations defined by different sepsis criteria is illustrated using a Venn diagram, with consistency described by the kappa coefficient. Spearman’s correlation coefficient was used to analyze the relationship between SIRS and Phoenix points. Binary logistic regression was used to analyze the association between sepsis criteria and 28-day mortality, adjusting for clinically relevant confounders: age, neoplastic disease, hematological disease, congenital heart disease, primary infection site, and septic shock. Discriminations of the regression models were measured with the C-statistic and compared with the DeLong test. Cochran-Armitage test of trend was used to determine if PICU 28-day mortality increased as SIRS or Phoenix points increased. We also conducted an exploratory subgroup analysis for septic shock based on the two pediatric sepsis definitions. p values < 0.05 were considered statistically significant.
Results
Patient selection
Between January 1, 2020, and November 21, 2023, a total of 1551 patients directly admitted to the PICU of Children’s Hospital of Chongqing Medical University were initially extracted from the hospital’s Sepsis Specialized Disease Database (SSDD). Of these, 1101 were preliminarily identified as confirmed or suspected infection cases. After manual chart review, 67 patients were excluded because they had multiple hospitalisation records and their PICU admission occurred prior to January 1, 2020 (the study start date), as the automated database system could not accurately identify PICU admission timing for patients with multiple hospitalisations. The final study cohort comprised 1034 patients with complete analytical data, among whom 613 (59.3%) met the Sepsis-SIRS definition and 744 (72.0%) met the Sepsis-Phoenix definition (Fig. 1).
Fig. 1.
Flow diagram of study cohort selection
Patient characteristics
Overall, the unadjusted 28-day PICU mortality was 15.2% in patients meeting the 2005 IPSCC SIRS sepsis criteria and 16.3% in those meeting the 2024 PSC. Pulmonary infection was the most common primary infection site in both groups, while bloodstream infection was associated with the highest mortality, followed by intracranial infection. Across both definitions, patients with neoplastic disease, hematological disease, or septic shock had significantly higher unadjusted 28-day mortality compared to those without these comorbidities. Notably, non-survivors had significantly shorter hospital stays than survivors, regardless of the sepsis definition used (Table 1).
Table 1.
Characteristics of study cohorts categorized by pediatric sepsis definition and by PICU 28-day mortality
| Variable | IPSCC definition | Phoenix definition | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Survivors (n = 520) | Non-survivors (n = 93) | Total (n = 613) |
Mortality (%) | p valuea | Survivors (n = 623) |
Non-survivors (n = 121) | Total (n = 744) |
Mortality (%) | p valuea | |
| Baseline demographic information | ||||||||||
| Age, months | 40.0 (10.1-107.8) | 38.0 (8.0-92.5) | 40.0 (10.0-105.0) | N/A | 0.580 | 25.0 (7.3–91.0) |
29.0 (7.3–80.0) |
26.0 (7.3–90.0) | N/A | 0.731 |
| Sex | … | … | … | … | 0.732 | … | … | … | … | 0.873 |
| Male | 292 (56.2%) | 54 (58.1%) | 346 (56.4%) | 15.6% | … | 371 (59.6%) | 73 (60.3%) | 444 (59.7%) | 16.4% | … |
