The Phoenix Sepsis Score
A significant development in pediatric critical care, we unpack the 2024 study that brought us the tool.
Journals: JAMA (55.0), Critical Care Explorations (2.7), Critical Care Medicine (7.6)
Article Years: 2024, 2020, 2022
In developing what would become the landmark 2024 Phoenix Sepsis Score, Schlapbach et al had three overarching goals: (1) acquire large amounts of EHR data that may accurately portray the international outcomes of sepsis, (2) conduct a systematic review and meta-analysis of published literature on sepsis outcomes, and (3) organize a group of experts to interpret these findings as they rolled out, follow a modified Delphi process, and arrive at agreement on new sepsis criteria. In other words, Research Goals 1 and 2 use two different sources of data to identify common features of infection clinical course leading to bad outcomes, and Research Goal 3 uses 1 and 2’s extrapolations to delineate the ways clinicians can identify it in the future. Diving into it:
Research Goal 1: Researchers wanted to develop guidelines with international reach, so to ensure their guideline development considered the realities of practicing in developed and developing communities, they sought multisite, EHR data from both high and low resource settings—the United States represented high resource healthcare and Bangladesh, China, Colombia, and Kenya represented low resource healthcare. Researchers identified infection by searching for cases of children who during their presentation received systemic antimicrobials and undergoing microbiological testing. Then, in these children, researchers used clinical and laboratory markers from the first 24 hours of presentation to calculate 8 organ-specific subscores of dysfunction (e.g., renal and cardiovascular). They also calculated qSOFA scores and IPSCC sepsis calculations and took these numbers for further evaluation by the panel in Research Goal 3. The study included data from over 3 million hospital encounters of patients younger than 18 years across various inpatient and ambulatory hospital locations. The study excluded data from birth hospitalizations & children whose postconceptional age was younger than 37 weeks. Their findings are as follows:
Suspected infection was associated with .7% and 3.6% mortality within the first 24-hours in high and low resource settings, respectively. (Note: suspected infection means infection of any severity, ranging from uncomplicated rhinovirus URI to disseminated meningococcemia—there’s no delineation yet.)
These numbers shoot up to 7.1% and 28.5% if the child scores ≥2 per Research Goal 3’s newly developed Phoenix criteria— ≥2 means the patient meets criteria for sepsis. (Note: Research Goals 1-3 occurred with chronologic overlap leading up to the paper’s publication.)
Sepsis with remote organ dysfunction is associated with different risks than sepsis affecting one organ system: mortality rates of 8.0% were observed in high resource settings compared to 32.3% in low resource settings.
Finally, these numbers rise again to 10.8 and 33.5% if the patient meets criteria for septic shock. Septic shock is defined as suspected/confirmed infection AND ≥1 point(s) earned via cardiovascular abnormalities.
This data demonstrated a strong association between Phoenix scores ≥2 and mortality; upon a vote of approval by the Delphi process panel (Research Goal 3), this is the cutoff we observe today.
Research Goal 2: The systematic review protocol can be found here. Essentially, researchers searched on Medline MeSH terms such as “sepsis,” “septicemia,” and “septic shock” (n=1243) before hand-combing through the papers and tossing out research that was either a duplicate, was at high risk for bias per QUIPS, or did not meet inclusion criteria. The most frequent reasons for excluding articles included they were not about sepsis (n=204), covered the wrong age range (n=151), or had no comparator group (n=130). They ended up with 81 papers to work with for meta-analysis which used pooled estimates to determine mortality risks based on multiple patient baseline characteristics, clinical characteristics, laboratory values, and organ dysfunction/illness severity scores.
