Editorial 

Surveillance of Sepsis in Children—Making Every Case Count

Luregn J. Schlapbach, Shu-Ling Chong, Nora Luethi

JAMA Published Online: March 22, 2026

doi: 10.1001/jama.2026.2814

Sepsis is a leading cause of hospital admissions, mortality, and health care expenditure worldwide. It disproportionally affects very young and very old individuals. Sepsis in young individuals carries a magnified societal impact through potential years of life lost and reductions in quality-adjusted life-years. Sepsis increases the risk of functional impairment that may persist for decades, translating into sustained health care needs and reduced lifetime economic productivity. While many countries maintain public surveillance systems for common infectious diseases such as influenza, the absence of systematic sepsis surveillance obscures the full societal burden of disease, limits benchmarking of care, and hinders effective public health measures.

To date, most estimates of pediatric sepsis incidence rely on smaller research databases¹ or models derived from administrative diagnostic codes, which have been shown to be variably applied, resulting in a number of limitations and biases.² In adults, the 2016 Sepsis-3 criteria³ shifted focus to objective measures of organ dysfunction rather than systemic inflammatory response syndrome criteria, enabling more robust and reproducible estimates of sepsis incidence, severity, and outcomes. Building on clinical criteria in Sepsis-3, the Centers for Disease Control and Prevention (CDC) developed the Adult Sepsis Event definition, a pragmatic approach for standardized extraction of sepsis cases from electronic health records (Table).⁴

Table.  Comparison of Pediatric vs Adult Sepsis Criteria, Their Components, and the According Sepsis Event Operationalization

Criteria	Phoenix	Pediatric Sepsis Event (PSE)	Sepsis-3	Adult Sepsis Event (ASE)
Age group	Pediatric	Pediatric	Adult	Adult
Purpose	Data-driven, globally applicable consensus criteria for sepsis and septic shock in children	Adapted Phoenix criteria for surveillance optimized for extraction from digital health care data	Data-driven consensus criteria for sepsis and septic shock in adults	Adapted Sepsis-3 criteria for surveillance optimized for extraction from routine digital health care data
Infection	Receipt of systemic antimicrobials and microbiological testing within 24 hours of presentation	Blood culture collection (or hospital transfer) and ≥4 QADs starting within ±2 calendar days of blood culture	Combination of oral or parenteral antibiotics and body fluid cultures within 24 hours	Blood culture collection and ≥4 QADs starting within ±2 calendar days of blood culture
Sepsis	Phoenix sepsis score ≥2 points in a child with suspected/confirmed infection	≥2 modified Phoenix organ dysfunction points within ±2 days of the blood culture plus ≥4 QADsCommunity-onset sepsis: any criterion met prior to hospital day 3Hospital-onset sepsis: all criteria met on hospital day 3 or later	Sequential Organ Failure Assessment score ≥2 points in an adult with suspected/confirmed infection	Infection plus organ dysfunction; ≥1 organ-specific criteria met within ±2 calendar days of blood culture plus ≥4 QADs
Septic shock	Sepsis in presence of cardiovascular dysfunction (cardiovascular Phoenix sepsis score component ≥1 point)	Sepsis with ≥1 cardiovascular point	Sepsis in presence of arterial hypotension, lactate >2 mmol/L, and treatment with vasopressors	Not specified
Respiratory dysfunction	Pao2/Fio2 ratio or Spo2/Fio2 ratio cutoffs and use of respiratory support (maximum 3 points by increasing severity)	New noninvasive or invasive ventilation (maximum 2 points by increasing severity)	Pao2/Fio2 ratio and use of respiratory support (maximum 4 points by increasing severity)	New invasive mechanical ventilation
Cardiovascular dysfunction	Vasoactive medications, lactate, hypotension by age-specific MAP (maximum 6 points by increasing severity)	Vasoactive medications, lactate, hypotension by age-specific MAP (maximum 6 points by increasing severity)	Hypotension (MAP) and vasoactive medication (maximum 4 points by increasing severity)	Vasoactive medication
Coagulation dysfunction	Platelet count <100 × 103/μL, INR >1.3, D-dimer >2 mg/L, fibrinogen <100 mg/dL (maximum 2 points by increasing severity)	Platelet count <100 × 103/μL and ≥50% decline from baseline, INR >1.3, D-dimer >2 mg/L, fibrinogen <100 mg/dL (maximum 2 points by increasing severity)	Platelet count (maximum 4 points by increasing severity)	Platelet count <100 × 103/μL and ≥50% decline from baseline
Neurologic dysfunction	GCS score ≤10, reactivity of pupils (maximum 2 points by increasing severity)	GCS score ≤10 (or ICD-10–based proxy in datasets without GCS) (maximum 1 point)	GCS score (maximum 4 points by increasing severity)	Not available
Kidney dysfunction	Not available	Not available	Creatinine or urine output (maximum 4 points by increasing severity)	Doubling of serum creatinine or decrease by ≥50% of estimated glomerular filtration rate relative to baseline
Hepatic dysfunction	Not available	Not available	Bilirubin (maximum 4 points by increasing severity)	Total bilirubin ≥20 mg/dL and increase by 100% from baseline

