Table 2.
ROC analysis results
| Indicator | Value |
|---|---|
| AUC | 0.937 (0.867–0.977) |
| Optimal threshold | 93.7 mg/cm3 |
| Sensitivity | 85.45% (47/55) |
| [73.3–93.5%] | |
| Specificity | 89.19% (33/37) |
| [74.6–97%] | |
| PPV | 92.2% (45/49) |
| [82.2–96.8%] | |
| NPV | 80.5% (33/41) |
| [63.3–88.8%] |
ROC curve analysis yields an optimal patient-level cut-off value of 93.7 mg/cm3 which shows high diagnostic accuracy for the detection of follow-up fractures in a 2-year interval (p < .001). Numbers in square brackets are confidence intervals. ROC, receiver operating characteristic; AUC, area under the curve; PPV, positive predictive value; NPV, negative predictive value