Table 3.
Performance of the automated classifiers with different testing sets/media.
| Agar Medium | Parameter | Classifier S. Paratyphi A |
Classifier S. Paratyphi B |
Classifier S. Paratyphi C |
Classifier S. Typhi |
|---|---|---|---|---|---|
| CBA | Accuracy | 99.5% | 87.0% | 99% | 99.9% |
| Error rate | 0 | 10.8% | 0 | 0 | |
| Failed classification | 0.5% | 2.2% | 1% | 0.1% | |
| Sensitivity | 100% | 77.8% | 100% | 100% | |
| Specificity | 100% | 94.8% | 100% | 100% | |
| Positive predictive value # | 100% | 77.8% | 100% | 100% | |
| Negative predictive value # | 100% | 83.1% | 100% | 100% | |
| TSA | Accuracy | 97.0% | 87.8% | 95.5% | 99.6% |
| Error rate | 1% | 10.1% | 2.8% | 0 | |
| Failed classification | 2% | 2.1% | 1.7% | 0.4% | |
| CHO | Accuracy | 100% | 86.6% | 98.8% | 99.1% |
| Error rate | 0 | 12.0% | 1.2% | 0 | |
| Failed classification | 0 | 1.4% | 0 | 0.9% | |
| MHA | Accuracy | 94% | 84.9% | 94.7% | - |
| Error rate | 0 | 11.7% | 0 | - | |
| Failed classification | 6% | 3.4% | 5.3% | - |
# positive and negative predictive values are largely affected by prevalence, so these values may vary in other sample collections.