Among children, there is no comparable construct, a gap that was until recently exacerbated by outdated sepsis criteria from 2005.⁵ Current estimates of the global burden of pediatric sepsis derive from diagnostic codes and yield substantially diverging results: The 2017 Global Burden of Disease study² estimated that children accounted for 25.2 million (52%) of 48.9 million incident sepsis cases worldwide and 3.35 million (30%) of 11.0 million sepsis-related deaths, corresponding to a case-fatality rate of 13.3%. The subsequent 2021 Global Burden of Disease study,⁶ which included implicit codes for infection and presumed organ dysfunction, reported more than a 2-fold higher incidence of pediatric sepsis but a substantially lower case-fatality rate of 5.65%. The methodological differences in these studies undermine comparability, clarity of messaging, and, ultimately, the strength of recommendations for policy change.

Now, with the updated Phoenix criteria for sepsis and septic shock in children,^7,8 clinicians, researchers, and stakeholders worldwide have data-driven criteria at hand, developed and validated across more than 3.5 million pediatric health care encounters, incorporating data from US, Latin American, African, and Asian health care settings. Pediatric sepsis is defined as confirmed or suspected infection in the presence of a Phoenix sepsis score of 2 or higher, which integrates up to 14 clinical parameters to quantify respiratory, cardiovascular, neurologic, and coagulation dysfunction. Although the score incorporates internal redundancy, permitting calculation with a much smaller subset of easily obtainable clinical parameters, not all data elements are consistently captured in structured fields available in the electronic health record, potentially limiting scalable case ascertainment for robust institution- and population-level surveillance of pediatric sepsis.

In this issue of JAMA, Rhee and colleagues⁹ present the first CDC framework for sepsis in children, optimized for routine extraction from the electronic health record. An interdisciplinary team of experts in pediatric intensive care, emergency, and infectious diseases, data science, and public health from the US developed criteria for Pediatric Sepsis Events (PSEs) derived from the Phoenix criteria. Using the criteria for the PSEs, the authors then estimated the incidence, mortality, and temporal trends of pediatric sepsis in US hospitals. The authors combined 2 US large electronic health record datasets as well as secondary datasets, yielding a combined 3 925 809 hospital encounters of children aged 1 month to 17 years across approximately 400 specialized pediatric academic and general health care systems, the majority contributing data from 2016 to 2023.

