Table 1.
Performance of the artificial intelligence algorithm for predicting rapid diagnostic test (RDT) results with respect to human visual reading in (1) the validation set, (2) the field study for reading antibody (Ab) RDTs, and (3) in the field study when reading antigen (Ag) RDTs.
| Evaluation data | AUCa (95% CI) | Sensitivity (95% CI) | Specificity (95% CI) | Tests, n | ||||||||
| Negative | Positive | |||||||||||
| 1: RDT manufacturer (Ab) and band | ||||||||||||
|
|
Abbott | |||||||||||
|
|
|
IgG | 99.5 (98.7-100) | 96.4 (94.1-98.8) | 100 (100-100) | 94 | 145 | |||||
|
|
|
IgM | 92.5 (85.4-99.6) | 80.8 (75.8-85.8) | 90.7 (87.0-94.3) | 184 | 55 | |||||
|
|
UNScience | |||||||||||
|
|
|
IgG | 100 (100-100) | 100 (100-100) | 100 (100-100) | 100 | 140 | |||||
|
|
|
IgM | 89.5 (83.7-95.2) | 80.0 (74.9-85.1) | 88.6 (84.6-92.6) | 214 | 26 | |||||
|
|
AllTest | |||||||||||
|
|
|
IgG | 99.8 (99.4-100) | 97.9 (96.1-99.7) | 100 (100-100) | 96 | 144 | |||||
|
|
|
IgM | 90.6 (85.0-96.1) | 79.6 (74.5-84.7) | 86.0 (81.6-90.4) | 186 | 54 | |||||
|
|
Global | |||||||||||
|
|
|
IgG | 99.8 (99.5-100) | 98.1 (97.1-99.1) | 100 (100-100) | 290 | 429 | |||||
|
|
|
IgM | 90.8 (87.4-94.3) | 80.0 (77.1-82.9) | 89.0 (86.2-90.9) | 584 | 135 | |||||
| 2: RDT manufacturer (Ab) and band | ||||||||||||
|
|
Roche | |||||||||||
|
|
|
IgG | 100 (100-100) | 100 (100-100) | 94.4 (92.8-96.1) | 18 | 154 | |||||
|
|
|
IgM | 99.6 (96.0-100) | 100 (100-100) | 95.8 (94.3-97.3) | 166 | 6 | |||||
| 3: RDT manufacturer (Ag) and band | ||||||||||||
|
|
Abbott | |||||||||||
|
|
|
Test | 100 (100,100) | 100 (100,100) | 100 (100,100) | 68 | 28 | |||||
aAUC: area under the curve.