Table 2.
Results of the REMS accuracy evaluations for each considered anatomical site. In all the calculations, DXA results obtained for the retained patients after the exclusion of medical reports containing errors were assumed as the reference ground truth
| Anatomical site | Lumbar spine | Femoral neck |
|---|---|---|
| Retained cases (n) | 508 | 512 |
| Sensitivity | 90.1% | 90.9% |
| Specificity | 93.6% | 94.6% |
| Diagnostic concordance | 82.7% | 81.8% |
| K | 0.71* | 0.71* |
| r | 0.91* | 0.90* |
| r 2 | 0.83* | 0.82* |
| Regression line slope | 0.90 | 0.90 |
| SEE (g/cm2) | 0.045 | 0.031 |
| Average difference (bias ± 1.96 SD, g/cm2) | −0.06 ± 0.60 | −0.07 ± 0.44 |
*p < 0.0001