Pooled estimates are represented as either positive or negative integers. Positive values mean that the higher the value, the more at risk a person is from dying. The 3 most positive pooled estimates from this meta-analysis are as follows:
Vasoactive inotropic score (VIS): 23.5 [3.4, 43.6]
Creatinine: 13.0 [4.6, 21.5]
PIM-2: 12.2 [9.3, 14.9]
Negative values mean that the higher the value, the less at risk a person is from dying. The 3 most negative pooled estimates from this meta-analysis are as follows:
Platelet count: -87 [-107, -67]
Albumin: -4.3 [-8.4, -0.2]
GCS: -4.0 [-6.2, -1.8]
Research Goal 3: Matters handled by the expert panel are detailed here. The expert panel comprised 35 nurse and physician experts hailing from Australia, Bangladesh, Brazil, Canada, France, India, Italy, Japan, Switzerland, South Africa, United Kingdom, and the United States with backgrounds in intensive care, emergency medicine, infectious diseases, general pediatrics, informatics, nursing, and neonatology. Perhaps superfluous to note, but we think the variety of disciplines is a strength of this study. A modified Delphi process was used to produce consensus responses to a variety of questions pertinent to study design, such as the definition of sepsis and inclusion/exclusion criteria.
The panel of experts considered outcomes calculated in Research Goals 1 and 2 before voting on matters such as which organ dysfunction criteria might best aid future clinicians using Phoenix criteria to predict sepsis and septic shock outcomes. Notably, the final model, which calculated levels of dysfunction for 4 organ systems—cardiovascular, respiratory, neurological, and coagulation—had comparable performance with scores generated from an 8-organ system model that also included renal, hepatic, endocrine, and immunological dysfunction (Phoenix-8 Score), so the panel reached consensus on simplifying Phoenix criteria to the four-organ system model of today.
The Spin: The introduction of the Phoenix Sepsis Score (PSS) represents the official change in the way clinicians think about sepsis: it officially abandoned the overly sensitive, non-specific SIRS criteria, which defined sepsis as just infection plus inflammation and was thus not helpful for differentiating serious infection from benign illness in kids. It refocused the entire definition on what actually makes sepsis deadly: life-threatening organ dysfunction caused by a dysregulated host response to that infection. This change is now in line with the way the pathology is viewed in adults due to Sepsis-3 and provides a much more specific and standardized way to identify patients at high risk of death.
It is important to underscore: PSS DOES NOT SCREEN FOR SEPSIS—it merely identifies it at any point in the disease course, perhaps early in disease course, but clinicians should not rely upon it in this way. (Research is still underway for a good screening method for sepsis, a topic we will discuss in a future article.) Instead, PSS is more reliable for matters such as research which can calculate PSS and in a standardized fashion to identify cases of sepsis and therefore identify increased risk for mortality.
To read more: Menon, K., Schlapbach, L. J., Akech, S., Argent, A., Biban, P., Carrol, E. D., Chiotos, K., Jobayer Chisti, M., Evans, I. V. R., Inwald, D. P., Ishimine, P., Kissoon, N., Lodha, R., Nadel, S., Oliveira, C. F., Peters, M., Sadeghirad, B., Scott, H. F., de Souza, D. C., Tissieres, P., … Pediatric Sepsis Definition Taskforce of the Society of Critical Care Medicine (2022). Criteria for Pediatric Sepsis-A Systematic Review and Meta-Analysis by the Pediatric Sepsis Definition Taskforce. Critical care medicine, 50(1), 21–36. https://doi.org/10.1097/CCM.0000000000005294
Menon, K., Schlapbach, L. J., Akech, S., Argent, A., Chiotos, K., Chisti, M. J., Hamid, J., Ishimine, P., Kissoon, N., Lodha, R., Oliveira, C. F., Peters, M., Tissieres, P., Watson, R. S., Wiens, M. O., Wynn, J. L., & Sorce, L. R. (2020). Pediatric Sepsis Definition-A Systematic Review Protocol by the Pediatric Sepsis Definition Taskforce. Critical care explorations, 2(6), e0123. https://doi.org/10.1097/CCE.0000000000000123
Schlapbach, L. J., Watson, R. S., … Society of Critical Care Medicine Pediatric Sepsis Definition Task Force (2024). International Consensus Criteria for Pediatric Sepsis and Septic Shock. JAMA, 331(8), 665–674. https://doi.org/10.1001/jama.2024.0179