Diagnostic accuracy of initial blood lactate in predicting mortality in children with sepsis/suspected sepsis*
| Setting | Population | Overall mortality (%) | Lactate cut-off (mmol/L) | Relative risk of death | Sensitivity | Specificity | PPV | NPV | Mortality timeframe |
| ED (USA)9 | <18 years (n=1299) | 1.9 | >4 | 2.9 | 20% (12% to 28%) | 92.3% (90.7% to 93.7%) | 4.85% (2.23% to 10.26%) | 98.3% (97.97% to 98.62%) | 30 days |
| PICU admissions (Europe)13 | 1 month to 18 years (n=444) | 5.1 | >2.2 | 4.95 | 78.1% (56.3% to 92.5%) | 59.9% (55% to 64.6%) | 9.6% (7.7% to 12%) | 98.1% (95.9% to 99.1%) | In-hospital death or survival |
| Hospital admissions (East Africa)11 | 60 days to 12 years with severe non-malarial febrile illness (n=1283) | 12.4 | >5 | 5.82 | 71.7% (64 to 78.5%) | 75.5% (72.9% to 78%) | 29.3 (24.8% to 34.1%) | 95% (93.3% to 96.3%) | 72 h |
| Hospital admissions (Uganda)10 | <5 years with a diagnosis of pneumonia (n=155) | 14 | >3.4 | 4.5 | 68% (45% to 86%) | 74% (66% to 82%) | 31% (23% to 40%) | 93% (88% to 96%) | 5 days |
| PICU admissions (India)12 | 1 month to 12 years(n=148) | 63.5 | >4 | 1.8 | 57% | 82% | 84% | 51% | In-hospital death or survival |
*95% CIs provided where available in text or via author correspondence. PPV and NPV combine test accuracy and disease prevalence and as such can be misleading and can vary between patient groups, specifically where death rate PPV will appear poor even if the test is actually quite good.
NPV, negative predictive value; PPV, positive predictive value.