VALIDATION OF THE PEDIATRIC SEPSIS SCORE IN A DUTCH NATIONAL COHORT OF CHILDREN ADMITTED TO PICU WITH SEPSIS
- Navin P. Boeddha, Netherlands
- Navin P. Boeddha, Netherlands
- I. H. Visser,
- N. J. Jansen,
- Luregn Schlapbach, Australia
- . On Behalf Of The Skic (dutch Collaborative Picu Research Network),
Abstract
Background
Early recognition of high-risk sepsis patients is crucial for enrolment in trials, and to select patients for targeted therapies. The pediatric sepsis score is a prediction model for sepsis mortality, using a simple set of variables available within 60mins of ICU admission (Schlapbach LJ et al. Intensive Care Med.2017 Aug;43(8):1085-1096).
Objectives
We aimed to validate this score in an independent cohort.
Methods
Retrospective multicenter cohort study based on prospectively collected data from the Dutch Pediatric Intensive Care Evaluation (PICE) registry. We included patients <16 years admitted with sepsis and/or septic shock to PICU from 2003-2016. The pediatric sepsis score was calculated using variables available at PICU admission. Lactate data was not available. The primary outcome was mortality and secondary outcome was mortality and/or PICU length of stay (LOS) ≥72h.
Results
We included 1929 admissions for sepsis and septic shock (57% male; median age 2.2y; IQR 6m-8.2y), of which 257/1929 (13.3%) died and 1162/1929 (60.2%) admissions resulted in mortality and/or PICU LOS ≥72h. Both primary and secondary outcome strongly correlated with the sepsis score (p<0.001). The pediatric sepsis score predicted mortality with an AUC of 0.698 (95%-CI 0.677-0.718) and mortality/PICU LOS ≥72h with an AUC of 0.711 (95%-CI 0.690-0.731).
Conclusion
Validation of the pediatric sepsis score in an independent dataset demonstrates that it is a robust tool to predict mortality, and mortality and/or prolonged ICU stay in children admitted to PICU with sepsis. The lack of lactate data in the validation dataset may have contributed to the lower predictive performance compared to the original study.