11. Comparison of visual inspection and dermoscopy for the detection of any skin lesion requiring excision.
| Testa | Studies |
Lesions (cases) |
DOR (95% CI) |
Specificity at 80% sensitivity (95% CI) % |
Sensitivity at 80% specificity (95% CI) % |
| In‐person evaluations | |||||
| Visual inspection | 2 | 3457 (151) | 38.4 (2.5 to 582) |
91% (39 to 99) |
91% (39 to 99) |
| Visual inspection + dermoscopy |
4 | 3880 (260) | 232 (16.0 to 3354) |
98% (80 to 100) |
98% (80 to 100) |
| In‐person evaluations: direct studies | |||||
| Visual inspection | 2 | 3457 (151) | 15.0 (0.18 to 1225) |
79% (4 to 100) |
79% (4 to 100) |
| Visual inspection + dermoscopy |
2 | 3449 (137) | 88.1 (1.1 to 7338) |
96% (21 to 100) |
96% (21 to 100) |
| Image‐based evaluations | |||||
| Clinical (macro) images | 3 | 547 (138) |
21.7 (4.8 to 98.9) |
84% (54 to 96) |
84% (54 to 96) |
| Dermoscopic images | 5 | 815 (217) |
37.5 (8.8 to 161) |
90% (69 to 98) |
90% (69 to 98) |
| Image‐based evaluations: direct studies | |||||
| Clinical (macro) images | 3 | 547 (138) | 12.1 (5.4 to 26.7) |
75% (58 to 87) |
75% (58 to 87) |
| Dermoscopic images | 3 | 546 (136) |
18.4 (8.1 to 41.7) |
82% (67 to 91) |
82% (67 to 91) |
| CI: confidence interval; DOR: diagnostic odds ratio | |||||
aEstimates are based on fitting models with symmetric receiver operating characteristic (ROC) curves, and no formal comparisons between tests are made due to paucity of data. It is noted that the estimates for the visual inspection studies change between the all data and paired data analyses for both in‐person and image‐based analyses. This is driven by differences in the heterogeneity in accuracy between the models, which affects all parameters in the analyses.