492 - Prospective external validation of the phoenix sepsis score and sepsis criteria

Publication Number: 492.6087

Elliot Long, The Royal Children's Hospital, Clifton Hill, Victoria, Australia; Meredith L. Borland, None on the list, Perth, Western Australia, Australia; Shane George, Gold Coast University Hospital, Gold Coast, Queensland, Australia; Shefali Jani, The Children's Hospital at Westmead(SCHN, Castle Hill, New South Wales, Australia; Eunicia Tan, The University of Auckland, Auckland, Auckland, New Zealand; Natalie Phillips, Queensland Children's Hospital, South Brisbane, Queensland, Australia; AMIT KOCHAR, WOMEN'S AND CHILDREN'S HOSPITAL AND UNIVERSITY OF ADELAIDE, ADELAIDE, South Australia, Australia; Simon Craig, Monash University, Clayton, Victoria, Australia; Anna Lithgow, The Royal Darwin Hospital, Casuarina, Northern Territory, Australia; Stuart Dalziel, University of Auckland, Auckland, Auckland, New Zealand; Ed Oakley, The Royal Children's Hospital, Melbourne, Victoria, Australia; Stephen Hearps, Murdoch Children's Research Institute, Parkville, Victoria, Australia; Ben Gelbart, The Royal Children's Hospital, Melbourne, Victoria, Australia; Scott L. Weiss, Nemours Children's Hospital, Wilmington, DE, United States; Nathan Kuppermann, UC Davis Health, Davis, CA, United States; Amanda Williams, MCRI/RCH, Parkville, Victoria, Australia; Franz E. Babl, MCRI/RCH, Parkville, Victoria, Australia

Background: The novel Phoenix Sepsis Score and criteria for sepsis diagnosis have been derived and validated in hospitalised children with infection using a large multi-country retrospective dataset.

Objective: The aim of this study was to externally validate the Phoenix Sepsis Score (area under the precision recall curve (AUPRC) 0.17) and criteria for sepsis diagnosis (sensitivity 56.5%, positive predictive value (PPV) 9,9%) in a prospectively identified cohort of children hospitalised with infection.

Design/Methods: Using data from the multi-centre multi-country Sepsis Epidemiology in Australian and New Zealand Emergency Departments (SENTINEL) study, we evaluated the Phoenix Sepsis Score as a predictor of in-hospital mortality (primary outcome) and the test characteristics of the Phoenix criteria for sepsis diagnosis for predicting in-hospital mortality and death or extra-corporeal life support (ECLS) within 72 hours (secondary outcomes).

Results: 6258 children hospitalised with suspected community acquired sepsis were included in the analysis (median age 2.1 years, in-hospital mortality 1.0%, death or ECLS within 72-hours 0.6%). The Phoenix Sepsis Score had an AUPRC of 0.17 (95%CI 0.09-0.26) for predicting in-hospital mortality, and 0.24 (95% CI 0.12-0.37) for predicting death or ECLS within 72-hours. Overall, 271 children met Phoenix criteria for sepsis diagnosis. Phoenix Criteria for sepsis diagnosis had a sensitivity of 53.3% (95%CI 40.0-66.3%) and PPV of 11.9% (95% CI 8.3%-16.3%) for predicting in-hospital mortality, and a sensitivity of 77.8% (95%CI 60.8-89.9%) and PPV of 10.4% (7.0-14.7%) for predicting death or ECLS within 72-hours.

Conclusion(s): In this prospective cohort of children, the test characteristics of the Phoenix Sepsis Score and criteria for sepsis diagnosis were similar to the original retrospective derivation and validation cohort with similar baseline mortality.

Figure 1. Consort diagram of included participants.
Fig 1 Consort Diagram.pdf

Table 1. Characteristics of participants from the SENTINEL and original Phoenix cohorts.
Table 1 PAS final.pdf

Table 2. Comparison of Phoenix Sepsis Score and Sepsis Criteria for predicting in- hospital mortality and death or ECLS within 72 hours between SENTINEL and original Phoenix cohorts.
Table 2 PAS final.pdf

Figure 1. Consort diagram of included participants.
Fig 1 Consort Diagram.pdf

Table 1. Characteristics of participants from the SENTINEL and original Phoenix cohorts.
Table 1 PAS final.pdf

Table 2. Comparison of Phoenix Sepsis Score and Sepsis Criteria for predicting in- hospital mortality and death or ECLS within 72 hours between SENTINEL and original Phoenix cohorts.
Table 2 PAS final.pdf