| Female | 228 (43.8%) | 39 (41.9%) | 267 (43.6%) | 14.6% | … | 252 (40.4%) | 48 (39.7%) | 300 (40.3%) | 16.0% | … |
| Weight, kg | 14.0 (8.5–25.8) | 14.0 (7.8–22.0) | 14.0 (8.3–25.0) | N/A | 0.363 | 12.0 (7.5–22.0) |
12.0 (6.8–19.8) |
12.0 (7.0-21.4) | N/A | 0.531 |
| Primary infection sites | … | … | … | … | < 0.001 | … | … | … | … | 0.003 |
| Pulmonary infection | 347 (66.7%) | 44 (47.3%) | 391 (63.8%) | 11.3% | … | 449 (72.1%) | 70 (57.9%) | 519 (69.8%) | 13.5% | … |
| Gastrointestinal infection | 60 (11.5%) | 12 (12.9%) | 72 (11.7%) | 16.7% | … | 65 (10.4%) | 14 (11.6%) | 79 (10.6%) | 17.7% | … |
| Bloodstream infection | 27 (5.2%) | 11 (11.8%) | 38 (6.2%) | 28.9% | … | 26 (4.2%) | 11 (9.1%) | 37 (5.0%) | 29.7% | … |
| Intracranial infection | 60 (11.5%) | 24 (25.8%) | 84 (13.7%) | 28.6% | … | 61 (9.8%) | 24 (19.8%) | 85 (11.4%) | 28.2% | … |
| Other infection | 26 (5.0%) | 2 (2.2%) | 28 (4.6%) | 7.1% | … | 22 (3.5%) | 2 (1.7%) | 24 (3.2%) | 8.3% | … |
| Comorbidities | ||||||||||
| Neoplastic disease | 9 (1.7%) | 7 (7.5%) | 16 (2.6%) | 43.8% | 0.004 | 9 (1.4%) | 11 (9.1%) | 20 (2.7%) | 55.0% | < 0.001 |
| Hematological disease | 21 (4.0%) | 10 (10.8%) | 31 (5.1%) | 32.3% | 0.014 | 18 (2.9%) | 10 (8.3%) | 28 (3.8%) | 35.7% | 0.010 |
| Immune deficiency / immune suppression | 24 (4.6%) | 2 (2.2%) | 26 (4.2%) | 7.7% | 0.420 | 24 (3.9%) | 3 (2.5%) | 27 (3.6%) | 11.1% | 0.636 |
| Malnutrition | 37 (7.1%) | 9 (9.7%) | 46 (7.5%) | 19.6% | 0.388 | 45 (7.2%) | 10 (8.3%) | 55 (7.4%) | 18.2% | 0.689 |
| Congenital heart disease | 33 (6.3%) | 13 (14.0%) | 46 (7.5%) | 28.3% | 0.010 | 57 (9.1%) | 21 (17.4%) | 78 (10.5%) | 26.9% | 0.007 |
| Septic shock | 179 (34.4%) | 62 (66.7%) | 241 (39.3%) | 25.7% | < 0.001 | 181 (29.1%) | 68 (56.2%) | 249 (33.5%) | 27.3% | < 0.001 |
| Hospital length of stay (d) | 19.0 (11.0–29.0) |
5.0 (2.0-11.5) |
16.0 (9.0–28.0) | N/A | < 0.001 | 20.0 (13.0–31.0) |
5.0 (3.0-11.5) |
18.0 (10.0–28.0) |
N/A | < 0.001 |
Note: Boldface values, p < 0.05. Data are n (%) or median (IQR) unless otherwise stated
Abbreviations: IPSCC = International Pediatric Sepsis Consensus Conference. N/A = Not applicable
ap values are χ² or Mann–Whitney U between survivors and non-survivors
Diagnostic concordance between the two criteria
The population overlap between the two sepsis definitions is illustrated in the Venn diagram (Fig. 2). Only 489 (47.3%) patients met both sepsis criteria, while 166 (16.1%) met neither. Among SIRS-positive patients, 79.8% also met the PSC; among PSC-positive patients, 65.7% also met the SIRS criteria. The agreement between the two criteria was poor (kappa = 0.202, 95% CI: 0.143–0.261), which was expected given their different pathophysiological foundations (inflammation vs. organ dysfunction). The distribution of SIRS and Phoenix points among all patients is shown in the bubble plot (Fig. 3), demonstrating poor correlation between the two scoring systems (r = 0.272). Notably, respiratory dysfunction was present in 95.3% (709/744) of PSC-positive patients, reflecting the predominance of severe pneumonia in our PICU cohort (Table S1). Table S1 further presents detailed comparisons of all individual SIRS and Phoenix components, demonstrating statistically significant differences in all parameters between patients meeting and not meeting each definition.
Fig. 2.
Venn diagram for patients with suspected sepsis
Fig. 3.
