6471 Performance of eight different severity scores to predict critical care admission in paediatric inpatients treated for sepsis: a retrospective cohort study

Objectives Sepsis is a leading cause of deterioration leading to unplanned transfer to critical care (UTCC) in hospitalised children, but there are few published studies evaluating the predictive performance of severity scores outside critical care. We aimed to compare the predictive performance of eight different severity scores in predicting UTCC and multiple organ dysfunction syndrome (MODS) in hospitalised children treated for sepsis. Methods Eligible inpatients who were treated for sepsis before having UTCCs were identified prospectively between 1 March 2018 to 28 February 2019, each patient was matched with two controls who did not transfer to critical care or have sepsis. The worst Alder Hey PEWS within 24 hours of UTCC were collected from the electronic patient record. National PEWS (Paediatric Early Warning Score), pSOFA (paediatric Sequential Organ Failure Assessment), qSOFA (quick Sequential Organ Failure Assessment), LqSOFA (Liverpool quick Sequential Organ Failure Assessment), P-MODS (Pediatric Multiple Organ Dysfunction Score), PELOD-2 (Paediatric Logistic Organ Dysfunction-2) and PODIUM (Paediatric Organ Dysfunction Information Update Mandate) criteria were calculated using the worst observations and laboratory results in the last 24 hours preceding UTCC. The primary outcomes were UTCC and MODS. The secondary outcomes were sepsis-related mortality, critical care (CC) stay>=5 days and invasive ventilation duration>2 days. Results A total of 75 UTCCs occurred in children treated for sepsis and 31 (41%) had MODS. Two (3%) children had sepsis-related mortality, 45 (60%) had CC stay>=5 days and 24 (32%) had invasive ventilation >2 days. PEWS had excellent predictive performances for UTCCs (AUCs 0.86–0.91) and PODIUM criteria and PELOD-2 showed superior performance for MODS (AUCs 0.72–0.79). Conclusion Overall performance of the severity and organ dysfunction scores were good in hospitalised children treated for sepsis with UTCC. Further validation is needed in hospitalised children who experience UTCCs from other causes.