- Design
- Systematic review and meta-analysis of prognostic accuracy studies
- Population
- 15 studies, 16 cohorts, about 2.6 million paediatric encounters with suspected infection
- Primary outcome
- In-hospital mortality
- Effect
- AUC 0.81 (95% CI 0.75 to 0.84) vs 0.71 for older criteria; specificity 99% in ED vs 48% in ICU
The Phoenix sepsis criteria replaced the 2005 International Pediatric Sepsis Consensus Conference definition in 2024. This meta-analysis pooled 15 studies (16 cohorts, about 2.6 million encounters) of children with suspected infection.
Phoenix predicted in-hospital death with sensitivity 77%, specificity 73% and an area under the curve of 0.81, against 0.71 for the older criteria (difference 0.10, 95% CI 0.04 to 0.19). Specificity varied sharply by setting: 99% in emergency department cohorts, 48% in intensive care, where most children are already sick.
The authors stress that Phoenix is a tool for identifying children at high risk of death, not a screening tool to catch early sepsis. In the emergency department, it should sit alongside, not replace, clinical judgement and local early-warning tools.
- Use the Phoenix criteria to identify children with suspected infection at high risk of death.
- Do not rely on Phoenix to screen for early sepsis; it was not designed for that.
- Keep using clinical judgement and early-warning scores to start antibiotics promptly.
- Expect fewer false alarms in the emergency department than in intensive care.
Why it matters
Phoenix is now the paediatric sepsis definition, and this pools the evidence on how well it predicts death.
Don't overread it
Mostly retrospective and high-income data; performance in lower-resource emergency departments is still uncertain.
The statistics, in plain English
An area under the curve of 0.81 means the criteria ranked a child who died above one who survived about 8 times in 10. The very different specificity by setting shows that the same test behaves differently depending on how sick the children being tested already are.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for emergency & critical care, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free