Prediction of pediatric sepsis mortality within 1 h of intensive care admission

Purpose: The definitions of sepsis and septic shock have recently been revised in adults, but contemporary data are needed to inform similar approaches in children. Methods: Multicenter cohort study including children <16 years admitted with sepsis or septic shock to ICUs in Australia and New Zealand in the period 2012-2015. We assessed septic shock criteria at ICU admission to define sepsis severity, using 30-day mortality as outcome. Through multivariable logistic regression, a pediatric sepsis score was derived using variables available within 60 min of ICU admission. Results: Of 42,523 pediatric admissions, 4403 children were admitted with invasive infection, including 1697 diagnosed as having sepsis/septic shock on admission. Mortality was 8.5% (144/1697) and 50.7% of deaths occurred within 48 h of admission. The presence of septic shock as defined by the 2005 consensus was sensitive but not specific in predicting mortality (AUC = 0.69; 95% CI 0.65-0.72). Combinations of hypotension, vasopressor therapy, and lactate >2 mmol/l discriminated poorly (AUC < 0.60). Multivariate models showed that oxygenation markers, ventilatory support, hypotension, cardiac arrest, serum lactate, pupil responsiveness, and immunosuppression were the best-performing predictors (0.843; 0.811-0.875). We derived a pediatric sepsis score (0.817; 0.779-0.855), and every one-point increase was associated with a 28.5% (23.8-33.2%) increase in the odds of death. Children with a score >= 6 had 19.8% mortality and accounted for 74.3% of deaths. The sepsis score performed comparably when applied to all children admitted with invasive infection (0.810; 0.781-0.840). Conclusions: We observed mortality patterns specific to pediatric sepsis that support the need for specialized definitions of sepsis severity in children. We demonstrated the importance of lactate, cardiovascular, and respiratory derangements at ICU admission for the identification of children with substantially higher risk of sepsis mortality.