RATIONALE AND OBJECTIVES
Acidosis and hyperlactatemia predict outcome in critically ill patients. We assessed BE and pH for risk prediction capabilities in a sub-group of septic patients in the MIMIC-III database.
Associations with mortality were assessed by logistic regression analysis in 5586 septic patients. Baseline parameters, lactate concentrations, pH, and BE were analyzed at baseline and after 6 hours.
MEASUREMENTS AND MAIN RESULTS
We combined acidosis (defined as either BE ≤-6 and/or pH ≤7.3) and hyperlactatemia and split the cohort into three subgroups: low-risk (no acidosis and lactate <2.3 mmol/L; n = 2294), medium-risk (either acidosis or lactate >2.3 mmol/L; n = 2125) and high-risk (both acidosis and lactate >2.3 mmol/L; n = 1167). Mortality was 14%, 20% and 38% (p<0.001) in low-risk, medium-risk and high-risk patients, respectively. The predictiveness of this model (AUC 0.63 95%CI 0.61-0.65) was higher compared to acidosis (AUC 0.59 95%CI 0.57-0.61; p<0.001) and lactate >2.3 mmol/L (AUC 0.60 95%CI 0.58-0.62; p<0.001) alone. Hyperlactatemia alone was only moderately predictive for acidosis (AUC 0.60 95%CI 0.59-0.62).
Acidosis and hyperlactatemia can occur independently to a certain degree. Combining acidosis and hyperlactatemia in a model yielded higher predictiveness for ICU-mortality. Septic patients with acidosis should be treated even more aggressively in the future.