Table 4.
Predictive accuracy of mortality of sepsis definitions with or without lactate value
| Sepsis-2 definition | Sepsis-3 definition | |||
|---|---|---|---|---|
| AUC (CI 95%) | Asymptotic P value | AUC (CI 95%) | Asymptotic P value | |
| Without lactate model | 0.650 (0.619–0.680) | <0.001 | 0.615 (0.580–0.650) | <0.001 |
| Lactate as a continuous variablea | 0.704 (0.668–0.739)b | <0.001 | 0.649 (0.608–0.689)b, d | <0.001 |
| Lactate > 2 mmol/L | 0.664 (0.631–0.697) | <0.001 | 0.620 (0.584–0.656) | <0.001 |
| Lactate > 4 mmol/L | 0.686 (0.653–0.719)b, c, d | <0.001 | 0.636 (0.598–0.674)b, c, d | <0.001 |
The receiver operating curves (ROC) were constructed using the binary logistic regression probabilities of death using sepsis definitions and lactate categories. AUC area under the ROC curve
aLactate as a continuous variable was logarithmically transformed in order to correct the positive skewness. Lactate > 2 mmol/L represents all patients with lactate > 2 mmol/L, including those with lactate > 4 mmol/L
bDeLong’s test P < 0.05 versus without lactate model
cDeLong’s test P = ns versus lactate as a continuous variable
dDeLong’s test P < 0.05 versus lactate > 2 mmol/L