Table 2. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) performance compared with RT-PCR (reference) in acute Dengue infection during 0 to 7 fever days.
| Fever days | ||||||
|---|---|---|---|---|---|---|
| 0 | 1 | 2 | 3 | 4 | 5 | |
| NS1 | ||||||
| Pos./all No. | 25/31 | 187/195 | 83/87 | 35/36 | 12/12 | 9/9 |
| Sensitivity | 80.65 | 95.90 | 95.40 | 97.22 | 100 | 100 |
| (95% CI) | (62.53–92.55)a | (93.11–98.68)b | (91.00–99.80)b | (91.85–100.00)b | ||
| Pos./all No. | 14/23 | 53/63 | 19/25 | 10/17 | 3/6 | 1/7 |
| Specificity | 60.87 | 84.13 | 76 | 58.82 | 50 | 14.29 |
| (95% CI) | (38.54–80.29)a | (75.10–93.15)b | (54.87–90.64)a | (32.92–81.56)a | (11.81–88.19)a | (0.36–57.87)a |
| Pos./all No. | 25/34 | 187/197 | 83/89 | 35/42 | 12/15 | 9/15 |
| PPV | 73.53 | 94.92 | 93.26 | 83.33 | 80 | 60 |
| (95% CI) | (55.64–87.12)a | (91.86–97.99)b | (88.05–98.47)b | (72.06–94.60)a | (51.91–95.67)a | (32.29–83.66)a |
| Pos./all No. | 14/20 | 53/61 | 19/23 | 10/11 | 3/3 | 1/1 |
| NPV | 70.00 | 86.89 | 82.61 | 90.91 | 100 | 100 |
| (95% CI) | (49.92–90.08)a | (78.41–95.36)b | (61.22–95.05)a | (58.72–99.77)a | ||
| IgM | ||||||
| Pos./all No. | 11/31 | 9/195 | 2/87 | 3/36 | 4/12 | 8/9 |
| Sensitivity | 35.48 | 4.62 | 2.30 | 8.33 | 33.33 | 88.89 |
| (95% CI) | (19.23–54.63)a | (1.67–7.56)b | (0.00–5.45)b | (0.00–17.36)b | (9.92–65.11)a | (51.75–99.72)a |
| Pos./all No. | 16/23 | 48/63 | 16/25 | 9/17 | 3/6 | 1/7 |
| Specificity | 69.57 | 76.19 | 64 | 52.94 | 50 | 14.29 |
| (95% CI) | (47.08–86.79)a | (63.79–86.02)a | (42.52–82.03)a | (27.81–77.02)a | (11.81–88.19)a | (0.36–57.87)a |
| Pos./all No. | 11/18 | 9/24 | 2/11 | 3/11 | 4/7 | 9/15 |
| PPV | 61.11 | 37.50 | 18.18 | 27.27 | 57.14 | 60 |
| (95% CI) | (35.75–82.70)a | (18.80–59.41)a | (2.28–51.78)a | (6.02–60.97)a | (18.41–90.10)a | (32.29–83.66)a |
| Pos./all No. | 16/36 | 48/234 | 16/101 | 9/42 | 3/11 | 1/2 |
| NPV | 44.44 | 20.51 | 15.84 | 21.43 | 27.27 | 50 |
| (95% CI) | (27.94–61.90)a | (15.34–25.69)b | (8.72–22.96)b | (10.30–36.81)a | (6.02–60.97)a | (1.26–98.74)a |
| NS1&IgM | ||||||
| Pos./all No. | 28/31 | 187/195 | 83/87 | 35/36 | 12/12 | 9/9 |
| Sensitivity | 90.32 | 95.90 | 95.40 | 97.22 | 100 | 100 |
| (95% CI) | (62.53–92.55)a | (93.11–98.68)b | (91.00–99.80)b | (91.85–100.00)b | ||
| Pos./all No. | 13/23 | 47/63 | 15/25 | 8/17 | 3/6 | 1/7 |
| Specificity | 56.52 | 74.60 | 60 | 47.06 | 50 | 14.29 |
| (95% CI) | (34.49–76.81)a | (62.06–84.73)b | (38.67–78.87)a | (22.98–72.19)a | (11.81–88.19)a | (0.36–57.87)a |
| Pos./all No. | 28/38 | 187/197 | 83/93 | 35/44 | 12/15 | 9/15 |
| PPV | 73.68 | 94.92 | 89.25 | 79.55 | 80 | 60 |
| (95% CI) | (59.68–87.68)a | (91.86–97.99)b | (82.95–95.54)b | (64.70–90.20)a | (51.91–95.67)a | (32.29–83.66)a |
| Pos./all No. | 13/16 | 47/55 | 15/19 | 8/9 | 3/3 | 1/1 |
| NPV | 81.25 | 85.45 | 78.95 | 88.89 | 100 | 100 |
| (95% CI) | (54.35–95.95)a | (76.14–94.77)b | (54.43–93.95)a | (51.75–99.72)a | ||
| NS1&IgM&IgG | ||||||
| Pos./all No. | 28/31 | 188/195 | 83/87 | 35/36 | 12/12 | 9/9 |
| Sensitivity | 90.32 | 96.41 | 95.40 | 97.22 | 100 | 100 |
| (95% CI) | (62.53–92.55)a | (93.80–99.02)b | (91.00–99.80)b | (91.85–100.00)b | ||
| Pos./all No. | 13/23 | 43/63 | 14/25 | 8/17 | 3/6 | 0/7 |
| Specificity | 56.52 | 68.25 | 56 | 47.06 | 50 | |
| (95% CI) | (34.49–76.81)a | (55.31–79.42)a | (34.93–75.60)a | (22.98–72.19)a | (11.81–88.19)a | |
| Pos./all No. | 28/38 | 188/208 | 83/94 | 35/44 | 12/15 | 9/16 |
| PPV | 73.68 | 90.38 | 88.30 | 79.55 | 80 | 56.25 |
| (95% CI) | (59.68–87.68)a | (86.38–94.39)b | (81.80–94.80)b | (64.70–90.20)a | (51.91–95.67)a | (29.88–80.25)a |
| Pos./all No. | 13/16 | 43/50 | 14/18 | 8/9 | 3/3 | 0/0 |
| NPV | 81.25 | 86 | 77.78 | 88.89 | 100 | |
| (95% CI) | (54.35–95.95)a | (76.38–95.62)b | (52.36–93.59)a | (51.75–99.72)a | ||
a Exact confidence limits, in contrast, are based on the binomial distribution, and they have better converage in small samples and/or rare events.
b The confidence limits appearing under the asymptotic standard error (ASE) are based on asymptotic (Wald) theory. If n is large, then the 95% confidence interval can be calculated using p ± 1.96×ASE.