Predictors of mortality and multiple organ failure in children with sepsis

Objectives: To assess the markers of perfusion which best discriminate survivors from non-survivors of childhood sepsis and to compare the information derived from gastric tonometry with conventionally measured haemodynamic and laboratory parameters. Design: Prospective clinical study of children with sepsis syndrome or septic shock. Setting: Paediatric intensive care unit in a tertiary referral centre. Patients: 31 children with sepsis syndrome or septic shock. Interventions: A tonometer was passed into the stomach via the orogastric route. Measurements and main results: The following data were recorded at admission, 12, 24 and 48 h: heart rate, mean arterial pressure, arterial pH, base deficit, arterial lactate, gastric intramucosal pH (pHi) and DCO2 (intramucosal carbon dioxide tension minus arterial partial pressure of carbon dioxide). The principal outcome measure was survival. The secondary outcome measure was the number of organ systems failing at 48 h after admission. There were 10 deaths and 21 survivors. No variable discriminated survival from death at presentation. Blood lactate level was the earliest discriminator of survival. Using univariate logistic regression, lactate discriminated survivors from those who died at 13 and 24 h after admission, but not at 48 h (p = 0.019, 0.044 and 0.062, respectively). The area under the receiver operating characteristic (ROC) curve for lactate was 0.81, 0.88 and 0.89 at 12, 24 and 45 h, respectively. At 12 h after admission, a blood lactate level > 3 mmol/l had a positive predictive value for death of 56 % and a lactate level of 3 mmol/l or less had a positive predictive value for survival of 84 %. At 23 h a lactate level > 3 mmol/l had a positive predictive value for death of 71 % and a level of 3 mmol/l or less had a positive predictive value for survival of 86 %. No other variable identified non-survivors from survivors at 12 h, Gastric tonometry could only be done on 19 of the 31 children, of whom 8 died and 11 survived. In these 19 children, DCO2 measured at 24 h, but not at 12 or 38 h, distinguished those who died from those who survived (p = 0.045 and p = 0.20, respectively). The area under the ROC curve for DCO2 measured at 24 h as a predictor of survival was 0.71. Neither the absolute value of pHi nor the trend of change in pHi at any time in the first 38 h identified survivors in this series. The mean arterial pressure distinguished survivors from non-survivors at 24 and 48 h (area under ROC curve = 0.80 and 0.78, respectively). The base deficit and heart rate did not identify non-survivors from survivors at any time in the first 48 h. Conclusions: Blood lactate level was the earliest predictor of outcome in children with sepsis. In this group of patients, gastric tonometry added little to the clinical information that could be derived more simply by other means.