Bubble plot of SIRS and Phoenix points in patients with suspected sepsis. Bubble area is proportional to the number of patients. The largest bubble (2 SIRS and 2 Phoenix points) represents 108 patients
Mortality and predictive performance
Overall 28-day mortality was 12.3% (127/1034). A total of 613 (59.3%) patients met the SIRS criteria with a mortality of 15.2% (93/613), and 744 (72.0%) met the Phoenix criteria with a mortality of 16.3% (121/744). Patients meeting both criteria had the highest mortality (18.6%), followed by those meeting only Phoenix criteria (11.8%) and those meeting only SIRS criteria (1.6%). Patients meeting neither criterion had a mortality of only 2.4%.
Mortality increased significantly with increasing SIRS and Phoenix points, as demonstrated by the Cochran-Armitage trend test (p < 0.001 for both). Figure 4A shows the distribution of patients and corresponding mortality by SIRS point, while Fig. 4B shows the same for Phoenix point. The receiver operating characteristic (ROC) curves comparing the discriminatory ability of the two criteria are presented in Fig. 5. The PSC demonstrated significantly superior discriminatory ability compared to the SIRS criteria (C-statistic = 0.809 vs. 0.589, p < 0.001). Combining both criteria into a single model did not improve predictive performance compared to using the PSC alone (C-statistic = 0.807 vs. 0.809, p = 0.750).
Fig. 4.
A, PICU 28-day mortality (bars) and number of patients (line) by number of SIRS points. B, PICU 28-day mortality (bars) and number of patients (line) by number of Phoenix points
Fig. 5.
Comparison of area under the receiver operating characteristic curves (AUROCs) to discriminate PICU 28-day mortality for SIRS points, Phoenix points and combined model
Consistent with our primary hypothesis, after adjusting for clinically relevant confounders (age, neoplastic disease, hematological disease, congenital heart disease, primary infection site, and septic shock), the PSC remained a strong independent predictor of 28-day mortality (adjusted OR = 5.123, 95% CI: 2.128–12.333, p < 0.001). In stark contrast, the SIRS criteria showed no independent predictive value after multivariable adjustment (adjusted OR = 0.937, 95% CI: 0.523–1.678, p = 0.827) (Table 2).
Table 2.
Univariate and multivariable binary logistic regression models for predictors of PICU 28-day mortality
| Variable | Unadjusted OR (95%CI) |
p-value | Adjusted OR (95%CI) |
p-value |
|---|---|---|---|---|
| SIRS criteria | 2.036 (1.345–3.080) | 0.001 | 0.937 (0.523–1.678) | 0.827 |
| Phoenix criteria | 9.193 (4.002–21.118) | < 0.001 | 5.123 (2.128–12.333) | < 0.001 |
| Age (months) | 1.000 (0.997–1.004) | 0.817 | 0.994 (0.990–0.999) | 0.009 |
| Neoplastic disease | 7.176 (3.198–16.103) | < 0.001 | 13.060 (5.122–33.296) | < 0.001 |
| Hematological disease | 3.015 (1.413–6.437) | 0.004 | 4.627 (1.917–11.169) | 0.001 |
| Congenital heart disease | 2.467 (1.467–4.148) | 0.001 | 2.508 (1.381–4.556) | 0.003 |
| Sepsis-SIRS shock | 3.879 (2.641–5.698) | < 0.001 | 2.284 (1.250–4.170) | 0.007 |
| Sepsis-Phoenix shock | 4.623 (3.146–6.793) | < 0.001 | 2.307 (1.389–3.832) | 0.001 |
| Pulmonary infection | 0.524 (0.357–0.770) | 0.001 | 1.093 (0.572–2.091) | 0.788 |
| Bloodstream infection | 2.680 (1.311–5.475) | 0.007 | 2.117 (0.807–5.555) | 0.127 |
| Intracranial infection | 2.196 (1.337–3.607) | 0.002 | 3.458 (1.595–7.495) | 0.002 |
Note: All predefined clinically relevant variables were entered simultaneously into the multivariable model using the enter method. bold values indicate p < 0.05
OR: Odds ratio; CI: Confidence interval; SIRS: Systemic inflammatory response syndrome
Component-level analysis showed that, with the exception of heart rate, all three other SIRS components were significantly associated with PICU 28-day mortality in univariate analysis (Table S2), whereas all four Phoenix components demonstrated significant univariate associations (Table S3). When all eight components were entered simultaneously into a single model to assess their relative prognostic importance, only the four Phoenix components retained independent associations with mortality (Table S4).
