Table 5. Estimated sensitivity, specificity and accuracy of staging of HAT diagnostic algorithms (worst-case scenario).
| Accuracy indicator | Gamboma, Mossaka, Nkayi, Republic of Congo | Kiri, Sudan (old algorithm) | Kiri, Sudan (new algorithm) | Adjumani, Uganda | Arua/Yumbe, Uganda |
| Screening and confirmation of infection, assuming perfect follow-up of suspects † | |||||
| Sensitivity if case is stage 1 (%) | 91.7 (83.5–97.1) | QBC: 81.8 (76.6–89.8)CTC: 80.7 (70.5–88.9) | QBC: 59.6 (47.7–72.5)CTC: 58.7 (46.2–72.1) | 74.3 (61.1–86.3) | 75.4 (62.8–86.2) |
| Sensitivity if case is stage 2 (%) | 91.7 (83.5–97.1) | QBC: 80.6 (70.4–88.7)CTC: 79.0 (69.0–87.6) | QBC: 59.0 (47.4–71.9)CTC: 58.1 (46.1–71.2) | 75.6 (63.1–86.7) | 76.1 (63.5–86.9) |
| Specificity (%) | 97.8 (97.1–99.5) | QBC: 99.6 (99.0–99.9)CTC: 99.6 (99.0–99.9) | QBC: 99.9 (99.8–100.0)CTC: 99.9 (99.8–100.0) | 99.7 (99.2–100.0) | 99.7 (99.2–100.0) |
| Screening and confirmation of infection, assuming no follow-up of suspects | |||||
| Sensitivity if case is stage 1 (%) | 85.2 (76.7–91.4) | QBC: 71.1 (60.0–81.7)CTC: 64.3 (52.3–77.0) | QBC: 53.3 (41.5–67.3)CTC: 49.2 (36.9–64.3) | 67.8 (55.6–80.3) | 69.5 (56.7–81.3) |
| Sensitivity if case is stage 2 (%) | 85.2 (76.7–91.4) | QBC: 77.8 (67.4–86.6)CTC: 74.7 (64.1–84.3) | QBC: 57.4 (45.8–70.5)CTC: 55.6 (43.5–69.1) | 73.7 (61.6–85.0) | 74.6 (61.9–85.6) |
| Specificity (%) | 97.8 (97.3–99.5) | QBC: 99.6 (99.1–99.9)CTC: 99.6 (99.1–99.9) | QBC: 99.9 (99.8–100.0)CTC: 99.9 (99.8–100.0) | 99.7 (99.2–100.0) | 99.7 (99.2–100.0) |
| Accuracy of staging | |||||
| Probability of being correctly classified as stage 1 (%) | 63.6 (44.6–74.1) | QBC: 47.5 (30.5–62.4)CTC: 45.8 (29.4–60.4) | QBC: 47.8 (31.1–62.9)CTC: 46.6 (30.1–61.3) | 47.5 (30.7–62.3) | 48.0 (31.0–63.0) |
| Probability of being correctly classified as stage 2 (%) | 89.2 (78.9–94.8) | QBC: 93.1 (83.0–97.8)CTC: 93.7 (84.4–98.0) | QBC: 93.0 (82.7–97.8)CTC: 93.5 (83.6–98.0) | 92.9 (82.8–97.8) | 92.8 (82.3–97.8) |
†: Four follow-up visits at three month intervals in all projects except for Adjumani (one visit at three months only). Under the new Kiri algorithm, suspect follow-up only occurred if the village has an observed prevalence >2%.
Values in parentheses are 95% percentile intervals.