Table 4.
BDI-II cut-points and associated test characteristics for optimization criteria for ROC analyses with specificity ⩾70%.
| Study | Sample, n (better/not) | Absolute change | Relative change | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Cut-point | Se | Sp | Correctly classified | AUC (95% CI) | Kappa | Cut-point | Se | Sp | Correctly classified | AUC (95% CI) | Kappa | ||
| GenPod | 240/88 | ⩾2 | 0.75 | 0.74 | 74.7% | 0.79 (0.7–0.8) | 0.43 | ⩾17% | 75 | 74 | 75% | 0.79 (0.74–0.85) | 0.44 |
| TREAD | 137/76 | ⩾1 | 0.68 | 0.71 | 69.0% | 0.76 (0.7–0.8) | 0.37 | ⩾18% | 61 | 86 | 70% | 0.76 (0.70–0.82) | 0.42 |
| CoBalT | 186/199 | ⩾ 3 | 0.62 | 0.71 | 66.5% | 0.75 (0.7–0.8) | 0.32 | ⩾32% | 65 | 86 | 76% | 0.78 (0.73–0.83) | 0.52 |
BDI-II, Beck Depression Inventory, 2nd edition; ROC, receiver operator characteristics; Se, sensitivity; Sp, specificity; AUC, area under curve; CI, confidence interval.
Cut-off values denote improvements in BDI-II scores required for optimal criterion for specificity ⩾70%. Youden Index calculated as J = max(sensitivity + specificity − 1). Kappa scores denote agreement between self-reported better/not and ROC classified better/not based on cut-off values.