Phenotypic overlap characteristics
Notably, 34.3% of Sepsis-Phoenix patients did not meet the SIRS criteria. These patients had higher Phoenix-respiratory points (p < 0.001) but lower overall mortality (11.8% vs. 18.6%, p = 0.016) compared to patients meeting both criteria (Table S5).
Conversely, 20.2% of SIRS-positive patients were not classified as Sepsis-Phoenix. This subgroup had significantly lower Phoenix organ dysfunction scores (p < 0.001) and an extremely low mortality of 1.6%, compared to 18.6% in patients meeting both criteria (p < 0.001), reflecting higher specificity of the Phoenix criteria for identifying high-risk patients (Table S6).
Logistic regression analysis revealed that only body temperature and white blood cell count (the two mandatory SIRS components) were significantly associated with Sepsis-Phoenix status (Table S7), while only coagulation and neurological scores were significantly associated with SIRS positivity among Sepsis-Phoenix patients (Table S8).
Septic shock subgroup analysis
A total of 241 (23.3%) patients met the SIRS septic shock criteria (mortality: 25.7%) and 249 (24.1%) met the Phoenix septic shock criteria (mortality: 27.3%). The agreement between the two septic shock definitions was moderate (kappa = 0.561, 95% CI: 0.502–0.621) (Figure S1). In patients with SIRS-defined septic shock, an increase in SIRS points was not associated with increased mortality (Cochran-Armitage p = 0.511) (Fig. 6A). In contrast, in patients with Phoenix-defined septic shock, mortality increased significantly with rising Phoenix points (Cochran-Armitage p < 0.001) (Fig. 6B).
Fig. 6.
A, PICU 28-day mortality (bars) and number of patients with septic shock (line) by number of SIRS points. B, PICU 28-day mortality (bars) and number of patients with septic shock (line) by number of Phoenix points
Consistent with the main analysis, the Phoenix criteria demonstrated superior predictive performance for mortality in the septic shock subgroup compared to the SIRS criteria. In patients with SIRS-defined septic shock, the Phoenix components had a C-statistic of 0.819 (95% CI: 0.754–0.884) for predicting mortality, compared to 0.618 (95% CI: 0.539–0.698) for SIRS components (Tables S9 and S11). Similarly, in patients with Phoenix-defined septic shock, the Phoenix components had a C-statistic of 0.807 (95% CI: 0.744–0.869), compared to 0.609 (95% CI: 0.529–0.688) for SIRS components (Tables S10 and S12). Combining SIRS and Phoenix components did not significantly improve predictive performance compared to using Phoenix components alone in either subgroup (Tables S13 and S14).
Subgroup comparisons of septic shock patients are presented in Tables S15 and S16. Compared with Sepsis-Phoenix shock patients without Sepsis-SIRS shock, those with Sepsis-SIRS shock had higher Phoenix points (p = 0.001), but there was no statistically significant difference in PICU 28-day mortality between the two groups (p = 0.180) (Table S15). Compared with Sepsis-SIRS shock patients without Sepsis-Phoenix shock, those with Sepsis-Phoenix shock had higher Phoenix points (p < 0.001) and PICU 28-day mortality (30.1% vs. 16.7%, p = 0.026) (Table S16).
Further analysis showed that only cardiovascular and coagulation system scores were significantly correlated with Sepsis-SIRS shock (Table S17), while all four SIRS components were associated with Sepsis-Phoenix shock (Table S18). This subgroup analysis was exploratory and should be interpreted cautiously due to limited sample size.