The PSE definition was constructed through an iterative process modeled on the adult approach, adapting the Phoenix criteria to facilitate extraction from the electronic health record. To enhance feasibility while preserving construct validity, respiratory dysfunction was defined by use of invasive or noninvasive ventilation rather than gas-exchange metrics; pupillary response was omitted; International Statistical Classification of Diseases and Related Health Problems, Tenth Revision codes were permitted to identify neurologic dysfunction in the absence of Glasgow Coma Scale data; and new relative declines in platelet counts were defined. A manual health record review of 581 cases in 3 tertiary pediatric academic centers served to construct a physician-adjudicated sepsis case reference standard, against which the PSE criteria demonstrated a sensitivity of 69.9%, a specificity of 93.1%, a positive predictive value of 70.7%, and a negative predictive value of 92.9%, outperforming administrative code–based case procedures. When applied to the large electronic health record datasets, the authors report an average hospital incidence of 1.3% for pediatric sepsis (72.6% community onset) and 0.8% for septic shock. In-hospital mortality was 4.0% for nonshock sepsis, 14.0% for septic shock, and 10.1% overall. Accordingly, the surveillance identifies more than 18 000 children with sepsis in the US each year, associated with more than 1800 in-hospital deaths—approximately 1 in 6 pediatric hospital deaths nationally. Epidemiological patterns were consistent across datasets, and the investigators did not observe a significant decline in incidence or mortality in recent years. Corroborating the substantial impact on health care utilization, children with sepsis had severalfold higher rates of intensive care unit admission and longer hospital stays.

The investigators should be congratulated for synthesizing and reporting the first national estimates of nonneonatal pediatric sepsis incidence through a scalable approach built on a Phoenix-derived standardized PSE framework. The Hospital Toolkit for Pediatric Sepsis Event Surveillance and the open-source code provided by the authors represent timely and much needed additions to the CDC’s Sepsis Core Elements.¹⁰ Specifically, they will enable standardized retrospective identification of pediatric sepsis cases, prospective tracking of community- and hospital-acquired sepsis epidemiology similar to other infectious disease threats, assessment of the impact of sepsis quality improvement initiatives,¹¹ as well as reporting and benchmarking of outcome data for stakeholders, policy, and planning. The PSE criteria are conceptually harmonized with the CDC’s Adult Sepsis Event criteria, facilitating the implementation of sepsis surveillance across age groups, which is of particular relevance for facilities caring for both pediatric and adult patients and for the population of patients transitioning from adolescence to young adulthood. Vendors of electronic health records should provide standardized PSE documentation templates to enhance scalability and interoperability of digital sepsis surveillance across health care systems and national borders.¹²

Several limitations merit consideration for the interpretation of PSE-based estimates and may inform future advancements in the field. First, the design of the PSE definition removed clinical components of neurologic dysfunction or made them optional, perhaps introducing bias in capturing sepsis-associated encephalopathy, which is a well-recognized manifestation of sepsis-related organ failure.¹³ Therefore, it would be desirable to enhance structured collection of the Glasgow Coma Scale score, as well as pupillary reaction, within electronic health record fields. Second, the PSE framework does not capture sepsis in newborns, a highly vulnerable population that accounted for nearly half of sepsis cases in prior pediatric estimates, but in whom optimal characterization of sepsis-related organ dysfunction remains a topic of ongoing debate. Third, the modifications for the PSE substantially reduced the sensitivity of capturing sepsis cases, for example, by omitting milder respiratory dysfunction. Importantly, the Phoenix criteria, developed using the area under the precision recall curve, had already given considerable weight to specificity in capturing life-threatening disease associated with higher mortality as outcome. Children meeting PSE criteria thus constitute a severe clinical phenotype with substantial mortality, and PSE-based surveillance will substantially underestimate the true disease burden by 30% or more. While the validation of PSE criteria against manually curated case reviews seems justified from the perspective of face validity, interclinician agreement can be taxing, and the performance of the PSE criteria in terms of sensitivity will also need to be compared against the consensus gold standard set by the Phoenix criteria. Fourth, clinical decision support systems for early recognition and treatment of children with suspected sepsis remain a priority, and their development may leverage the Phoenix criteria as objective outcome measures.^14,15 Finally, independent validation of the PSE framework in international settings will be necessary, given differences in epidemiology, comorbidity patterns,¹⁶ and levels of health care digitization. Such validations could inform context-specific adaptations, for example, to enable future European Centre for Disease Prevention and Control pediatric sepsis surveillance.

In summary, the PSE definition created by Rhee and colleagues sets a pragmatic standard to facilitate electronic health record–based surveillance of pediatric sepsis in the US. One of the leading causes of death as well as short- and long-term morbidity in children can thereby be robustly monitored for public health. This should spur advances in quality improvement and research both in the US and internationally to reduce the excessive burden of sepsis on child health.