Table 4. Performance results of the proposed diagnostic algorithm with different ACR score thresholds, T, for HIV-uninfected (N = 252) patients only.
| ACR threshold T | Selected for Xpert (%) | Sensitivity (%) | Specificity (%) | PPV (%) | NPV (%) | CSS ($) | CNTBC ($) |
|---|---|---|---|---|---|---|---|
| 0 | 100 | 93.9 | 99.1 | 93.9 | 99.1 | 13.09 | 106.41 |
| 24 | 90 | 93.9 | 99.1 | 93.9 | 99.1 | 13.25 | 107.72 |
| 42 | 80 | 93.9 | 99.5 | 96.9 | 99.1 | 11.95 | 97.16 |
| 50 | 70 | 93.9 | 99.5 | 96.9 | 99.1 | 10.60 | 86.19 |
| 67 | 60 | 90.9 | 99.5 | 96.8 | 98.6 | 9.30 | 78.15 |
| 76 | 50 | 90.9 | 99.5 | 96.8 | 98.6 | 8.01 | 67.24 |
| 85 | 40 | 87.9 | 100.0 | 100.0 | 98.2 | 6.71 | 58.28 |
| 93 | 30 | 78.8* | 100.0 | 100.0 | 96.9 | 5.41 | 52.41 |
| 96 | 20 | 69.7* | 100.0 | 100.0 | 95.6 | 4.11 | 45.02 |
The first row matches a scenario where all subjects undergo Xpert and no ACR. ACR, automated chest radiography; CNTBC, cost per notified TB case; CSS, cost per screened subject; NPV, negative predictive value; PPV, positive predictive value; Xpert, Xpert MTB/RIF.
*Sensitivity significantly different from Xpert standalone, T = 0. (p < 0.05 considered significant).