Discussion
In this single-centre retrospective study of 1034 PICU patients with confirmed or suspected infection, we found that the Phoenix criteria identified a higher proportion of patients as septic (72.0%) compared to the SIRS criteria (59.3%), with poor diagnostic concordance between the two (kappa = 0.202). After adjusting for important confounding factors, the Phoenix criteria remained an independent strong predictor of 28-day mortality, while the SIRS criteria showed no independent predictive value. These findings support the use of the Phoenix criteria for risk stratification in critically ill children with infection.
The poor agreement between the two criteria was expected, as they are based on fundamentally different pathophysiological concepts: SIRS reflects systemic inflammation, while Phoenix focuses on life-threatening organ dysfunction. This difference in design explains why the Phoenix criteria have superior prognostic performance. Importantly, this does not represent circular reasoning. The Phoenix criteria were intentionally designed to incorporate life-threatening organ dysfunction, which is the immediate pathophysiological precursor to sepsis-related mortality. This represents a deliberate and evidence-based design advantage rather than a methodological limitation, as the primary objective of sepsis definitions is to accurately identify patients at highest risk of adverse outcomes who require urgent, targeted intervention.
A key finding of our study was that 20.2% of SIRS-positive patients were not classified as septic by the Phoenix criteria, and this subgroup had an extremely low mortality of 1.6%. This demonstrates that the Phoenix criteria effectively exclude patients with isolated inflammation but no significant organ dysfunction, thereby reducing overdiagnosis and unnecessary treatment. This is particularly important in the PICU setting, where resource allocation and treatment intensity must be guided by accurate risk assessment.
Another important phenotypic finding was that Sepsis-Phoenix patients who concurrently met the SIRS criteria had significantly lower Phoenix-respiratory scores than those who did not (p < 0.001) (Table S5). Furthermore, respiratory dysfunction was present in 95.3% (709/744) of all Sepsis-Phoenix patients in our cohort. This notably high prevalence aligns with the global epidemiology of paediatric critical illness, where respiratory dysfunction remains the most common organ dysfunction encountered in PICU settings [17–19]. In our institution, severe pneumonia is the leading indication for direct PICU admission, and this local epidemiological profile likely accounts for the higher sepsis incidence observed with the Phoenix criteria compared with the SIRS criteria. Specifically, a substantial proportion of children with severe pneumonia may fail to meet the SIRS definition, which mandates at least two abnormal inflammatory parameters including a mandatory abnormality in either temperature or white blood cell count. However, these same patients frequently have life-threatening respiratory dysfunction that meets the Phoenix sepsis criteria. This indicates that the Phoenix criteria are better able to identify critically ill children with isolated respiratory organ dysfunction who would have been missed by the older SIRS-based definition.
By subgroup analysis of septic shock, we found that the incidence (23.3% vs. 24.1%) and PICU 28-day mortality (62/241 vs. 68/249) were comparable between the Sepsis-SIRS shock criteria and Sepsis-Phoenix shock criteria. Furthermore, the agreement between the two septic shock phenotypes was moderate (kappa = 0.561, 95% CI: 0.502–0.621). Although further research is needed to explain this, it may be related to the diagnostic criteria for the two types of septic shock. Using Sepsis-SIRS shock patients as a subset of the Sepsis-Phoenix shock population, we found that not only were 32.4% of Sepsis-SIRS shock patients not classified as Sepsis-Phoenix shock, but the PICU 28-day mortality among these patients was slightly lower (16.7% vs. 30.1%, p = 0.026). This may limit the utility of using Sepsis-SIRS shock patients as a subset to study Sepsis-Phoenix shock population. Additionally, we observed a similar consistency: the Phoenix model demonstrated superior predictive ability for PICU 28-day mortality in patients with septic shock compared to the SIRS model.
