Retrospective Pediatric Study Finds Divergent PICU Stays at Equal Phoenix Scores

Children with sepsis who received the same total Phoenix Sepsis Score but had different organ-system involvement profiles experienced markedly different stays in the pediatric intensive care unit in a retrospective study from China. Among 107 patients admitted to one PICU in 2024 and 2025, 29 with respiratory involvement alone and 59 with combined respiratory and coagulation involvement both had a median Phoenix score of 2, yet their median PICU stays were 10 and 6 days, respectively. The difference was associated with invasive mechanical ventilation, which was required by every child in the respiratory-only group but by 25.4% of those in the respiratory-coagulation group.

The finding indicates that the aggregate Phoenix score may not fully describe a child’s clinical course or intensive-care resource needs in a respiratory infection-heavy cohort. Respiratory and coagulation dysfunction accounted for 89.7% of Phoenix score positivity, while the respiratory-coagulation combination alone represented 55.1% of all 107 cases. The researchers therefore conducted an exploratory comparison of the two most common profiles, both of which met the same median total score.

Eligible patients were 29 days to 18 years old, met a Phoenix Sepsis Score of at least 2 within 24 hours of admission, and were admitted consecutively to the PICU at Sichuan Provincial Maternity and Child Health Hospital between January 2024 and December 2025. Children were excluded if records were incomplete, they died or were discharged within 24 hours, they were postoperative, or a primary hematologic disorder affected coagulation. The Phoenix score combines respiratory, coagulation, circulatory, and neurological components, with a total of 2 or more meeting the sepsis threshold under the 2024 criteria used in the analysis.

The institutional ethics committee approved the chart review and waived informed consent. Respiratory scoring used oxygenation indices, including the ratio of arterial oxygen partial pressure to inspired oxygen fraction or, when oxygen saturation was no more than 97%, the ratio of oxygen saturation to inspired oxygen fraction, together with invasive ventilation. Coagulation scoring considered platelet count, international normalized ratio, D-dimer, and fibrinogen, with one point assigned for each abnormal measure and a maximum of two points.

The circulatory component combined vasoactive-agent use, lactate, and age-adjusted mean arterial pressure, while neurological scoring used the Glasgow Coma Scale and pupillary response. The most abnormal values recorded during the first 24 hours were used to calculate each subsystem score, and the sum provided the total Phoenix Sepsis Score. Patients were grouped according to the number of subsystems scoring at least 1: 37 had single-system involvement, 62 had dual-system involvement, and eight had involvement of three or more systems.

The single-system group comprised 29 children with respiratory dysfunction alone and eight with coagulation dysfunction alone. Among dual-system cases, 59 involved respiratory and coagulation systems, two involved coagulation and circulation, and one involved respiration and circulation. The eight children with multi-system involvement had the highest median total Phoenix score, 4.5, a median lactate level of 5.30 mmol/L, and mortality of 37.5%.

No deaths were recorded in either the single-system or dual-system groups. Severity did not translate into a simple ordering of time spent in intensive care. The single-system group had the longest median PICU stay at 10 days, followed by six days for dual-system disease and 4.5 days for multi-system disease.

Median total hospital stay followed the same pattern at 13, 10, and 8.5 days, respectively. The shorter stays among the most severely affected children are an observed feature of this cohort and do not demonstrate that broader organ dysfunction reduces the need for care. Across the three groups, the multi-system cases also had higher total Phoenix and pediatric Sequential Organ Failure Assessment scores, elevated lactate and liver enzymes, and more diverse infection sites, while the single-system cases had the greatest use and duration of mechanical ventilation.

The focused comparison showed that the isolated-respiratory group had a median PICU stay of 10 days, with an interquartile range of 9 to 12, compared with six days, ranging from four to eight, in the respiratory-coagulation group. Prolonged stay, defined as at least 10 days and corresponding to the cohort’s 75th percentile, occurred in 65.5% of the isolated-respiratory group and 18.6% of the respiratory-coagulation group. In a model adjusted for admission lactate and age, the isolated-respiratory pattern was associated with prolonged stay with an odds ratio of 6.92 and a 95% confidence interval of 2.48 to 20.80.

Lactate and age were not significant factors in that model. The subgroups also had different pSOFA scores, with medians of 3 and 2, respectively, further showing that an identical Phoenix total did not imply identical profiles of organ dysfunction. Adding invasive mechanical ventilation to the model changed the subgroup association from an odds ratio of 6.92 to 1.27, with a 95% confidence interval of 0.30 to 5.07 and P = 0.739.

Ventilation itself was strongly associated with prolonged stay, with an odds ratio of 18.80, a wide 95% confidence interval of 3.78 to 123.24, and P below 0.001. Median ventilation duration was six days in the isolated-respiratory group and zero days in the respiratory-coagulation group. A sensitivity analysis treating length of stay as a continuous variable produced a similar pattern: the subgroup association weakened after adjustment for ventilation, while ventilation remained associated with longer stay.

These findings do not establish a formal mediation pathway because invasive ventilation is itself part of the Phoenix respiratory subscore, meaning the predictor and subgroup classification share information. Several limitations constrain how far the results can be generalized. The study was retrospective and single-center, and the multi-system group included only eight children, making estimates for that group particularly uncertain.

PICU length of stay also depends on discharge practices, bed availability, and local organizational factors that were not measured. Phoenix scores were assessed only during the first 24 hours rather than repeatedly as illness evolved, and the prolonged-stay threshold was derived from this cohort rather than a universal pediatric standard. The analysis did not include the renal and hepatic dimensions of the extended Phoenix-8 criteria, and unmeasured differences in severity may remain.

The results support further prospective, multicenter assessment of whether the respiratory subscore consistently provides useful information beyond the total score for estimating intensive-care duration and resource demand, rather than establishing a clinical prediction rule on their own.

Subject of Research: Phoenix Sepsis Score subsystem profiles, invasive mechanical ventilation, and pediatric intensive-care length of stay in children with sepsis

Article Title: Divergent PICU lengths of stay despite identical phoenix sepsis scores: heterogeneity in pediatric sepsis outcomes according to system involvement profiles

Article References: Jiang, Z., Wu, L., Zhou, X., Xu, J., Xiong, X., Chen, J., & Hu, Y. (2026). Divergent PICU lengths of stay despite identical phoenix sepsis scores: heterogeneity in pediatric sepsis outcomes according to system involvement profiles. Frontiers in Pediatrics, 14, Article 1986840. https://doi.org/10.3389/fped.2026.1986840

Image Credits: AI Generated

DOI: 10.3389/fped.2026.1986840

Keywords: pediatric sepsis, Phoenix Sepsis Score, PICU length of stay, invasive mechanical ventilation, respiratory failure, coagulation dysfunction, organ-system involvement, critical-care resources, retrospective cohort, pediatric intensive care, Divergent, PICU