Table 2 Diagnostic accuracy of electronic nose analysis
From: Exhaled breath profiling for diagnosing acute respiratory distress syndrome
Cohort | Comparison | ROC-AUC | Specificity | Sensitivity |
|---|---|---|---|---|
Training | ARDS vs. Control | 0.72 (0.63-0.82) | 42% | 95% |
External validation | ARDS vs. Control | 0.71 (0.54-0.87) | 50% | 89% |
In-set: subgroup analysis | Moderate/severe ARDS vs. Control | 0.80 (0.70-0.90) | 62% | 91% |
| Â | Mild ARDS vs. Control | 0.67 (0.56-0.77) | 44% | 89% |
| Â | Moderate/severe ARDS vs. Mild ARDS | 0.69 (0.55-0.83) | 46% | 91% |
| Â | Pulmonary vs. non-pulmonary ARDS | 0.52 (0.37-0.67) | 74% | 40% |
In-set: competing diagnoses | CPE vs. ARDS | 0.65 (0.51-0.79) | 58% | 74% |
| Â | Pneumonia vs. ARDS | 0.69 (0.49-0.88) | 83% | 64% |
| Â | CPE vs. Moderate/severe ARDS | 0.76 (0.61-0.92) | 74% | 74% |
| Â | Pneumonia vs. Moderate/severe ARDS | 0.76 (0.57-0.95) | 91% | 64% |