Sepsis is recognized as a heterogeneous clinical syndrome and this heterogeneity contributes to the complexity and uncertainty surrounding its research [20]. Different syndromes or phenotypes may reflect distinct responses across organisms, prompting efforts to identify clinical and biological characteristics that classify patients into subtypes potentially more responsive to specific therapies [11, 20]. Although our study is similar to the Sanchez-Pinto et al. study [9] that confirmed that the Phoenix criteria was superior to the SIRS criteria in determining mortality in PICU patients with infected and suspected infection, we have found few studies evaluating overlap between Sepsis-SIRS and Sepsis-Phoenix populations [10]. In a study evaluating the application of a new pediatric sepsis definition in previously registered severe sepsis patients (n = 401), Kernan et al. [10] found that approximately two-thirds of patients previously diagnosed with severe sepsis based on SIRS criteria also met Phoenix criteria. Although the overlapping cohort had higher Phoenix organ dysfunction scores than the cohort meeting only SIRS criteria (p < 0.001), there was no statistically significant difference in PICU mortality between the two groups (12.6% vs. 7.6%, p = 0.17). However, their study was a retrospective analysis of historical cohorts. Due to missing data, the authors used a modified Phoenix sepsis score to define the study cohort and did not systematically explore the extent of overlap between populations identified by the two pediatric sepsis definitions. Currently, studies comparing different sepsis definitions [11, 16, 21–24] have primarily focused on adults.
Our study has several strengths. First, this was the first systematic investigation of the extent of overlap between populations defined by two pediatric sepsis criteria. Second, our center is the largest PICU center in Southwest China and our hospital’s big data center has established the SSDD containing extensive clinical data. However, this study also has certain limitations. First, this is a single-centre retrospective study conducted in a tertiary care children’s hospital in Southwest China. Our study population were patients admitted directly to the PICU with infected and suspected infection. Such patients often present with organ dysfunction and generally exhibit higher overall severity of illness, potentially introducing selection bias. While our large sample size enhances the internal validity of our findings, the results may not be fully generalizable to other regions with different epidemiological profiles or to lower-resource settings. Second, our study did not include patients with confirmed or suspected infection from settings such as general ward or emergency department, which may have limited the SIRS criteria’s ability to identify more sepsis cases. Third, our operational definition of confirmed or suspected infection, based on the temporal relationship between blood culture collection and antibiotic administration, may introduce misclassification bias. Fourth, we used PICU 28-day mortality rather than the in-hospital mortality endpoint used in the original Phoenix validation studies, which may reduce comparability with previous literature [16]. Fifth, we did not include an external validation cohort, and our results require confirmation in multicentre studies.
Conclusions
In conclusion, our findings demonstrate that the SIRS and Phoenix criteria identify partially overlapping populations with distinct risk stratification profiles and show poor diagnostic concordance. The Phoenix criteria exhibit superior independent predictive performance for PICU 28-day mortality and may therefore 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 Phoenix-defined populations. Future research should focus on multicentre prospective validation of the Phoenix criteria across different healthcare settings and geographical regions.
Supplementary Information
Below is the link to the electronic supplementary material.
Acknowledgements
None.
Abbreviations
- PICU
Pediatric intensive care unit
- IPSCC
International pediatric sepsis consensus conference
- SIRS
Systemic inflammatory response syndrome
- PSC
Phoenix Sepsis Criteria
- SCCM
Society of critical care medicine
- SSDD
Sepsis Specialized Disease Database
- IQR
Interquartile range
- OR
Odds ratio
- CI
Confidence interval
- ROC
Receiver operating characteristic
Authors’ contributions
All the authors were responsible for the content of this manuscript and approved to submit. LT and HY designed the study and revised critically the manuscript for important and intellectual content. LZ performed data collection, data analysis, and writing of the paper, SL participated in the revision of the paper, QD collected and organized clinical information, TT performed data extraction and retrieval of bibliographies, while JC analyzed clinical data.
Funding
This research was supported by The Key Projects for Technological Innovation and Application Development in Chongqing Municipality (CSTB2024TIAD-KPX0034). LT received this funding. The funding source had no role in the study design, data collection, data analysis, or writing of the report.
Data availability
The datasets generated and analyzed during the current study are available from the corresponding author on reasonable request.
Declarations
Ethics approval and consent to participate
This study was approved by the Institutional Review Board of the Children’s Hospital of Chongqing Medical University (IRB No. 2025367). This was a retrospective study based on the de-identified administrative database; therefore, informed consent was waived.
Consent for publication
Not applicable.
Competing interests
The authors declare that they have no competing interests. Neither the entire paper nor any part of its contents have been published or accepted by another journal. The paper has not been submitted in its entirety, or in part, to any other journal.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Contributor Information
Haiping Yang, Email: oyhp0708@163.com.
Liping Tan, Email: tanlp0825@hotmail.com.
References
- 1.IHME Pathogen Core Group. Global burden associated with 85 pathogens in 2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet Infect Dis. 2024;24(8):868–95. https://pubmed.ncbi.nlm.nih.gov/38640940/. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.GBD 2019 Child and Adolescent Communicable Disease Collaborators. The unfinished agenda of communicable diseases among children and adolescents before the COVID-19 pandemic, 1990–2019: a systematic analysis of the Global Burden of Disease Study 2019. Lancet. 2023;402(10398):313–35. https://pubmed.ncbi.nlm.nih.gov/37393924/. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Menon K, Schlapbach LJ, Akech S, et al. Criteria for Pediatric Sepsis-A Systematic Review and Meta-Analysis by the Pediatric Sepsis Definition Taskforce. Crit Care Med. 2022;50(1):21–36. https://pubmed.ncbi.nlm.nih.gov/34612847/. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Shamavu GK, Mohamoud F. Rethinking Pediatric Sepsis and Septic Shock: Beyond International Consensus Criteria. Pediatr Health Med Ther. 2025;16:61–5. https://pubmed.ncbi.nlm.nih.gov/40065805/. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Lanziotti VS, Ventura A, Kache S, et al. New Phoenix criteria for pediatric sepsis and septic shock: the strengths and the future of a comprehensive perspective. Crit Care Sci. 2024;36:e20240058en. https://pubmed.ncbi.nlm.nih.gov/39046059/. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Goldstein B, Giroir B, Randolph A. International pediatric sepsis consensus conference: definitions for sepsis and organ dysfunction in pediatrics. Pediatr Crit Care Med. 2005;6(1):2–8. https://pubmed.ncbi.nlm.nih.gov/15636651/. [DOI] [PubMed] [Google Scholar]
- 7.Sanchez-Pinto LN, Bennett TD, DeWitt PE, et al. Development and Validation of the Phoenix Criteria for Pediatric Sepsis and Septic Shock. JAMA. 2024;331(8):675–86. https://pubmed.ncbi.nlm.nih.gov/38245897/. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Schlapbach LJ, Watson RS, Sorce LR, et al. International Consensus Criteria for Pediatric Sepsis and Septic Shock. JAMA. 2024;331(8):665–74. https://pubmed.ncbi.nlm.nih.gov/38245889/. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Sanchez-Pinto LN, Daniels LA, Atreya M, et al. Phoenix Sepsis Criteria in Critically Ill Children: Retrospective Validation Using a United States Nine-Center Dataset, 2012–2018. Pediatr Crit Care Med. 2025;26(2):e155–65. https://pubmed.ncbi.nlm.nih.gov/39982153/. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Kernan KF, Shaik M, Horvat CM, et al. Application of New Pediatric Sepsis Definition to a Multicenter Observational Cohort of Previously Enrolled Severe Sepsis Patients Defined by SIRS Plus Organ Dysfunction. J Intensive Care Med. 2025;40(12):1260–8. https://pubmed.ncbi.nlm.nih.gov/40518943/. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Engoren M, Seelhammer T, Freundlich RE, et al. A Comparison of Sepsis-2 (Systemic Inflammatory Response Syndrome Based) to Sepsis-3 (Sequential Organ Failure Assessment Based) Definitions-A Multicenter Retrospective Study. Crit Care Med. 2020;48(9):1258–64. https://pubmed.ncbi.nlm.nih.gov/32618690/. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Seymour CW, Liu VX, Iwashyna TJ, et al. Assessment of Clinical Criteria for Sepsis: For the Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). JAMA. 2016;315(8):762–74. https://pubmed.ncbi.nlm.nih.gov/26903335/. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Wösten-van Asperen RM, la Roi-Teeuw HM, Tissing WJEA, et al. The Phoenix Sepsis Score in Pediatric Oncology Patients With Sepsis at PICU Admission: Test of Performance in a European Multicenter Cohort, 2018–2020. Pediatr Crit Care Med. 2025;26(2):e177–85. https://pubmed.ncbi.nlm.nih.gov/39982155/. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Dowell JC, Parvathaneni K, Thomas NJ, et al. Epidemiology of Cause of Death in Pediatric Acute Respiratory Distress Syndrome. Crit Care Med. 2018;46(11):1811–9. https://pubmed.ncbi.nlm.nih.gov/30095498/. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Zhou L, Li S, Tang T, et al. A single-center PICU present status survey of pediatric sepsis-related acute respiratory distress syndrome. Pediatr Pulmonol. 2022;57(9):2003–11. https://pubmed.ncbi.nlm.nih.gov/35475331/. [DOI] [PubMed] [Google Scholar]
- 16.Cheng B, Li Z, Wang J, et al. Comparison of the performance between sepsis-1 and sepsis-3 in ICUs in China: a retrospective multicenter study. Shock. 2017;48(3):301–6. https://pubmed.ncbi.nlm.nih.gov/28448400/. [DOI] [PMC free article] [PubMed]
- 17.Sanchez-Pinto LN, Bembea MM, Farris RW, et al. Patterns of Organ Dysfunction in Critically Ill Children Based on PODIUM Criteria. Pediatrics. 2022;149(1 Suppl 1):S103–10. https://pubmed.ncbi.nlm.nih.gov/34970678/. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Egbuta C, Easley RB. Update on ventilation management in the Pediatric Intensive Care Unit. Paediatr Anaesth. 2022;32(2):354–62. https://pubmed.ncbi.nlm.nih.gov/34882910/. [DOI] [PubMed] [Google Scholar]
- 19.Kortz TB, Holloway A, Agulnik A, et al. Prevalence, aetiology, and hospital outcomes of paediatric acute critical illness in resource-constrained settings (Global PARITY): a multicentre, international, point prevalence and prospective cohort study. Lancet Glob Health. 2025;13(2):e212–21. https://pubmed.ncbi.nlm.nih.gov/39890223/. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Meyer NJ, Prescott HC. Sepsis and Septic Shock. N Engl J Med. 2024;391(22):2133–46. https://pubmed.ncbi.nlm.nih.gov/39774315/. [DOI] [PubMed] [Google Scholar]
- 21.Fullerton JN, Thompson K, Shetty A, et al. New sepsis definition changes incidence of sepsis in the intensive care unit. Crit Care Resusc. 2017;19(1):9–13. https://pubmed.ncbi.nlm.nih.gov/28215126/. [PubMed] [Google Scholar]
- 22.Takauji S, Hayakawa M, Fujita S. A Nationwide Comparison Between Sepsis-2 and Sepsis-3 Definition in Japan. J Intensive Care Med. 2020;35(12):1389–95. https://pubmed.ncbi.nlm.nih.gov/30636495/. [DOI] [PubMed] [Google Scholar]
- 23.Khwannimit B, Bhurayanontachai R, Vattanavanit V. Comparison of the performance of SOFA, qSOFA and SIRS for predicting mortality and organ failure among sepsis patients admitted to the intensive care unit in a middle-income country. J Crit Care. 2018;44:156–60. https://pubmed.ncbi.nlm.nih.gov/29127841/. [DOI] [PubMed] [Google Scholar]
- 24.Shankar-Hari M, Harrison DA, Rubenfeld GD, et al. Epidemiology of sepsis and septic shock in critical care units: comparison between sepsis-2 and sepsis-3 populations using a national critical care database. Br J Anaesth. 2017;119(4):626–36. https://pubmed.ncbi.nlm.nih.gov/29121281/. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The datasets generated and analyzed during the current study are available from the corresponding author on reasonable request.